EP 164 - Kat Kozyrytska: AI in Pharma Is a Yesterday Problem – Why Ethical Frameworks Can’t Wait

Interactive Transcript

spk_0 Artificial intelligence is already inside your company, whether you know it or not.

spk_0 Employees are feeling confidential data into tools like chat-truity, chasing efficiency, without realizing the risks.

spk_0 And here is the hard truth.

spk_0 Protecting your company from AI misuse isn't a tomorrow problem.

spk_0 It's already a yesterday problem.

spk_0 Because what happens when those hidden queries leak your intellectual property to competitors?

spk_0 Or when algorithms confidantly give you the wrong answer, and you make decisions and even worse, act on it?

spk_0 The efficiency may feel irresistible, but the cost of blind trust could be devastating.

spk_0 People working in the space are already using AI, right?

spk_0 It's the temptation of getting an answer fast out of your spreadsheet is...

spk_0 That Kosyri Diska has lived this tension from the inside, from winning the Merck Prize at MIT to neuroscience research in a Nobel laureate lab at Stanford.

spk_0 To leading Nobel biotextradity at Demo Fisher and Satorios, she has seen how fragile trust can be when technology collides with human lives.

spk_0 And today is co-founder of Collings and the Cell Therapy Manufacturability Program, she's building frameworks for confidential collaboration, where artificial intelligence accelerates breakthroughs without sacrificing privacy or ethics.

spk_0 But if you are working on some obscure target, this is your super highly confidential IP.

spk_0 You want to get some intelligence about the research, for example.

spk_0 In this conversation, cat and bags, house eos, investors and innovators can turn artificial intelligence from a reckless risk into a tool that embeds values, safeguards IP, and even amplifies humanity's best instincts.

spk_0 If we know that a personalized therapy will work better for the patient, it is almost our obligation.

spk_0 Believe me, this episode is not just about artificial intelligence, it's about responsibility and decision making.

spk_0 It's about how leaders can harness technology to build a future where therapies are safer, more personal and more ethical.

spk_0 But before we dive in, let me ask you one thing.

spk_0 I've looked at the numbers today and found out that only about 20% of you who listen regularly actually follow the show.

spk_0 If this episode sparks your curiosity, please do me a favor. Hit that follow button.

spk_0 It helps me grow the show and bring you more conversations like this with rare, deep-tech, leaders entrepreneurs and venture capitalists.

spk_0 And tell you for free to you. And now let's get started.

spk_0 Then let's go online and talk a little bit about dark personalities, which is a great. How did you come up with this term?

spk_0 Why didn't this is an industry term?

spk_0 Oh, really it is.

spk_0 Yeah, industry.

spk_0 I'm telling you, we have so much research on this.

spk_0 We already know.

spk_0 You have an impressive background. Thank you very much for your detailed preparation material.

spk_0 I'm just for the audience, you started biology at the MIT in Boston.

spk_0 How about how about that?

spk_0 Well, I thought I was good at math. So I started on math.

spk_0 And then I met people who were good at math and then so I made me physics is for me.

spk_0 So I kept moving through engineering and chemical engineering.

spk_0 And then eventually I found biology and it just seemed like such an opportunity to model and learn in this kind of physics, rational based approach.

spk_0 And it really felt like at that point we weren't really there. I mean, I would argue we're still not really there.

spk_0 I think we're we have a vision. We can now we have such an explosion of these physics based cell models.

spk_0 So I think we're finally getting to this place of understanding.

spk_0 So I think it was a good choice in the end.

spk_0 Yeah, I believe that and I read in your material, you earned the mark price for outstanding research.

spk_0 Yes, I was in this great lab that for me, I think it was a real preview of this fusion of biology with core mechanism understanding.

spk_0 So as I'm Kathy Drennan's lab working on crystallography, protein engineering, looking at modifying active sites.

spk_0 You could look at it at the very granular level of electron density. You could see where the electrons were moving as the binding occurred.

spk_0 And it was it was just so beautiful.

spk_0 And you know, you have the 80% training set the 20% test set. So it was kind of already a foundation for for me for understanding how we how we train models.

spk_0 Reminds me of the theories big band theory because there was this one guy from MIT, the MIT engineer and you pretty much replayed it in your life because you started at MIT and then decided to move to the west coast.

spk_0 Yes, to Stanford, but how was the cultural shock for you from East Coast to West Coast?

spk_0 Yes, so these are very different schools and I was teaching while I was at MIT and then I was teaching again at Stanford and just the different approaches with the undergraduate students I found to be fascinating.

spk_0 So, you know, MIT, we have some good weather, but most of the year it's bad weather. You're in the library, at least back in the day, you were in the library.

spk_0 And Stanford, it's sunny, the vast majority of the year. So you walk out of the classroom and you kind of lie down on the grass with your device or your textbook and it was shocking to me to see students outside enjoying themselves enjoying the weather is unthinkable, but completely different approach.

spk_0 I think the sun definitely helps in general for species to live the spirit. So, yeah, I think from that perspective, it's a good place, California.

spk_0 Yeah, absolutely. I mean, it's so fascinating what you said, MIT Boston, Krawliwater, Rainy weather, not much sunshine. It's perfect for studying. And then I always wondered in California, I mean, you have the ocean, you have the beach, surfing, music, reminds me of Baywatch in the 80s, 90s.

spk_0 This year is, which is just on Amazon Prime. How, how, how, how do people do that in California? Did they say no, I go to the to the lab and get my work done and rather than enjoying the sun?

spk_0 So it's Stanford, there's a kind of an expression or hypothesis that it's all this kind of duck approach where you pedal very hard beneath the surface where it's invisible, but you're working really, really hard and then from the surface, it's, it's, it's all very still and peaceful.

spk_0 And I think it's that internal drive to get things, especially in science, right? I mean, we don't get paid a whole lot of scientists. So it's clearly not for some external motivation that we're doing things is really that the wish to explore, find answers, make this world a better place, save lives.

spk_0 Some of those things, I think, drive a tremendous amount of progress that happens in a sunny location at an academic institution.

spk_0 That's fantastic. So you have specs on mathematics, biology and standard in your material, you then worked in neuroscience in a noble laureate slab. How was that?

spk_0 Oh, it was fantastic. I mean, he, he's told me Tom Zudov, he got his Nobel Prize in 2013 and he is, his brain is definitely the fastest that I haven't counted.

spk_0 And I mean, I think of myself as quite sharp, but talking to him was so way so many steps and sentences ahead. I remember just in these conversations, trying to catch up, like I sort of see how we got from here to here.

spk_0 Okay, yeah. But he's just so bright. He sees the big picture. He's a very molecular person, but I also think about very, let's say abstract ideas. And so I love the lab because he was, he built it to be everything from crystal on or if you all the way through to animal behavior with some profound implications for human disease.

spk_0 And so it's, it's the spectrum that can be kind of horizontal and vertical nature of his enterprise. It's a huge lab, 70 people, right? So you can get a lot of things done with 70 people, but it's just such an explosion of ideas. I am hoping to say it's an anniversary of his lab this year.

spk_0 So I'm headed to the symposium to reconnect with with some of these bright minds and bring some of the thoughts that we're going to cover today to that community to get their feedback.

spk_0 That's fantastic. So from possible back to, you know, impossible, I guess, and then back to sunny California, enjoying the, the travel or even the whole movie on packing up my things into a few boxes.

spk_0 This is a fact who could possibly resist that? So you moved from mathematics to biology from biology to neuroscience and now landed in artificial intelligence and ethics. What's part you're interested in that area?

spk_0 I mean, it's sort of a part of my natural journey. I think for me, when I think about understanding something, it's really understanding it to that molecular level. And so that's why neuroscience has always been so fascinating because you observe the high level of human behavior.

spk_0 And through my unfortunate many ways experience, I've observed a range of human behavior. So I think for me, it's trying to understand how does that work molecularly.

spk_0 And so neuroscience is tempting because we're on the cost of molecular understanding, but not really quite there yet. I don't know if we will be, I hope that we will be, but it's getting these answers at that protein level or atomic level resolution to the very big.

spk_0 Picture things.

spk_0 That's fantastic. In your material, there was another interesting point. You said that there is a lot of yesterday's problems in artificial intelligence. What do you mean with that?

spk_0 Especially in life sciences, we kind of think of this, you know, in a few years, we'll get to artificial intelligence eventually will collect the large enough data sets. Oh, we have to go find this collection of et cetera, et cetera.

spk_0 But the truth is that people working in the space are already using AI, right? It's the temptation of getting an answer fast out of your spreadsheet is so vast, nobody can resist it really.

spk_0 I mean, a few people do, but vastly it's I want this answer and I'm not going to spend the five hours on it. I can plug it into chat GPT and then here it is, you know, the response.

spk_0 So I think from that perspective, it's we have to face the reality that, you know, employee education and employee adherence to your policies has to become a priority because people will and this has already happened in the industry right people will put confidential documents into chat GPT.

spk_0 And so, you know, I think it's I would argue not with malicious intent, but it's due to the lack of understanding of the profound impact of sharing that kind of information with an AI tool. So yes, it cuts some corners and makes your days smoother, shorter and so on. But the big picture of the impact of that is, yeah, we have we really have to be thoughtful about how we guide employees in using AI based tools.

spk_0 And I would argue that you're not offering an alternative that isn't up for this with your compliance policy governance and so on.

spk_0 They're still going to use it. It's just not going to be according to your rules.

spk_0 Yeah, very good.

spk_0 And the end of the day is really an interesting part when we take this conversation into a cafe and I asked in the next two hours, what's the big idea you want the world to understand right now, what would you say?

spk_0 Stay with us will be right back.

spk_0 Hey, listen, it's me, your podcast host, Christian Soshna here. Have you ever felt the need to elevate your strategic vision, refine your execution or master the art of leadership, now's your chance.

spk_0 Dive deep with me in exclusive coaching sessions tailored just for you, whether it's a power packed 30 minutes session and intensive 60 minute deep dive or a refreshing walk and talk in Vienna, I'm here to guide you.

spk_0 We for over 35 years of business experience, let's unlock your potential together. Visit the link in the show notes and book your session today. Let's make strategy come alive.

spk_0 Yeah, I think the big picture of this is that artificial intelligence is a great way to scale throughput amplify right bring efficiency and so on.

spk_0 But I think what I would love for us to pause and think about what is it that we want to amplify what is it that we're high group, because we have studied ourselves as a species, we understand there's a range of behaviors, there's the good behaviors, those bad behaviors, I would argue that we should really focus on amplifying the good behaviors and try our very best to slow down or stop the amplification of the bad behaviors.

spk_0 Right, because it kind of you think about this from the extreme view, if you used to have those people in Craigslist, right, who they tried to get your check and so on. Now they can do that at high throughput, because you no longer have to have the human behind it.

spk_0 So how can we put some guard rails in place and safeguards to make sure that that that's quite down and all of the goodness in this world that takes more effort and so on that we bring efficiency to that that we do better more ethical things.

spk_0 Isn't that an unusual thinking for a scientist when we look on LinkedIn.

spk_0 In the discussion is mostly driven by science and they read a lot about we have a new target AI makes finding molecules so much easier and we can simulate everything so it's a lot of tech and features and you come from a different angle, say no, it's all about ethics and moral.

spk_0 How did you come to the conclusion that we need a different discussion in this space, not just tech driven, but above it.

spk_0 I will say that I'm also technologist and a scientist so I want answers I can't wait for us to have the answers. For example, you know, part of my work is in southeast manufacture ability.

spk_0 I think if we could get the answer that faster, that'd be amazing.

spk_0 But I think so much of the conversation like you said is driven by people who are really focused on tech or people who really have that question about science.

spk_0 If you just take a step back and think about the impact of some of these answers that we're going to get, we can think about that from the perspective of reliability right so you know we implement the tech but how reliable is what we're getting out of it.

spk_0 I think there's a gap on the implementation side that we're often not asking the question is this good or not.

spk_0 And I think for more conventional technologies, right, if we think about a capital equipment, if you have a box and you ship it and the bolts fall out, it's very obvious that it's not good or if you have a microscope that's not aligned very clear.

spk_0 There's no there's no question about that. I think with software and use gen AI, especially it's just harder to evaluate.

spk_0 So you have to think about that upfront. And then the other side of it is on the patient's right side.

spk_0 So we can get into more depth on this, but there are so many studies that show that that that the identification, anonymization of data is really a myth.

spk_0 So when you ask a question of a data set and we're going to bring up 23 me right because I've been trying to get my data set out of that for the longest time and it just hasn't worked yet.

spk_0 But some of the answers that you're going to get will impact real humans on the other end.

spk_0 So we might be studying clinical impact or even manufacture ability like in South Africa's right, I believe that we would get tremendous insights if we pulled in clinical outcomes data.

spk_0 But that's going to reveal something about those humans. And if a certain entity, so in the US, a big player to this is health insurance.

spk_0 If for maybe your auto insurance, so one right if they get that insight somehow because you know things get get uploaded into AI all the time.

spk_0 But that can impact their strategy for pricing for insurance. It's just one example, right. So now there's real financial impact for the individuals who are within that data set.

spk_0 And I think we have to think about that. We as much as we want to get to a scientific answer, we have to think about the humans involved.

spk_0 That's a good point. Let me challenge you on this. So I'm now I'm now proud to say that I'm 51 in my 52nd year on this planet. And I can now say to you young woman.

spk_0 I'm living in Europe. And this regulation is driving me crazy. I experienced the internet and we are just fine at the end of the day. So we can play around. We should experiment at the end of the day. Nothing serious happened. And Europe is thriving the US is thriving. And we really have a great time with technology. And we need it all these mistakes that we made to come to this point where we are now. And it's a great thing.

spk_0 And we are upstream. So the West evolved in a good way. Global poverty went down in also other countries and technologies are blessing for humans.

spk_0 When it look now at European Union, we want to regulate everything which leads to the position that AI in Europe is in that and development accelerated in the US and also in China. Interestingly, communist country doing better than Europe.

spk_0 Now what you propose what you understood is no, we need to be more careful with artificial intelligence. So how would you convince someone like me who says completely arrogantly young lady. I have experienced the last 30 years. And we will be fine in future.

spk_0 I should be more careful with that.

spk_0 Yeah, so maybe you want to see some to become more worried about the topic of human rights, the patient rights and so on.

spk_0 Equality based on another big one. You want to see some massive full out. You want to see a big impact on someone that that's what we're assuming.

spk_0 And we're not going to be able to see some of the human rights. But the our neurobiology we actually need extreme.

spk_0 But with some of the impact is with AI, especially it's less obvious, less visible. And you may in fact never see it.

spk_0 Especially because some of the models are less transparent and so on. Right. So I think an example to bring up here is with mortgage eligibility assessments.

spk_0 You will never know why somebody truly got rejected by an algorithm. But it might be because the algorithm was built on some historical data where we have historically discriminated against the particular category.

spk_0 That category is probably not you know the ruling class. So they will have you even less of a voice in expressing that. But I think it's that if we if we want to build the future to be better at the very core of this.

spk_0 If we want to build a future that's better than the past. We have to think about this. I know it's it's less you know dollar driven.

spk_0 But in some ways, I think the implementation of AI and the development of this technology is an opportunity for us to reset.

spk_0 And again, I amplify the better things that we do.

spk_0 The way I think about it is that we're really here every one of us for such a split second. So yes, I understand the cat we need cash to survive. I get that I live in Boston. So I understand this in the profound way.

spk_0 But it's also that you can't really take cash with you after you die. I think that's very clear.

spk_0 So why are you here for the split second? Do you want to leave this place a better place than when you came here? So I think fundamentally it's that and I think it's just such a chance for us to do better.

spk_0 Yeah, let's hope that when you're not going to have an store that someone doesn't ask you to credit card and show me a bit of a wallet.

spk_0 We never know. But let's go back to more serious things. I mean, but this is a this is a huge problem at the end of the day because we need data to use artificial intelligence to improve the human human society's across the globe.

spk_0 On the other hand, it also puts us at risk. How can we do it? I mentioned that's just to think about creating new therapeutics.

spk_0 Scientists always complain that they can't access data. There is so many data around fake clinical trials with the beneficial for scientists, but this data sets connected to human beings and you have the names and you have the genetic profile sometimes in there and all that health history.

spk_0 But there is also the danger of how can we deal with that in your opinion, but what's the solution for that problem?

spk_0 So I think to your point, the rare disease is a really interesting example of that right where it's if you have 17 patients around the world, you're probably going to know which role and the table is who I'm.

spk_0 It's fairly straightforward to find out. But in fact, because the options are very limited for disease, there's a massive drive on the patients that they want they they're happy to donate whatever data you want their mother's maiden name.

spk_0 Yes, have that if that's going to help with the therapy, right? So talking to also pediatric cancers, for example, if you talk to the parents, they will say we will give anything our own the child's data and so on.

spk_0 Because we want the answer. So I think it's that balance of the benefit versus the cost and the risks involved, right? And I mean, if the people within the data set want to donate their data in order to get something it's their data.

spk_0 Certainly, I mean, they can they could decide, but I think they need to also understand the impact of sharing that data. I think we do not spend enough time explaining that there's not a real way to do identifying and analyzing so much research has been done on the topic of re identification, even before the loans became, you know, this so widely accessible with such a potential to uncover insights and identities.

spk_0 And we it's inconvenient to share that information with the patients or donors of data. So I understand, you know, kind of practically why we don't want to do that.

spk_0 But you know, I think I think we can think about better ways of conveying that information than like a five-page legal contract.

spk_0 And we think we can think about how we can send these patients. So for example, my situation, I was in the hospital, I was convinced I was a good chance I was not going to make it to the end of the day.

spk_0 And somebody gives me this paper. Do you want to participate in the research? You know, we're going to cut this and that out of you. And then we want to analyze this tissue.

spk_0 I haven't slept in three days. I have I don't even know what the words are that you were saying. So that's not it's not a good way to consent people, right?

spk_0 So I think we have to think about how we do this really informed decisions on the part of patients.

spk_0 And especially people who were donors, right? Who just want to do this for the goodness of the planet of humanity.

spk_0 For longevity research, for example, the UK by bank is such an amazing data sets of lots of healthy people there.

spk_0 And they're donating their data to help the world move forward in this research.

spk_0 So I think there are a lot of people who want to make that happen.

spk_0 It's an interesting time. Let's stay a little bit at the problem level first before we talk.

spk_0 FDPA into solutions and also into your history. Have a very interesting with the to talk about this the next part.

spk_0 But let's take a defining the problem a little bit better.

spk_0 When I think back to the 80s, basically the online world didn't matter when I traveled to the United States.

spk_0 Basically, man, nobody can find anything about me anywhere online because the online world didn't exist.

spk_0 So I had really the chance and opportunity to create the first impression about myself.

spk_0 And you person then became then came the internet and we started feeding a little bit of data in the 90s.

spk_0 It was not much. It was mostly science driven in 2004 Facebook started and it was the first time when people were motivated upload your private pictures upload your drink at the party and share it with friends.

spk_0 Usually people don't should not do that. It's not very smart because you have to speak just online decades in the future.

spk_0 In the 90s, it didn't exist luckily for people like me.

spk_0 And when we look further down it, there was a complete shift in messaging online.

spk_0 It started around 2021 with the pandemic in the pandemic where social media platforms.

spk_0 I would say convinced and motivated people right articles post every day.

spk_0 And I think in my opinion maybe some AI companies funded some influencers to spread the message, come on post every day.

spk_0 And they described the benefits of it. So you find new contacts, you find better business partners, works everything well.

spk_0 And this is the reason why I started the podcast on one hand.

spk_0 But on the other hand, it means when we post and when we comment, we leave a huge footprint about how we think, how we feel, what we do, you can analyze a lot of data from everybody and also nailed down mental illnesses and health problems.

spk_0 And also when you analyze speech, I think there is also some some some studies and scientists who want to identify does this person have any progress in goodness.

spk_0 As long as it is a local company is probably fine, but companies are bought sold acquired, shut down date, but the status stays there.

spk_0 Yes, yes, yes, yes, yes, to other people. So this is something you think nobody sees today. The state is there forever.

spk_0 And who wants to data can access it or is do you see artificial intelligence more as a black box where it's not possible to get access to the data.

spk_0 Oh, I think even from some of the digging that you've done, right, very obvious that there's some documents that the AI can find that maybe human would have with enough time, but it's it's that throughput piece, right.

spk_0 How much effort, how much diligence would a human be willing to do to really understand somebody.

spk_0 And then if it's a malevolent player, use that information to manipulate them into a certain decision.

spk_0 So you're talking about mental illness, but that's clinical right if you take non clinical features.

spk_0 There are lovers that you can pull with a certain person to get them to say yes.

spk_0 And as a commercial person, I certainly think hope that every salesperson out there trying to close a deal is doing all the digging to find everything about their customer and find a way to convince them.

spk_0 But on the other side of it, right, from that customer perspective, they might not want to be manipulated perhaps.

spk_0 And it's a balance. So I think sharing your beliefs is tricky because you want to build that connection. I would say that LinkedIn with all of the information that I've been putting about my own beliefs about kind of how the world should work in the perfect scenario.

spk_0 What I believe about AI and so on algorithm is beautiful. It brings me profiles and posts of people who believe the same. It's such a wonderful bonding experience.

spk_0 I love it. I am so grateful for it. Do I also understand that somebody can look at all of this stuff and then convince me of things much more easily. Yes, absolutely.

spk_0 So I think it's a it's a two way street right and also as an individual understanding this about yourself that there's your beliefs are out there in the open.

spk_0 I think you can then be more thoughtful, take proactive approach and protect yourself from people who might be trying to maneuver you to go one way or the other.

spk_0 Sure. I think it's it's from the both both sides.

spk_0 Do you think it's possible to manipulate people via social media?

spk_0 I think it's it's a neurobiology right. So I think of that in the frame reference of the person themselves.

spk_0 There are some previous experiences that make people more vulnerable right some foundational neuronal signaling pieces that perhaps make them more prone to risky behavior, decision making right.

spk_0 The whole of that is open science questions. What comes to mind for me is when I was at MIT we were so huge deal because it was the first casino that was approved in Massachusetts and the backlash of that right the academic discussion about that topic was so prominent because it's it was all about the addiction and how gambling triggers this fundamental neural circuits.

spk_0 It's just it's difficult to overcome some of that because it's that lizard brain. It's who we were before they have we had the massive thinking analytical cortex.

spk_0 So once somebody taps into that inside it's it's hard. It's just practically very hard to overcome that.

spk_0 And so some of the conversations that were live at the time right was how can be protect people how can they identify people who are more vulnerable to these influences.

spk_0 And I think some of it is the people themselves understanding okay I'm vulnerable from this perspective so I'm going to be more protective of myself. I'm going to put these features in place of my life to make sure I'm not influenced.

spk_0 I don't gamble because I think that that's probably a weakness for me so I don't even I don't want to try it. I don't want to touch it.

spk_0 But that's something I'm doing for me. And I think taking some of that control into your own sphere building your life to protect your office.

spk_0 It's just it's key. We can regulate everything to death. But in the end of the day, you have to be protective of of your own self.

spk_0 Yeah, it's a balance act between getting the benefits out of a new technology, but avoiding the risks and it I think it's the next time.

spk_0 You mentioned before that it was yeah, I created about you when I prepared the meaningfuls really tempting you brought a lot about steep personalities and ethics.

spk_0 And it was for that one point. I haven't done that for a long time because chat GPT a couple of months ago didn't really respond well to questions about people who always feedback.

spk_0 No, it's not positive. We're like, no, it's not positive. We're like, it's a private person and not the public person, not the superstar, so I just gave up on that. And when I read your messages, I thought, when we talk about dark person, there's now a play the dark personality and try to find everything out about you.

spk_0 So with chat GPT deep research and I didn't really have high expectations that I had something meaningful back. And suddenly I got a thing it was 10 pages or something like that with references and links and everything together.

spk_0 Where you grew up that you were an MIT at Stanford, which left every single video and conferences, various books. So all this data came back in a lot of documents. It was really impressive.

spk_0 And great. You in fact, I'll cover some of the documents. I did not know were in line. And what I did after that is I emailed those organizations to take those documents down.

spk_0 We will see how well that works. I think that's a lot of it because right, taking control of your own data, I think what we often don't think about is that in fact, we still have ownership of the Facebook Instagram and so on data. So you can take it down.

spk_0 It's completely within our range.

spk_0 But the thing is, I mean, it works really well with a would say for every version of chat GPT, but you can also ask other questions, you can also dig into what weaknesses does this person have?

spk_0 How does the personality function of this person? What are the spots where I should take them aside from we talk about dark person is, I mean, we have to admit it.

spk_0 There are not always always just good humans out there and humans evolve in the lifetime. And sometimes they're good, sometimes they are bad and sometimes they don't know that they are bad or doing something badly in society.

spk_0 Does I disagree with this I disagree with, I think they they know we have evidence to suggest that they know they just choose not to do the right thing to be have the evidence.

spk_0 Yeah, the research if you ask people who do bad things that they know I think for us it's much more comforting to think that they didn't know.

spk_0 Like if you could listen to the interviews of victims of Epstein, they want to tell him some of them are sad that he's dead because they won't be able to tell him now what damage he delivered to their lives.

spk_0 I think research suggests he knew he just sort of didn't really care and he was happy in fact, perhaps you have delivered the damage.

spk_0 It's it's it's a very different framework of thinking. I don't know Epstein, but again, looking at the academic research, I think there's just every bit of evidence to suggest that it's it's a different operational mindset.

spk_0 So basically what we have to accept is the fact that there will always be humans who are bad actors and do something that harms other people on purpose.

spk_0 Yeah, so the current estimates are that in the general population that's about 5% in prisons is between 16 and 25 and the reason there's such a large readout scale right is because we just thought a hard measure.

spk_0 And interestingly in leadership, the percentages enriched to about 10% according to academics who study this. So I think the way that our society's built sort of reward some of these behaviors that are very self focused, self oriented.

spk_0 But maybe even beyond that, right, that it's in in their very different mindset, the other people are like 2D, they're not the same kind of a person or a being, they're very flat.

spk_0 And I think that it's so hard to embrace this for me, I think it was kind of like learning that Santa Claus doesn't exist, right? It's it's that very traumatic experience of understanding that there are there's a whole there are people will look the same, but they're operating so differently.

spk_0 Once I got on board with that, everything made so much more sense, right?

spk_0 Like so I grew up in Ukraine for us World War 2 was like it was happened yesterday. It was it was always in the daily operations. We always talked about it Hitler was top of mind, right?

spk_0 It was not something that was a number of decades ago, right? It was daily presence. And we always were talking to was just this one extreme really bad human.

spk_0 But against studies suggest that it's you can have a range of behavior or some range of execution, a ability.

spk_0 Sometimes you can have that intent, but you can't you know you just are not good operations persons who you can't execute on your bad wishes, but it's it's a much higher fraction of the population every current throughout history we can see examples in the day today.

spk_0 The idea of the degree of interval to where you and Germany and also in Austria is a good example of a really bad system. You can't really show a code that it's just a bad system harmful to society across the world.

spk_0 And when I think back to this time, also when I was in school in the 80s and the university in the 90s, it was a constant reminder that we should not do that again and bring that to humanity ever, ever again. And Austria had the you trolling that actually he was Austrian unfortunately.

spk_0 But when I think back to the 80s or so, I mean, when we had a conversation and we wanted to have a private conversation, it's just house owners could be almost certain that nobody's listening.

spk_0 And this was also the situation in Germany in the 30s and in the 40s. So there was no internet. There was no modern communication and you state that this bad actors like in the third Reich still exists today.

spk_0 But when you look now at today's technology, we feed everything into the internet. You have a mobile phone. So it's camera, microphone, I have Alexa devices in my house.

spk_0 Anybody could easily tune in.

spk_0 We're very brave to have Alexa devices.

spk_0 Really?

spk_0 You know what it is?

spk_0 I mean, maybe it's my background, but I am the kind of person who stays very manual, you know, with a new block and so on. So I do not have this right home.

spk_0 Already.

spk_0 I have every day.

spk_0 I think it's for that risk of monitoring, right? So an entity can then get a hold of your data.

spk_0 One of the very interesting examples is this study. I'm not going to remember the year at the moment, but they they look at the electricity consumption by the power meter.

spk_0 And essentially with a model, you can build a model that predicts how many people are in the household and what TV show they're watching just from that measurement of the power meter, right?

spk_0 You can you can you can predict the TV show.

spk_0 They are watching from the measurements.

spk_0 The pixels.

spk_0 His pixel brightness.

spk_0 Really?

spk_0 So now you think you think about the ownership of that data and now you have an electricity company that knows how many people are in that whichever apartment at what time of day, what they're watching.

spk_0 I mean, if I were not just a company trying to make money, it would probably sell that data to some TV station right to help them understand, get some insights and so on.

spk_0 So it's, you just have to be very mindful for that footprint, right, because somebody might own your data in ways that you do not expect.

spk_0 And this, for example, it was very surprising the time when they carried out the study right in this single readout is predictive of so many things that are with some argument in violation of your privacy of your home.

spk_0 So I think the point that you're making about the connectivity of that data and kind of the pool into some omniscient being your system is, is a whole new level in some of what I've been thinking about is how public videography or photography, right.

spk_0 It's not really been regulated in the profound ways.

spk_0 So you can have a security camera somewhere and so on, which is not an issue if it's just staying with the security guy who's watching the, you know, whatever 15 screams.

spk_0 But if now that's pulled into some more systemic oversight, it's very different picture because now you can track people going here and there and again, the impact through health insurance or auto insurance and so on.

spk_0 You can already see how that information can feed and manifest for you very obviously in financial impact.

spk_0 But it's not like you're going to be able to say, oh, they charged me more because they saw me engaging in some bad behavior somewhere else.

spk_0 It's the intransparency of the output.

spk_0 So to your earlier point, right, like how can we convince ourselves that data is important and we have to care about our privacy.

spk_0 It's a far game and because it's never in your face, right, the insurance companies like I say, we charged you $100 more this month because of these things that we saw through these data sources.

spk_0 So I think you just kind of have to think for yourself of the impact of sharing some of these things.

spk_0 That's an interesting point. So it could influence pricing strategy when we proceed at your frequently travel to for example, just the time example to Monaco.

spk_0 Most of your guys must be rich and this could be a way to there just what since he travels every week to Monaco and just to weekend somewhere to Italy to a cheaper house that could just assume this is a millionaire or billionaire.

spk_0 Different pricing. That's amazing. When we go back to your to your dear roots, you grew up in Ukraine you mentioned earlier.

spk_0 How did you build your perception towards trusting that environment? Could you share some stories of your youth?

spk_0 I think it was just a really confusing and difficult time. I'm not sure that trust really was a part of my world view.

spk_0 I mean, starting with some fundamentals with the government taking away savings one day, right, with a collapse of the union that you said, well, you won't be able to withdraw from your savings account.

spk_0 And it seemed that I remember just being outside and standing in this line with my grandma lots of other grandmas standing in that line crying because there are whole lives just just went to pieces.

spk_0 And it's such a betrayal from the government. It was very difficult to live through and so I mean for me the way that I process that at the time is that all they probably needed the money for something more informed than the lives of these people because it's just easier to think of it that way.

spk_0 And then the crime was so gruesome and so rampant in retrospect. I think it's because of that loss of oversight on the law enforcement became very weak at the times of people didn't have the consequences for their bad actions.

spk_0 And so the stuff that was happening in this is just is too much and you can show you watch that's nothing compared to what was happening there in the 90s.

spk_0 And so again, for me was I think easiest to process it in that well, these people were stealing money items, people's whatever it is because they needed it more.

spk_0 In retrospect, having myself then since then lived through situations where I didn't have anything to eat, it's a really high threshold that I just I cannot imagine how you cross where you take somebody's something.

spk_0 I think it's again going back to the dark personalities you have to be of that mindset, but somebody else is not a real person only you are a real person. So you're authorized to just take from them.

spk_0 So since then I've adopted a different point of view that I think it's it takes a different approach and different way of seeing or.

spk_0 That's an interesting point. I always thought it was a blessing to change from the communist system to to the open the array and it was more smooth process and sunshine into capitalism, so it was not really hard.

spk_0 It was sunshine. There was just no running water, but sunshine we had.

spk_0 Yeah, yeah.

spk_0 Yeah, but I think it was a very rough transition for the country and I know we think of the horrible things that are happening in Ukraine now is this, you know, a unique experience, but it's in my time there, it was never really good.

spk_0 So you also experienced the last days then basically of the old Soviet Union. How how was that?

spk_0 Well, I will also practice this with that because you know, so the union is a very large country and in the Zaris time was also very large with poor communication systems. So when you think of the collapse of the union actually took a few years to propagate that to remote regions like where I'm from.

spk_0 It wasn't like an instant change for us in the mindset.

spk_0 It was still present there. So we're talking about how, for example, commercial activity was so shamed for so judged if you sold something somewhere, if you bought something somewhere and then sold it somewhere else for more.

spk_0 Public shamed for that kind of an activity. And so for me, I think moving to the US with this very commercial mindset and you know, I've been in a lot of commercial functions. They are through an ROI person.

spk_0 It's just a massive shift in thinking that in fact, yes, you're adding value by even let's say moving goods from over here to over here. That's a value head for someone because they maybe not are not able to access this location, just as simple as that.

spk_0 So there was, but how did the Soviet Union then function at the end of the day when there was no commercial activity here?

spk_0 Excellent, excellent question, but I think it was a lot of show off, right?

spk_0 I mean, just like during the hunger times, just demonstrating things are going really well inside the country. So whatever by selling wheat and other goods to other countries, well, you have your own people dying of hunger.

spk_0 There's a lot of showing demonstrating this external success while internally the operations were less less than good right lots of deficits.

spk_0 Again, people standing in line, this is before my time, my parents time, lots of waiting in lines for you know, the limited groceries at the store.

spk_0 And then I think it was a bit better after that, but then my time again, you couldn't even buy anything. There were no, there was no food to buy. And for a country that's married to cheese and sausage to not have cheese and sausage one day, you could not buy for any amount of money.

spk_0 You could not buy these two key ingredients for for the kind of music. It was very difficult. It was a completely due reset people rethought how they operated what they ate.

spk_0 The whole lives changed in major ways.

spk_0 What were the main differences in how the system was set up when you compare you know both sides, the Soviet Union and you know the US.

spk_0 What were the main differences on the economic side from the setup perspective?

spk_0 I mean, I think in some ways, the Soviet Union operated in a non-sustatable way, which is sort of why we ended up where we ended up in the end.

spk_0 So I'm not sure that operationally the system could succeed.

spk_0 But it's, I think it's the, again, the trust, right going back to the trust, the government to take care of you.

spk_0 If you needed support, I think that social piece was more in place. But also if you put all homeless people in prison that really reduces your homeless population.

spk_0 You know, that's an easy fix right there. So I think some some solutions were again on the external let's show off how good things are side and so.

spk_0 You know, I, I think all the movies you want right it's you have the tourists kind of going down that main street.

spk_0 And then just as they go off the street, it's is a completely different set of ruins. Now you don't have paid the building in the ruins. I thought this was the thing of the past, you know, maybe the 60s.

spk_0 I went to St. Petersburg in 2013. I think it was the same. You go down the main street, painting the buildings, you turn off the main street. It's nobody has painted that in the very long time.

spk_0 So I think it's still very much this or at least at the time was very show based, very external perspective.

spk_0 And you know, I think it's just not a sustainable way of living because essentially you're not building the inside. You take that to the personal framework, right, a way to forget the government.

spk_0 But for yourself, if all you're projecting all day long as success, but at the core, you're, you know, you're dying on the inside, it's not a sustainable way of living.

spk_0 You have to have some positivity inside, then you deliver it to the outside.

spk_0 And in comparison to what experienced in the old Soviet Union, when you contrast that with your experience in the United States, how would you describe that?

spk_0 You can, in some ways, it's a completely different way of living, right? I mean, commercial is so front and center in the US.

spk_0 But I think there's still a lot of this surface level, everything's very good. How are you doing right the answer to that for me was shocking, talking somebody for me.

spk_0 And I think that's the reason you're up coming in. And I really struggled with this the first year and I deeply apologize to people that I interacted with during that time. But when they asked me how I'm doing, I thought the answer was to actually then describe how I'm doing, which was not very good those first few years. Let's put it that way.

spk_0 But the answer to that is great. I'm doing great or I'm doing well. So it's some of some of the future are actually really similar. And as on the side, there are people who've written about the similarities between the US and the Soviet Union infrastructure wise, just how maybe it's a big country thing, but there are academic studies, there are theses that have been made about how similar the countries are.

spk_0 And I think those are just really fascinating pieces to read because we think of them as so polar opposite right called war and so on, but in some ways, these are really really similar countries.

spk_0 That's interesting. It's interesting. I mean, at the end of the day, it's all back to the artificial intelligence when we talk about the Soviet Union, when it was so to write, it was more a centralized system, where basically you

spk_0 have to let the government decide what's best for you, and you don't have to think they make the decisions you just have to agree that's everything that you have to do at the end of the day.

spk_0 You just have to live. I'm not even sure they're green as so much a part of it. Just kind of you know, go on.

spk_0 Yeah, and for there was no room for personal development at the end of the day. So every commercial activity was shamed selling something to another person to make more money at the end of the day to invest in yourself in your personal development was not an option.

spk_0 Compared to the US, which is completely opposite at the end of the day. So it's more selfish.

spk_0 People are looking more for themselves and their benefit and their profit make deals and have to be sure that's the other side, the scare of their win and they just have to take care of the wrong win and the government has a role to

spk_0 combate the Soviet Union to put a little bit of framework in place so that people can't transact and the economy thrives and that's all. But it's more self-centered than the Soviet Union.

spk_0 What was good in the Soviet Union and what was good in the US when you could combine that the past of two worlds into one system, what would you choose?

spk_0 Yeah, I think that's a super interesting question. Then maybe just comment what you and what you said. I think part of the reason that when happening very capitalist society, right, you have to be taking care of yourself is because there is not a social system that will take care of you.

spk_0 If you have a dip in the Conmore, when you retire and so on. So it's a two-way street.

spk_0 I'm sure a lot of people would be willing to spend less time building own wealth if they knew that there was that support network.

spk_0 Sure, I think in terms of combining the two approaches, maybe I'll lean on education here. I think some of the really different things for me.

spk_0 And maybe it was I was student when I came here first, but that in a lot of classes, it was about this analytical thinking, right?

spk_0 You take math, you take English, you think about what is it that you're doing.

spk_0 I think the Soviet Union is very much about memorization. So there is a right answer. It's that right answer and all you need to do, memorize what that is and then deliver that answer.

spk_0 For example, exams for us, right, it was always you memorize word for word.

spk_0 What the responses are to the 30 questions you come in you pull out the ticket, you know, which you know the number of tickets, so you already know what the questions are because you memorize them.

spk_0 And then when I came here, I was to do every week we were writing an essay on what the author thought and what did you think about what the author thought and it was overwhelming just in like how much space there is to think about things.

spk_0 I think for me, I mean, I've been here very long time, but for me, this thinking part, it's really only, I would say fairly recently that I've really come to embrace that right that one is authorized and perhaps even rewarded for having one's own point of view.

spk_0 And other people may disagree with that point of view, but that's okay.

spk_0 So I think there's some value to the diversity of thinking and I think there are a lot of other countries that are built in that memorization framework that have very similar people coming from those countries, you guys have that very similar experience to mine, where you just, there's diversity of thought is very exciting.

spk_0 So I think that's to me that's something really valuable bringing this back to a high right some of the conversations are, you wanted to in the past past two years ago, if you wanted to get an answer to something you asked, you know, five, 10, 30 different experts and you got different answers they aligned on something they diverge from something else.

spk_0 Now, if you want to answer you go to chat, you can get one answer.

spk_0 I have yet to see a post on LinkedIn about somebody doing a same query for chat and GPT, let's say for science, 10 times or maybe between chat, GPT, for vaccine, so on, right to get that same flavor of diversity of thought.

spk_0 So I think this convergence on the single point of view is in some ways quite terrifying and big shift for a country like the US.

spk_0 So we'll see how we handle that and you're earlier point about kind of maintaining that source of truth, right, if the one answer that you rely on is coming from your AI tool.

spk_0 How can we make sure that the information that it's out is reliable and just true and there's some let's a global efforts right to maintain a source of truth somewhere so that we can reference that source of truth so the history does not get rewritten by AI tools through generative means.

spk_0 That's the two interesting points I think this discussion about.

spk_0 Just two questions just too I left for so long.

spk_0 Just to not more.

spk_0 I don't get that to give you more but it's just it's more a time sorry for being in polite it's more a time restriction otherwise we discuss eight hours so I try to pull out pull out to it not not devagued the others but to spark the curiosity.

spk_0 One thing what you mentioned is understanding of of social systems is important to understand the risks of AI.

spk_0 You experienced you are one of the rare people who experienced both systems the east the oldest and the west and I don't think human humans changed much in the last hundred years technology evolved but I think on a on a psychological level we are still the same as humans and humans gravitate together in groups and when they are in groups they create societies and we have so many.

spk_0 So many examples of different societies and now we amplify that with the I.

spk_0 This is why I think this discussion about what systems were there out there specific state they have what's the comparison what's the good side in the US what's the good side in the Soviet Union what's the bad side is really important to understand what we then amplify the end of the day with the I.

spk_0 Yes super interesting point and just just the corral worry of either a lot of the training data that went in was English based so I think it'll be interesting to see do different regions then develop once they localized right in their summer efforts now in the Nordics to do that once you localize that and you train with your special they let let's just say we focused on the language side.

spk_0 But are you going to get different answers based on that and is that going to be more reflective of the kind of social system that is in place right I mean I did not come up with this idea at all but some of the folks who look at the globe say that this community versus self focus is a gradient as you go from east to west and you can agree or disagree but I think maybe AI is where we're going to be able to test some of that those hypotheses.

spk_0 So we're interesting.

spk_0 Yeah, not all agree and the other points that I found compelling was what really didn't think about so far.

spk_0 Trusting AI too fast the old days were basically when I had a problem I had to ask myself through a lot of people just thinking back in the 80s in school when I couldn't solve homework I had to call friends.

spk_0 And I have to identify one who solved it from 30 possible hits so I got a lot of information I had to evaluate the information it's not that when I asked someone and said that this is the answer that I could assume it's the right answer and our teachers in the 80s and 90s said Christian start thinking for yourself always come up with your own opinion and now we have the eye and the I risks if you understood you right that people just take an answer from an AI as already proved.

spk_0 And don't think twice.

spk_0 Yeah, and I think against some of this is the fundamental neurobiologist that lizard within us I think that trust in another.

spk_0 This is well studied you can take specific behavioral steps to build trust with another hopefully you're doing it for the right reason.

spk_0 But we can talk about the 11 players as well and so I think when we look at AI we think this is an expert because it's analyzed a lot of data.

spk_0 And that's one piece right this perceived expertise perceived knowledge then for us through the heuristics that we use translates to trust.

spk_0 The other side of it is the confidence if we see somebody confident.

spk_0 We start trusting them more because of that tone of delivery and chat you could use always really confident in the answer that it delivers that it's like the absolute truth that there is I did not yet write an article about this but I just thought this is fascinating so I you know I'd like to stand up at the tech turns.

spk_0 I think of myself as you vary your data Germans hold so I wanted to optimize the performance of my posts on LinkedIn and I asked you to be for some suggestions with confidence it delivered a whole list of suggestions.

spk_0 Meanwhile some of them involve buttons that do not exist.

spk_0 I mean it's not possible to execute on the suggestions because the button is not even on the interface so anyway but it's not heuristic that we translate confidence to knowledge in our minds.

spk_0 And this is this is ancient knowledge this the Greeks have figured this out that if you are on stage and you you list some credentials like that you have analyzed a lot of data and then you project with confidence and opinion that just drives trust in the people who are listening to you because of the fundamental neuro biology so it's.

spk_0 It's like we have to protect but to the earlier conversation right we can focus on fixing chat GPT to express confidence levels for example it could say this is my answer in about I'm about 60% certain so that's a technological solution to this.

spk_0 A much faster solution for us to implement is to say chat GPT has analyzed a lot of data but may not be an expert we acknowledge the confidence with which this tool delivers information however this may just be a tone that it's built to use that is not reflective of the its own confidence in the answer so we can take a proactive approach here and protect ourselves from these influences.

spk_0 Once we become aware of the little things that for us they're translated to trust.

spk_0 That's an interesting point that's really interesting point they always fought chat GPT and LLM search in my or grog just our retriever system that just do the same like search engines going to a library pick the book that I need to answer my question and just reproduce the content but it's not actually it's not.

spk_0 Yeah I mean it's the you typically you do not say I want the specific book you say I want an answer and then the system sort of has to figure out which books to pull from the shelf to answer that question so you have some uncertainty already there and it's might be different time to time or run to run within a given system or between systems and then stuff that it pulls out and shows you that's that's a whole other side of it and there's some problems.

spk_0 So I think it's not a problem with the list of the list of nature to that so you're going to get variability with that as well and maybe you know like with so sorry going.

spk_0 No, just keep going.

spk_0 I think maybe maybe it's like with science you know we run the experiment three times so maybe that's something that we could do in the future you take different ways of measuring concentration maybe that's something that we can apply to this tool but it's we have to do it.

spk_0 We have to think about reliability as core and especially as we apply these tools in our professional domain right within farmer biotech life sciences we have such a strict code of conduct that is all about reliability very fiability of information.

spk_0 And so yes to implementation of AI but but I think taking that same approach that we have the same hesitancy that we have had for all the other tools that we've ever brought in right and really assessing it from from that reliability compliance privacy, whole suite of other things before we move forward with the tech.

spk_0 Again as a technologist i am all about implementing technology i just want to be very clear i'm all about reaping the benefits of it but also really putting the framework in place that make sure that you protect yourself protect your company IP protect your know how.

spk_0 You know make sure that whatever you deliver to the regulators is true and correct these fundamental things we have to put in place.

spk_0 And what you say is contrasted by the marketing of open AI for example or croc the one to convince everybody that it's the best tool and it's always right and will replace human it's the best experts in the room expert in the room and I use it for example for analyzing stocks companies and

spk_0 it's interesting companies needs a lot of data and it's really convenient to just upload it and say make a deep research about this company but the funny thing and interesting thing is when you ask for recommendation should a balanced company or not.

spk_0 Public list of companies chat GPT or croc or jemen air come back and say yeah it's such a good buy and it's such it's currently at the low price and has such a prosperous future ahead and you will bake a lot of money.

spk_0 They convince me and then I said this can't be right they just had a 400% uptake market capitalization the last six months so I go back to this tool.

spk_0 But what about the price now oh yeah the data of the last three months is missing so it didn't look at that and then I then go to chat GPT and said please look at this data they completely changed their opinion so it's not a really good decision making to let the end of the day currently at the stage.

spk_0 Yeah and I think that's why we are right now very focused on this human and the loop approach that you have the human like yourself are actually thinking about some of the other things that maybe the algorithm didn't take into account when delivering information or suggesting a particular path.

spk_0 I do think the long term human and the loop is going to be very painful and maybe not even practical to maintain that implementation right because first of all if you have to walk into every single thing that your model it pulls out and implement and suggest.

spk_0 So if you are just a lot of interest it's a lot of extra work.

spk_0 Yeah.

spk_0 That's not very interesting you're doing QC so I think in some ways it's more interesting to actually just do it yourself because then you're on the creation sign you're less on the QC sign I think in in front of my biotech you know QC is a very particular skill set and some people are exception at it but most people are not.

spk_0 For me this is reminiscent of my experience when I went from individual contributor to manager.

spk_0 And my work as a manager became a lot of this QC right I look through the decks that other people build and like I would have done this way but okay let's go with your approach and then all I'm doing is pulling out looking for errors right because I'm really responsible for the output I'm the human in the loop.

spk_0 But I have to make sure that whatever comes out of our team is the highest quality that we're operating with the newest information that we have.

spk_0 But it's just a very different flavor of a job you're going through and on every slide every number you're like that that that's that correct versus if I had to come up with a number myself it's it's a very different nature of work.

spk_0 So I think it will be difficult to retain a workforce that all it's doing is there fine whether the insights provided by the algorithm are correct or not.

spk_0 Yeah personal opinion.

spk_0 You may be looking at support and I think we need that interesting thing with this LLM still is when they are wrong and you tell them sometimes they are wrong they try to convince me that I wrong.

spk_0 The kickback is what are you doing there and they don't the down of feeling so they don't feel responsive where they can come to a wrong conclusion insists that the wrong conclusion is right and the company goes fast and the interesting thing in your VTS you not only were at MIT stand for it and have experience in the old Soviet Union you also work in the industry.

spk_0 So it's not academic discussion that we have here you worked with them official and Saturius this just just to but correct me if there if I'm wrong it forgot something and the trust the chat you be too much with the analysis.

spk_0 The problem now we have with the eyes A eyes out there people use it and people love it. I mean when I think I think of me I went in a couple of months to 100 million users so there was a huge up there.

spk_0 So there is a need on the market and then you need to lead these companies but generate wise to the leadership teams what should they do now in this situation there are risks they're not perfect at all but people want to use it how is that for leaders.

spk_0 Yeah I mean I'll talk about let's say the less as a terror things I think we feel out of the news about how we can see has to let go of lots of consultants because now they can do this that and the other with with a model great.

spk_0 I love it. As a consultant part of my work is on the commercialization side so I sometimes around these experiments right what would change the fees adjust for commercialization of a particular tool.

spk_0 And once again it comes out with suggestion that it's very confident in and so I can see how for leadership it's often tempting to say consultant so expensive I'm going to go with this much more

spk_0 much deeper way of doing this right I'm just going to ask to have a team.

spk_0 The problem with that is that you could then pursue a path that steers you much away from an optimized way of doing things or it's a completely wrong direction and you won't know until many months down the road so now you've invested lots.

spk_0 But down the wrong path and so my suggestion is still talk to somebody who's commercialized things in the past even if maybe maybe a new way that we work in is that the company comes up with our own strategy with the assistance of an agent and then they bring that to an expert and experts suggest so this might work this might not work maybe that's anyway I don't know what path forward is but I think going with just what the agent suggested to you is very risky.

spk_0 And then I think we have this other kind of idea that we're going to implement AI and now we can compensate for you know the 40 50% reduction in the workforce that we've had to go through because of the winter.

spk_0 In biotech that we're experiencing and I think it's not quite as easy as that unfortunately right because the expertise that's lost in layoffs I think I think that's that's going to be very trying to do that.

spk_0 And that's a very dramatic for the companies.

spk_0 I think in the past it's going to attach a knowledge management agents but in the past you know you we've gone through some phases where we maybe we did we had to lay people off for two three months and let's just go with some basic thing we had a piece of equipment that nobody remaining at the company knew how to fix but you can probably get by for three months.

spk_0 Now we're in a much longer stretch of this so we're having to find ways to operate without without that expertise and I think that's very very tough.

spk_0 The experience that I'm leaning on from my life as I was leading a team and there was somebody on my team who had been at the company for 30 years he was retiring so my job was to in these two months.

spk_0 Before his retirement download all of his knowledge so that when he leaves we don't have to bother him with how to run this not machine and so on how to fix this protocol right but it's it's such an immense task to do that in a short time frame so I think there's a massive opportunity for a base companies to come in and do this download of knowledge from the workers I mean my dream is that we have a little camera as people execute on the on the SOP that gives some visual basis.

spk_0 Some tacit knowledge that we can add into the written documents to help us.

spk_0 Run processes with a leaner with a leaner workforce so I think it's not going to be as simple as we can now compensate for all that's lost because of the economic situation that we're in we can probably gain some efficiency but it's.

spk_0 Again going back to the simplification you have to know what the best way of doing something is and then you can automate and provide a song if you don't know then it's it's you're just doing more efficiently something that might not be on the good track.

spk_0 So yeah investigating what the good track is I think is key.

spk_0 Yeah I don't think we are still early with other lamps and this is just the the the try to solve every problem with other lamps currently which is usually at the peak at the first peak of the panic Ruger cycle before we crash and maybe we have too much of a try to use this tool in a sales context and was really interesting for me was that in the last year chat you be the seem to have been trimmed towards pleasing people.

spk_0 What that means in a sales context is that chat GPT tries to agree to everything a customer wants which means the customer doesn't wants to pay very often so chat GPT didn't comes back with yeah you get this you get this for free you get this on top of that they're you agree on that.

spk_0 And the thing is when you just work with chat bots in a sales context and you don't understand what the risks are and remove all the humans who intervene and say no no no you can't give our car so we for free to a customer because the customer wants it you really risk ruining your company at the end of today.

spk_0 Yeah and I mean I think you can put some constraints right just say this is the space within which you should be operating and I think in the sales context it can be these are these are the things we will never give away this is your kind of have this for the sales teams now right we tell them you can operate within 5 or 10% of the price but you cannot really drop below 10% discount.

spk_0 And give these rules to the agents but I maybe want to even abstract further away from that to organizational values so I think implementation of these constraints would be key in the broader context.

spk_0 So I'll start with an example let's say we're 20 years from now and we have another biotech winter so we're now fully agentic and I am a CEO goes to the agent and I say give me give me a path for this winter.

spk_0 I want to survive and it says yeah fire 90% of your workforce well that's definitely going to give me more of a runway because I've just laid off all these people but maybe that's not really consistent with my values of retaining talent with my values as an organization of respecting humans there are other ways that we can approach this for low the production of paid right all these different things other than just lay everybody off.

spk_0 And I think that that's a kind of code of ethics internal to our organization might my organization's value that we can bake into the the algorithm as a constraint another way another one of these to think about right a sustainability I can maybe ask 20 years from now I asked my agent what's the best way to synthesize this and maybe it's going to give me the cheapest way but it's going to use all kinds of toxic chemicals now if a value of my word.

spk_0 The organization is sustainability and building green processes I have to tell that to the agent right I wanted to know what my values are so that it gives me a balanced optimized suggestion that yes gives me good our eyes so that I can.

spk_0 Power commercially saving people's lives that's key but I also say the environment is important to me I want to always be in 30% less energy right something something like this so a I actually see this as I know I know we think of values is sometimes fluffy but I think this time is an opportunity for us to really lean in and understand who we are as an organization what do we truly believe about ethics morality humans the plan is to be a good example of what we are doing.

spk_0 We are going to plan it and then feed all of that into the AI that we're implementing.

spk_0 It's a very good point when you talk about agents you mean artificial intelligence agents.

spk_0 The interesting thing was we had to say agents well I think back to the 80s when I started programming with basic the great thing with basic was you structure a program and every time when I pressed run

spk_0 and wrote run and pressed it enter and the program executed I could be 100% certain that the computer executed the program in exactly the order that I wanted it to and to challenge your assumption a little bit with with these agents.

spk_0 I also hope to understand the process well enough and let's stay with the sales example it minutes of a company needs to sell when you understand your sales process when you understand the customer person or

spk_0 the problem what's the buying tool that they want to pay and not you can codify it in a textbook you can write the process at the script you can outline every single email and what I was hoping for with

spk_0 the analysis then you can take this folder this binders and just throw it into the LMS and you have your perfect agent that executes exactly what you want but it doesn't this is so frustrating.

spk_0 Nine out of 10 times yes it might work but then you have this one time where it doesn't work and it comes back with something completely ridiculous and this can derail your company at the end of the day because when it's just a big agreement so yes take all my equity

spk_0 and I'm fine with it then it derives you do you think agents are really capable of executing instructions according to the plan they got.

spk_0 I will challenge you on this somewhat because I think the value of agents and let's just focus on the commercial space is that they're going to be able to tailor that framework that you described the book they're going to tailor that to the specific customer or potential customer that you're that they're working on with and so on.

spk_0 And it's going to be high for putting this tailored customized approach in ways that you just could not possibly pay a commercial team to do right so I think that's that's sort of the beauty of it but it's that balance between how to make sure that the tailoring doesn't then disrupt the process that you know works well.

spk_0 But it's such nascent state at this point I don't think it's possible to answer your question as to whether the agents will be breaking or not breaking.

spk_0 I mean maybe right sometimes you hire say you're sticking with sales you hire salesperson has a very different approach to like the way that you're your organization typically runs.

spk_0 There's a huge value to that person because maybe they will find another way maybe thinking of this from the math perspective maybe the path that you have in the organization is a local minimum maybe there is a global minimum that you are new different way of doing things can get you to local minimum meaning that you will put even less or global mean right you bring less effort than you were before to get the same ROI or better.

spk_0 So I think if the agents we can maybe map it up right say again from the math perspective you have variability you have the leeway to vary that textbook process by 10% 20% whatever it is you can explore the space around or maybe go to completely different path and there could be huge value to that in of itself again through putting your testing helping you.

spk_0 You know explore the design space so to say going to the kind of by processing language but I played to the commercial framework.

spk_0 I agree to the point that you made with testing was okay I is a great for a great testing ground so you can test different reactions to different emails to different conversations you can find on your tone you get 10 suggestions you how does it sound when I say it harsh Lee how does it sound when I say it politely.

spk_0 And you don't have the full scale and this works fine but I'm not really convinced in what I tried since in the sales context to really use the iron for the maybe we can accelerate the process and when we write I mean the hard thing in selling is you need to talk to people you need to listen to people closely you can't just

spk_0 say you talk for 30 minutes and then send you something in the deal is done you really need to relate to them understand the problem and then provide a solution especially in the biotech p2p context and in farmer p2p but LLM's really don't work well when I just try to feed an old email into the LLM and say okay this is the minutes from the last call and now please create an offer tailored to this person and you have the other 10 offers already.

spk_0 The just sometimes come up with just ridiculous things in the mail and when I trust LLM's and say I now just send it to the customer to the customer without double checking it.

spk_0 Yeah, you might have given a years worth of services away for free and LLM's does do that.

spk_0 So couple points to that I think again this is a very young tool so I do for as it's a technocratic hard I do for only believe it will get better.

spk_0 In terms of that consistency but I think the other side maybe counter to what you were saying the agents can exhibit empathy that is better than what you get from another human.

spk_0 So you have lots of the news kind of in the normal human domain but thinking about regulated industries and healthcare.

spk_0 Google presented on this on their tool it's called Amy but they compared empathy ratings by patients of their AI tool versus their physician and the AI tool scores better.

spk_0 So I think there's a massive opportunity in terms of that tuning into the customer and being able to give them a good experience right because the AI agent can be a great listener.

spk_0 So I think I think it can be massively helpful for for the commercial function but again right it's each these the risk of it going badly with one of the customers when you were talking about giving away a year worth of services I wouldn't even argue that that's not near as bad as you have a key account.

spk_0 That your agent just said something completely unrelated to and it probably used your name your email and now you've lost the key account because they they think that you're insane or didn't read the previous messages right I mean it's it's a I think at that relationship level it could be much more detrimental so for now I think we're still you know in this proofreading and so on but I have every hope that the tech.

spk_0 We'll match a higher standard and we'll get to a less supervision more reliability situation.

spk_0 Now it's really interesting discussion and challenge for companies at the end of the day I mean you feed data into systems confidential data I'm not so sure the point that you mentioned about the empathy just read I really like to conversate I mean I use it a lot to this tool to understand them better and what they can do and cannot do it.

spk_0 So I really love this discussion so to have with chat you are broke.

spk_0 We're gonna say please check this post for me. Rock for example started with that and bring me more information is this true and various the source coming from.

spk_0 And then they come back with a solution next time you wait and you get a solution and I say no this is not right this is definitely not right and then I sit down and say so now I tell the system that you are wrong.

spk_0 And the interesting thing is they come back with empathy I understand your frustration as a city I'm not frustrated I'm not frustrated so stop insinuating that I'm frustrated and then the second part is very interesting there's a lot of information is right.

spk_0 And as a very very simple video I mean it's just wrong information with telling the user that they are frustrated and many people just take it at face value and say okay now I'm frustrated.

spk_0 This is is a tool that you can use to manipulate people at the end of the day the read the feed and then they act.

spk_0 There are studies on how convincing the agents can be an advertising and again there are academic groups for studying this at scale and it's the same like two way street flip side of the coin right where it can be a tool it's at the end of the it's just technology.

spk_0 You can use it for good you can use it for not so good so I think this is such a key time to focus on the values the correct approaches that work and amplify and scale that I know this is probably the third time i'm saying this on this call but it's it's it's it's just if you think about is a tech.

spk_0 I think it it may be shift your framework and hopefully.

spk_0 Helps pivot towards this understanding of who you are what you actually want to achieve.

spk_0 I think it's a different kind of technology than some of the other ones because usually when you bring something you like automation by trying to do automation and manufacturing for the longest time there's a resistance from your workforce.

spk_0 And so you have to overcome that hurdle here you're in a place where people want to use it it's very different tech from that perspective adoption is already rolling.

spk_0 So you have to take a somewhat of a different address some different set of risks adoption is no longer your main concern right you need to be driving the framework.

spk_0 The governance making surely you protect yourself protect your company protect your IP and know how as the tech gets adopted.

spk_0 Yeah let's take the conversation back to the corporate library.

spk_0 How should we deal with AI what's what's your framework that you recommend for CEOs in the biotic in the farmer industries.

spk_0 Yeah I think that there are many pieces to this right but to me they're very first and the most important one is privacy making sure that your data does not leak anywhere and you know most companies have been working with some data security companies consultants they're in a good place from that perspective but the AI brings in new risks.

spk_0 So you have to think about all of the ways that the data can flow out of your organization and then work with your technology developers to your AI tech providers to make sure that they put the guard rails in place to accomplish your goals of privacy confidentiality compliance or all the other things that you want to put in place.

spk_0 I'll just give one example because I think it's it's an easy one an approachable one so let's say we have a database that you are implementing and this database has AI driven suggestions for search queries so maybe you put in your parking name and then it suggests some additional search terms to give you some insights great.

spk_0 We want this we want the intelligence within organizations to be built awesome problem is that the AI suggestions for search terms are trained on user data so now you have information data flowing out of your organization to train this algorithm to suggest search terms.

spk_0 If you're searching for you want so hot lots of companies are searching for it probably not as big a deal but if you're working on some obscure target this is your super highly confidential IP you want to get some intelligence about the research for example that's been carried out on this pretty.

spk_0 Yes of course you do but if you query that and you have to keep in mind that now you've disclosed essentially to the system and everybody who's working with the system that you're working on this if you have one other competitor who's working on this target they essentially can see what you have searched for.

spk_0 So of course you can then guide the technology provider to help you protect your search queries but it's not a default at least in the world that we live on in today as a customer as a therapy developer you have to ask for it you have to ask the questions how is your provider handling that privacy and confidentiality side and you know I work with both sides with the AI tech companies and with therapy developers.

spk_0 And it's as well as with regulators right so I think it's it's an ecosystem question today how can we get these tools to be usable for us in a regulated industry with strict IP requirements so I think it takes action for more parties we need the customers therapy developers to mandate I need this or else I won't be able to work with you.

spk_0 We need the tech developers to be willing and open to make amendments implement strict systems and then of course have the regulators to say these are the standards that you have to have in place this is something that law for us to have is a kind of certification some test run that we run on every AI tool to say okay it meets maybe it's a 10 points course system or something like this right it meets seven eight points out of 10 points.

spk_0 So then as a customer the 30 developer you can assess the risk level of using a particular tool versus the benefit that you're going to get out of this tool because it's always an equation right you have to you have to balance these things but if there are any technology people out there thinking along the same lines of setting standards reflective of the code of conduct that we have in this regulated industry.

spk_0 I would love to connect please please reach out and then I am just really grateful to the regulators the industry organizations who've been having this conversation with me I think this will be tremendous for our space to have some some level of standard.

spk_0 We can think of it as for example you know ratings for parts let's say for Boeing right there have to be a certain level of quality and we have some of those implemented for hardware we just don't yet have it for AI but I'm confident that we can get to a place where we have some standard system of assessing tools.

spk_0 Yeah there's a lot of risk and uncertainty involved at this stage for the understand from from your explanation but what does that mean our father for the biotic company actually researching a new therapy I mean there are no standards on the market and using chat GPT is tempting so when you do a clinical trial just throwing the patient data into an LLM is tempting to get.

spk_0 Please don't please don't throw patient data into chat GPT.

spk_0 Yeah I mean I think again we're so as scientists we're so driven to get an answer we want the insight and I think the potential that AI opens up for these insights especially for able to look across these massive data sets that are comprehensive again going back to South air if you think about that vein to vein workflow.

spk_0 If we are able to analyze a large enough data set we can finally get at the answers for what matters to be able to deliver an efficacious therapy that you can in fact manufacture with reliability I think this would be tremendous I'm going for it I want this when I'm a patient I want to have this as an option available to me but it's just have I think as a therapy developer you have to ask questions of your technology provider and I'm happy to say that some of the

spk_0 AI tech providers I work with are so focused on quality reliability privacy I love this I want them to also then be more focused on commercial so that they can propagate this great technology that they have sure I'm not going to say anything groundbreaking but if you build it they will come it's not true you have to bring them to see your great thing that you've built commercial is not an afterthought it's a core part of your approach

spk_0 so that you can then make the money to feed better development so I want the tech companies to think about if you I mean if you believe in this highly reliable tool that you're building please please make some commercial efforts so that your potential customers can see the great value but I think where we are today is that the potential customers are not we don't have the list of questions to ask of the tech providers I think we have to get to a point where there's some

spk_0 understanding of the risks and so therapy developers in the routine of asking these 20 30 however many questions maybe the burden will be alleviated once we have some regulatory framework we have a certification program might we can say

spk_0 out of these 10 points to 9.5 so you probably don't have to worry about the 30 questions you know maybe you ask two or three and the rest have already been by default answered with the certification program but as we're not there yet

spk_0 and as we're so innovative and we want to be on the edge of technology I think you have to again you have to take control and protect yourself and your organization in a very thoughtful way

spk_0 Yeah, yeah, that's right then people are seduced by open AI by alphabets the big big companies out there who also wants to sell their solutions and promise everything and it's just not working in farmy and biotech so easy.

spk_0 Yes, yes, we're I think we're very different from the day to day consumer because the impact of wrong answers of misinformation is absolutely devastating the impact of leakage of data is devastating and like we talked about before right sometimes you won't know that your data has left your organization.

spk_0 You might not know for years that's really scary there's not an immediate readout so you have to ask the questions upfront and maybe if I can just say here I think that I'm such a believer in collaboration I think we can get by working together we can get to answers so much faster and at a lower cost a lot of hesitancy that I hear from therapy developers obviously is that they don't want to share their data.

spk_0 But I want us to think about this in the bigger context so if you're collaborating the tech is here to do this more privately and confidentially and securely that you're actually retaining your data and not sharing it with the pen 20 companies that you're collaborating with.

spk_0 Meanwhile, as you implement AI not asking questions about privacy you might be sharing your data with the whole wild world so it's the balance of the balance of the data.

spk_0 The discrepancy of those two things I think existing in the same time and space I find to be a very interesting place but I it is my firm hope and belief that as we develop a better understanding of data privacy security confidentiality in this new world of AI internally we will then become more comfortable collaborating with others in that private and confidential way.

spk_0 I think there are key points in there so one key point is that LLM's are just a new technology and also artificial intelligence agents are new technology in their infancy it's not a mature system yet.

spk_0 When they the funny part is you talked about therapy companies so there are some times where we hesitate to collaborate with other companies but experience a little bit less hesitant to upload contracts into LLM's to analyze them especially when it's contracts from from Asian countries for example data and Chinese and in English.

spk_0 Just upload them and then you have the content even an LLM like chat GPT is basically public domain so when the data the data left the company and I'm not really sure that people are aware of this risk.

spk_0 When you have a contract in with some details when you can read a lot out of a contract and know what stage the company is what they are working on sometimes there is confidential data in there so you have the confidential data then also in LLM's it's really complicated so what we need at the end of the days not just the LLM from open may as just the base layer so you need really like in old times like in the 90s with the internet you need this layer in between that's.

spk_0 Make sure that this base layer fulfills all regulatory needs that we have in farmer and this is a huge space for new companies and unfortunately mentioned it before they are not expert in selling so they produce the solutions and really invest with their investors in commercialization in getting out to the customers in understanding the customers.

spk_0 How can we fix that.

spk_0 Well I think as we talked about before I think for new founders it's this recognition that revenues key I know you're really excited about the tech that you're developing what to get more data in get the better model maybe you're in stealth mode you don't want to reveal all these things but you sort of have to sell what you have today to feed the development of what you want to build.

spk_0 I think there's a lot of excitement on the investors side now about the usage of AI since like if you put the word AI in your pitch then everything's more seamless but we have to imagine that this is not an eternal family that it will inevitably will see some of those loneliness of adoption slow as outcomes and so on and we'll get to AI being less of a trigger for cash this is.

spk_0 Oh hi say hello and so and I think that founders who think about how they will feed themselves more sustainably feed their companies they will invariably do better this we have seen in time and again right this is not anything new.

spk_0 Yeah it's just a new kind of technology so I think focusing on that commercialization i'm sure everybody's tempted to just look at chat GPT and ask it what's the best strategy to commercialize.

spk_0 I would argue the more niche your application is the worst the answers that you're going to get.

spk_0 I will be so talk talk to somebody who.

spk_0 Who has commercialized that before obviously you know I am in the states I love working with companies on this and I think there's so many ways that we have seen the things do not work so.

spk_0 Moving around those obstacles and making sure that you get to successful commercialization scheme so but you know focus on bringing the money in it has to be a sustainable operation that's my suggestion.

spk_0 Yeah that's true that's true the other any best practices in track development that you see in your work.

spk_0 We are company successfully use artificial intelligence to improve certain areas of the workflow instead already work and fulfill record all compliance risks that we have.

spk_0 So I will maybe separate those two things out i think we have some really exciting tech in drug development we have yet to see some truly don't know what.

spk_0 Candidate in the clinic is at least as of today so that will be that will be huge.

spk_0 I think that on the compliance side that's more in your gmp space so away from drug discovery and there's emerging and evolving guidance i think from my perspective the email is leading that effort I love the way approaching it right that it's a separate guidance so you can update it more rapidly we understand that this will be.

spk_0 Rapid space so you know they're continuously working on the guidance i see an opportunity for the middle east to also lead some of those efforts and develop collaboratively with eMA and other regulators framework.

spk_0 I think on the regulatory side harmonization would be tremendous even in terms of cost reduction for the developers right if we could have some alignment across different regulatory agencies that be great so i think about compliance being key in your regulated space and then on the.

spk_0 So that's the priority there while privacy obviously is a priority but then on the drug development side there's new and exciting things I would again before you bring it back inside would think about privacy would think about reliability of those insights these are questions we don't see people asking very often they really should and then once again building in your values into the algorithms that you're bringing in.

spk_0 Let's look at the entire drug development process from drug discovery up to bringing it to the market and let's look at the benefits and risks that the eye has in this area and what it means for collaboration when we look at drug discovery at the early stages there are currently companies like in Silicon Medicine for example using AI tools to finding your targets to.

spk_0 So finding also new solutions new molecules to model them for these targets I always wondered how it works with patterns patent ability when it just feeds the data into LLM's I mean what what I thought at the beginning was this just unique models for one company but then I learned no they want to use it for more companies but then it's just sort of you feed something in their database that's really new innovative.

spk_0 But on the other hand you just close the innovation already what's the risk there how what do I miss in this area.

spk_0 Think on this patenting side show good but had a really spectacular law firm they had a really spectacular AI day at bio earlier this year and they had a really thoughtful approach with that human in the loop because aside from I think at least as of just a few years ago.

spk_0 I was just South Africa where you didn't have to assign a patent to a human everywhere else we have the human has to be on the patent so you can strategically place people along that workflow to ensure that whatever you discover is patentable and I think that's such an interesting area from the regulatory perspective from legal perspective right.

spk_0 I think for drug target discovery we have seen such immense impact of these collaborative approaches where you take a massive data set the project that I had the great fortune of interfacing with was the farmer proteomics project where at the time it was 10 largest pharma pulling together I think now it's 14 participants in the round two.

spk_0 But it's pulling cash together so you can fund these massive studies and so reading that first round of proteomic analysis of the UK biobank samples to discover I think over 10 and a half thousand new potential drug targets so maybe now we've been we've been some AI in there and which I think was happening for around two and you just feed so much more data into the algorithm you are going to get.

spk_0 So to me, drug discovery is such a big topic it's so expensive it's so difficult to do but so important so that we don't all continue with the gold rush after the one target that we see works in the clinic and then everybody floods floods that space right we want to get away from that because they're all those patients or maybe not served because we're all going after this one thing.

spk_0 So but even if you have a great model it's still going to be better if you pull the data together from across companies so to me the fact that these large players have been able to come up with a framework legal framework to share those insights in the pre competitive collaborations they do me that's wonderful alignment of the goodness of humanity with the individual company values and goals I mean it's it's so beautiful and so the only collaboration right there are many collaborations.

spk_0 That exists and again I may have mapped out maybe 18 that are between various combinations of companies but we know how to do this we have legal teams that have figured out how to share data and keep the IP I think the upgrade that we're going to get with AI and some of the tech developments in terms of decentralized data analysis that now we can keep the data private we don't actually have to put the data that we have across companies to one spreadsheet so it's better.

spk_0 It's more we can be more confidential in the way that we collaborate so I think it's just such an open and wonderful space from here on out and I think AI will play a massive role in this in the collaborative approaches.

spk_0 So it basically helps getting people more to collaborate with each other and breaking up the silos and there are some benefits when a company I also with smart contracts you can also make sure that the investments they made when it is close to data other and also returns to them when something hits the market so there's a lot of potential and it's moving in the right direction if you understood the right from what you see in the industry.

spk_0 I mean just the number of collaborations that we've already executed on as a Qtodon as an industry the number that are in progress I think are all really good indicators what I find is the challenge to propagating that collaborative approach is that in large organizations sometimes folks working over here don't know that folks over here have figured out a collaborative approach so it's it's this brand awareness almost for collaborations that some of that internal some of the external awareness that we have to do is that we have to do that.

spk_0 I mean that's what I talk about in my talks that we have a way we have a framework for collaborating and working together we just need to apply to these new domains like self therapy manufacturers one I think figuring out which organoid fits with which clinical outcome right what's the best read out most predictive of a particular clinical outcome in a given indication.

spk_0 I think that's another great area to collaborate and we all of the momentum that now is in organoins because of the FDA regulation from a gay pro I think that's prime I'm excited for us to really look closely at that space and figure it out.

spk_0 Now absolutely I mean with the eye with social media with also these new tools like zoom that we are using now we really can bring people together and help them collaborate better this is for the drug discovery part so hopefully at the end of the day we get better targets we get more solutions and we get faster through this phase that really takes a long time to pre clinical candidates and to candidates that I proved then at the end of the day by the FDA so they can go into clinics.

spk_0 What benefits do you see on the clinical stage we talked about risks earlier so no let's just stay on the benefit side what what what can I do on the clinical site in practical development.

spk_0 Yeah there's done a lot of investment on the biomarker side of things right and personalizing therapies you know from NCI match study that biomarker associated therapy selection gives six times more people benefit so I think it's the evidence is here where it's clear that we'll do better if we personalize therapies right so I think leveraging eye to map out from the end of the day.

spk_0 What are the predictive biomarkers and then finding people to deliver these therapies to again it's a balance of patient privacy rights will have to we'll have to navigate that landscape in a thoughtful way but for people who want the therapy but don't know that it exists.

spk_0 Right it could be life saving to help them even know about the existence of a given therapy with still herpes we've gone through this for a few years where you know their massive efforts to figure out how can we make doctors aware of the existence of these novel therapies how can we be patient so where of them.

spk_0 A good friend Tracy Ryan who's an NK core she's a mother of a child and she developed therapy together with the team of course to help this child survive and they've been battling this for many many years but she talked about how in the initial stages when her daughter was just diagnosed she spent weeks in front of the laptop just trying to find the right way to do it.

spk_0 So she's great that she's so resilient that after this traumatic diagnosis she can move on and be in front of the laptop and be searching that's not everybody that's not every person so I think we have to it's our obligation to make it easier to find this information and so actually conversation that I had just earlier today was about how perhaps chat GPT could be a tool to raise awareness of some of these therapies and it's certainly not focused on that.

spk_0 But maybe it's another channel that we can bring that information closer to the patients because we sort of know that everybody's tempted and using chat GPT to find information.

spk_0 This is one of the reasons why I started this podcast I mean people use LLM to get information so we need to make sure that the right information reaches the people have been to search for it and podcasts are a very good example very good way to distribute information to LLM's and data.

spk_0 Hopefully the retrieval it's then when somebody's using it but what I thought when you were talking is I experienced the farmer industry as an old fashioned industry I mean it started somewhere in the 1800s with the industrialization and then was designed basically for one concept over decades to develop one solution and then deliver it to many patients.

spk_0 And now what you said my understanding is that we have a really paradigm shift in the industry that we move away from this we want one molecule, put it into one pill and then just chip it out on the market to a more personalized approach is farmer really ready for that structural.

spk_0 I think that's an excellent question and I would love to bring items to get some folks from our district to this I think there's such an ethical side to this question right if we know that a personalized therapy will work better for the patient.

spk_0 It is almost our obligation to make it available to them to help them find it and to deliver this personalization and and that maybe propagates also to investor side who all want maybe to see a blockbuster and I understand as a commercial person I understand this but it's also that if you know that a tailor drug will work better than a blockbuster can we shift the way that we think about investment in the industry.

spk_0 In this space to have math that takes into account pure dollar value ROI but also this humanitarian ethical aspect and I love the way that RobExo's trust reform Excel is talks about it where he says that we are in health care not money care if you were purely focused on the cash money cares the finance industry right that's the place.

spk_0 But in health care it's more complex in equation it's about improving people's lives and I think the companies therapy developers when they fund various right of course they're struggling with the hockey stick how do we get hockey stick work.

spk_0 So I'm working on an article with Ryan Murray from a val source hopefully we'll be coming out here in the in the next two weeks on ethics side of it right how can we think of this perhaps someone differently than other types of investment.

spk_0 What's what's your concept to rethink investing in the end of the day investment is always about making money so this is a precondition not for an ethical reason and not because people are greedy but at the end of the day you have to reinvest in the next generation so if you don't make a profit from the old generation at one point in time you lack simply the capital to reinvest in the next generation.

spk_0 So this is a profit driven industries are profit driven for a reason to be able to improve the quality of the entire system to you have an idea how a new concept could look like.

spk_0 That super question I think once I figure that out I will sell that and I will be off the buying time.

spk_0 I was hoping for you.

spk_0 But I think some of it right is is this just the complexity of the math and then that if you so more of therapies that are less efficacious the actual impact on the patient population might be not as good as if you sell smaller batches but are very personalized therapies and we're still moving to that right we don't really have to make an argument for this.

spk_0 With all of the novel types of therapies with cell therapies it's it's very clear that it's so much more precision medicine oriented so I think we have that understanding it's just how can we how can we get away from the hockey stick 10x which is just such a difficult slide for companies to build as they fundraise so I think it'll be interesting to see how this evolves.

spk_0 And the hockey stick 10x is driven mostly by the high failure rate in the end of the day when you need to invest money and get a high heat rate of failures you need to be a crystal flanger brought in in his speech spring is the example that from 10,000 molecules.

spk_0 Yeah only one hits the market at the end of the day but you need this 10,000 number to find the one.

spk_0 I think this will change because of AI right because you can de-risk and make that pool smaller in the beginning with all the predictive tools and there's so much tech being developed on that front to reduce essentially reduce failure rates that even before we get to humans we have so much a better understanding of the toxicity profile and the binding profile right all of all the different pieces that routine testing that we can approach from that predictive.

spk_0 So I think that that equation of the 10,000 will change this is this is something I'm really curious to see in the future the big question when you look at these two approaches you have this sick signal approach but you have just an existing product and make it better and what you do there is reducing the failure rate which really works well you understand the process you understand the product and Elon Musk is an expert in that to just improve and see what you're doing.

spk_0 So you can just streamline every single process and take it out but then you have this creative process where really find try to find a new solution or identify a new problem and they are naturally the failure rate is high.

spk_0 And when it came from artificial intelligence companies they're okay let's reduce the failure rate in the creative party as you can do that but in the old system you also risked that you reduced the hatred I mean what you want is basically the one finding the one molecule that works.

spk_0 It's one of the things that you want to invest in the one molecule and one to drive down the investment in the molecules that don't work but in the old system without the I this 9,999 failures were necessary to get to the one when you reduce the failure rate then you also reduce the risk to find the one molecule that works in the end of the day and what you end up with is 100 potential candidates but none of them works.

spk_0 So a couple of pieces to that I think it helps us separate that into multiple phases right so in the discovery side it's I think in many ways the same question as we discussed earlier for this is your 10 20% exploration space your.

spk_0 We have something that we think will work about can you work around explore that design space and more to find new candidates identifying new targets and so you give more of a leeway to the algorithm the development side I think is predicting development ability for example we've done this for small molecule to some degree we have a standard and proprietary algorithms for predicting the availability of maps.

spk_0 I would love for us to get to this point with cell therapies but I think it's in the end it's we could have a checklist make you've discovered some new compound modality and so on you have a checklist user 10 criteria you have to fulfill in order to be able to manufacture it so then your development can be much more streamlined if you know what these criteria are on and so you won't end up in a space like we've hit recently right where you're in the clinic but you finding that actually manufacturing the girl.

spk_0 It's not it's only some of the time that you're able to do that I think that this whole this this part we can figure out and to me that's a very collaborative space because we're all going to face their similar challenges and manufacturing just given the biophysics biochemistry of the molecule that we're working on.

spk_0 Now just moving in the thing to potentially is the automation and just doing this small things can also the going for the quick wins can really when we look at the market side in the other therapies on the market but doctors are usually not aware of that they have to treat patients and no time for education so if you can improve that you already benefits the patient personal is a personalized medicine if you just can find the right medicine for that person with the eye it's perfect and when you look on the track development I'm not so sure.

spk_0 I'm sure that we really get down from the fair rate at the end of the day but can run much faster through the entire process I don't feel I'm not convinced that really should aim at reducing the fair rate fair is always high fair is always a

spk_0 a marker of high innovation and then innovative areas of fear is good in the space and I would really not like to just try to streamline something that is necessary for innovation and get away that you can think about reduction of cost and reduction of

spk_0 pipeline for that is you're doing this in silico so you're still trying a bunch of different things but it's much cheaper to do because you don't actually have to pay for the media biorectric all these things right and there again lots of companies working in that space fewer than I would like to see but there's some really interesting innovative in silico early PD

spk_0 I'd like first to have deeper understanding on the actual my proper manufacturing side so we will get there as well but I think in this new age will be able to condense a lot of the lab work by expanding our predictive capabilities again reliability will be key in that part because if you're predicting a process that then you discover

spk_0 two years down the road doesn't actually work it worked great on the computer but in the lab it doesn't work right so that then you're you've lost a lot of time and money so reliability once again in our space is just key so as you're bringing in an Encelico modeling provider ask them questions.

spk_0 Yeah and we need more we see in that space what willing to take risks at the end of the day and bet on these companies and in silico is I think a huge opportunity but we have not enough investment capital in that space I'm helping in what what is your opinion on that.

spk_0 When I looked at investments it seems that a lot of money has gone into AI on drug discovery side and patient stratification therapy personalization from that perspective I think this development side maybe it's not perceived as as big of a cost which is maybe why there's less investment going in there but to me when I look at it in much more let's say stream line stream line similar operationalizable so I think it's a lot of money.

spk_0 It's actually a prime area to be developing technologies in we see some of the big players making investments there would love to see more more BC mining going in I think my assessment for why this has been an underfunded area is that we kind of want to solve the biggest problem the biggest value the biggest cost driver but so I basis off of this book how to be a billion how to be a millionaire and that whole book is about how to be a million there.

spk_0 So I think that's a lot of save in tiny little pieces it's not how about how to make the massive paycheck so that you save a little bit and a little bit on the end you have a lot of money so I think of this development side is that we're going to save a little bit here and there but in the end acceleration that you're going to get the cost savings are going to be significant.

spk_0 I think it's a say sorry to interrupt you but I think it's a sales problem from from from the company side not so much from the VCs I mean VCs site in Europe is underfunded so this is a regulatory part but what I see from the company side is that they not sell what we see are looking for at the end of the day VCs always need to make this are looking for his 10 x returns this is pretty simple it's the mechanic of the industry because they want to take high risks high risk of failure any day you return what I remember.

spk_0 I really see when I talk with founders of AI companies for example you brought the example with

spk_0 I'm a entire lab models run in silica so you don't need a red model or a dog model or an email models in general you just run it on a computer in the cloud get the results and can move on and FDA seems to move in that direction

spk_0 the company that nails that can really say we can simulate the entire drug discovery part and pre-clinical development part and you get in no time a candidate out of this that goes directly in the clinic creates an entire new market captures the entire market but really an AI company says that to investors and say look I mean what we do is remodeling the entire market and when we win and when you make us win we throw everybody else out of the market

spk_0 that operates in that space I think it's a sales problem at the end of the day when I look at the sales pitch is the mostly have it with a little bit of here and a little bit of there and it's like a project sales part but we see I really recommend reading the power of for example Britain passipastian Marlop he brings a lot of excellent examples how to take started

spk_0 why we see invested in China tech and this is basically the blueprint for how to sell to vc's and 20 founders are not convinced they created an X case for a vc they likely not get funding and then we have the problem that's they are not funded and then the solution doesn't hit the market and then they need to go on the customers but then you have also the problem that it don't want to sell to customers so it's a little bit the same problem of both ends

spk_0 and also it's a you know when they go to customers I find that the language is still very messy it's very algorithm focused but which is understandable right the founders is some tech it's on level technical in AI but then you have to translate that language the customer there are biologists most of them some of them are data scientists but they're still not AI engineers and I again this is not a novel problem in any way I have worked with lots of mechanical engineers who also when they make a firm

spk_0 and they're not going to fly about their new catbacks it's very focused on that engineering language and it takes some iterations to get away from the build and more towards the impact and I think everybody knows they have to do that but it's it's hard to do for yourself when it's your own product right so it's the value of bringing somebody who's a third party and kind of sits between you and the customer that translator I think is a mess

spk_0 but then they need to learn it's a sales to investors it's also sales problem is always this 10x they need to deliver returns to get investors it's pretty simple so when you found out doesn't bring this 10x story then the deal doesn't close with the investor the same on the customer side what you said a customer needs to solve a problem and if the technician doesn't connect with the problem of the customer the customer won't buy the solution so it's pretty much the needs to both mindsets in the needs to get people on the teams

spk_0 who love doing that yes I completely agree it's really it's really time flies when you have fun we are 10 minutes over our two hours thank you for your time no I really love that before I come to the final question is there any topic that you want to emphasize it's the end of our our livestream

spk_0 I don't big themes right one I cannot say enough as a doctor of AI tech asking questions and really thinking deeply about the privacy confidentiality of your data and then if you're in the regulated state think about compliance and thinking about what are some of the next things that the regulators are going to do

spk_0 are going to ask for inevitably in six months 12 months and so on thinking about implementing your values to make sure the suggestions that your employees get that you get from the algorithms reflective of your core beliefs

spk_0 and going to the personal side protecting yourself from the malevolent players out there they are out there whether you believe in their existence or not so in the same way thinking about your privacy of your data making yourself less vulnerable

spk_0 and there's so much research into the very consistent steps that these people take in order to get you to do whatever it is to serve them so information's out there I could point anyone who wants that but we have to take advantage of the academic research that's been done

spk_0 and then you know bigger picture I love for us to be more collaborative in this industry I think the progress that we see with other industries where they're able to standardize and collaborate I mean it's it's just immense so like for example I tech right how fast they were able to move because of open source code I think for us we're

spk_0 lagging in that and I really hope that this difficult economic climate can help us think outside the box because we still want to develop therapies that will save people's lives so collaborating will offer a cheaper and likely faster way of getting those therapies to patients

spk_0 absolutely and I hope I mean if somebody thinks bad actors don't exist what recommend going on social media online for an hour x.com for example and you see a lot of back and forth between people so they exist but when we leave this at the beginning of the conversation and now look into the future and this is my final question for you

spk_0 the next 10 years I mean we are now in 2025 in 10 years we are in 2035 and there is a lot of potential out there with AI with robotics with all these innovations that are currently in development what's your most optimistic scenario when you can think it through how will biotech and farmer look like in 2035 when everything happens according to your plan and you have unlimited money

spk_0 and can make this one scenario this best case scenario really happen in 10 years how does the future look like for you

spk_0 yeah so if you get the core is that we have gone through these 10 years without massive losses of information by specific companies we have gone through this without a collapse because we were thoughtful and really forward planning in our privacy confidentiality compliance et cetera strategies

spk_0 I think in this 10 year view I would love for AI to become a tool technology to implement good things to amplify the good behaviors good part of our population so that I kind of almost think of AI and humans as being one populist maybe in some ways in the future I don't know if it's 10 years or longer

spk_0 but I would love for us to reduce that you know average malevolent player rate of 5% down by amplifying the good algorithmic side and then just this collaborative approach I think will help move the industry so much more better reliably right if you even think about measurements that different companies are taking in different ways if we can analyze across the different ways

spk_0 the outcome the inside that we're going to get is going to be much more reliable than the one person with magic hands in your lab who can execute on this experiment and no one else we both seen this the type of experiment so I think reliability will go up the cost will go down the time lang will go down if we can bring ourselves align ourselves better with that overarching initiative of bringing their best patients we have done this before so I know we can do it we can do this more often

spk_0 I enjoy this conversation a lot I like your positivity and I love your balance approach so not on the one hand you are not only positive and looking to build a better future and bring people to a more collaborative mindset but you also have a lot of risks and not everybody is a good player that there are also bad players on the market and wants to help companies to understand both sides the positive sides what AI can do for them but also the other hand make them a bit of risk and they love this balance approach that you bring to the table

spk_0 thank you thank you life has made me balanced in that way I have to most of my choice but but I think it's a more realistic perspective that will enable more actual stuff

spk_0 yeah absolutely agree with that thank you very much for your time and for this amazing conversation and let's catch up soon

spk_0 thank you thank you so much Christian thank you everybody who's doing I did really really appreciate this and please reach out on any of the topics where any of the other topics

spk_0 was such a pleasure thank you for the great questions thank you and have a great day and talk to you soon bye

spk_0 we have heard today that artificial intelligence isn't just technology it's trust privacy and values encoded into code

spk_0 that showed us how the wrong use of data can derail entire companies but also how the right frameworks can transform medicine speed up discovery and bring therapies to patients faster than ever before

spk_0 and here is the real insight AI will not replace humans but let us who know how to work with AI ethically, strategically, courageously

spk_0 will replace those who don't if this conversation sparked something in you here's one simple action please hit the follow button

spk_0 I've seen the stats today and most of you listen without following by following you help this show grow and you make sure more rare deep tech voices like cuts reach the world for free

spk_0 and if you found value here don't just follow come back next week we have another remarkable guest who will challenge how you see the future business and yourself

spk_0 because this show is not just about listening it's about learning how to lead and build the next generation of great deep tech companies in a world that's changing faster than ever

spk_0 have a great day

EP 164 - Kat Kozyrytska: AI in Pharma Is a Yesterday Problem – Why Ethical Frameworks Can’t Wait

Interactive Transcript

Topics Covered

Related Episodes

Out on a Limb | Nov 2nd | Hillhurst United Church

How Do You Respond to Jesus? John 5:16–47 :Daily Devotional, Daily Bible Study

Episode 240: Otto Aerospace CEO Paul Touw talks teardrop, laminar flow, and Phantom 3500

Burnout, Breakthroughs, and the Power of Getting Back Outdoors | Gun Talk Hunt

Share Episode