Dr. Tom Mather: De-risking Drug Development with AI

Ram Yalamanchili (00:04.29)

Hey, Tom, how are you?

Tom Mather (00:19.245)

Great, happy to be here.

Ram Yalamanchili (00:21.42)

Very excited to have you here. A little bit about Tom. Dr. Tom Mathers is a physician, entrepreneur, investor, my favorite, we describe it as a national level champion athlete as well. So quite a bit to look up to out there, Tom. So I'm very glad to have you here and talk about your experience in biotech and sort of building the space.

So let me start with my first question, which I'm really curious about. What's the journey been to be where you are? You've had a very successful career, it seems like, in multiple different areas. And can you tell us a little bit about how you started and where you are today in that journey?

Tom Mather (01:11.085)

So I'm a board certified ophthalmologist, oculoplastic surgeon, and I enjoyed very much the clinical side of medicine, still do, I still practice clinical side of medicine for 42 years. And I'm always looking for something that might be out there new and different. And one of my friends said that maybe I should think about looking at pharma development, that maybe it would be really

good for me to consider that. And I started working in the farmer development field in 2010, with various companies in various different roles. And you kind of stair steps away and you start learning more in the field and gain experience, gain knowledge.

Currently, I am Chief Medical Officer for MCAL Therapeutics, which is a clinical stage company working on a treatment for evaporative dry eyes. It's got a novel, straight forward approach to lipid replenishment on the tear film. I also run a venture capital firm, which is called Optifund, and it's dedicated just to

ophthalmic pharma and devices development. And that's, it's been really interesting to kind of leverage my clinical knowledge and gain other aspects of knowledge in other fields. It's really interesting how that deep base of clinical knowledge really pays off. It's, hard for anybody. It's hard to duplicate some of that knowledge. It's hard to duplicate the knowledge that you learn in

farmers and development, the people that have in there for 40 years. It's also hard to duplicate the knowledge of what you learn and trying to evaluate these products that would get developed and devices that get developed. But you have to evaluate them to see if you want to bring them in your practice. It's interesting, the confluence of all that.

Ram Yalamanchili (03:35.04)

sense. something I'm curious about is since you've taken this approach of being coming from a clinical practice into being an entrepreneur now, potentially an investor somewhere along the way, what are some assumptions which you probably made early on before you made the next move, which maybe turned out to be wrong at this point, right? Be curious what's some of the learning there.

Tom Mather (04:01.185)

Well, the learning over all that is phenomenal. mean, you ask anybody that starts a career, when you start as a physician or start at, have a certain amount of preconceived notion that this is how it's going to be. Well, it's not exactly that. And you kind of learn the new way. in pharma development,

It's much the same thing. I guess that the most important thing I sort of thought of is how disjointed the whole field is. And that it's all these little islands that sit out there, the people that come up with the drug or create the IP and the money and the development, all that is so disjointed. And those various parts don't really relate coherently with the other parts.

And, and you look at that, you're like, wow, I really thought that this was going to be more coherent. And the more you get into it, the more, the more you realize it is even less coherent. My appreciation now is that it's not getting more coherent. It's getting less coherent as it, as it works along.

Ram Yalamanchili (05:16.91)

That's interesting. So you feel there are more silos than you would expect, essentially coming in into this field.

Tom Mather (05:22.475)

Yeah, and sometimes the silos are not even connected. can't even, there's not a walkway between the silos. And to get one silo to talk to the other silo, it takes a lot of effort to get the chemists to talk to the development people, to talk to the CROs, to talk to the strategics. It's not like they, it's a challenge to get all that accomplished.

You never get more than two of those entities in one room at one time. You might get two, maybe you'd get three, but you're never gonna get five. Well, they put together a meeting, but they're never there to have talked to one another about a specific drug or product to help move it along. That just doesn't hardly occur.

Ram Yalamanchili (06:12.91)

I find that really interesting and I'd love to bring this up in a different context later. So coming into this field, you started off as a physician. Was the next journey to be more on the investor side or was it as an entrepreneur biotech? How did the next step in the journey go?

Tom Mather (06:36.192)

Well, working with the companies, you know, I would say that I was a started my own practice and started a, a, surgery center. So I would say that early on people, physicians that run a practice to run their own practice are entrepreneurs so that they were, they've started a business, started that. So you already have that and you're already working on those tools that you need to.

run a successful business. And so that was somewhat already embedded. And then as I became more acquainted and gained more knowledge in the pharma field, then it was pretty an easier step then that, well, I would like to be a little more, have a little more entrepreneurial spirit because that's really where you come from is that so many parts of the medicine that I had.

was all started, everything that I worked on and had, I started myself.

Ram Yalamanchili (07:41.358)

Got it. So, you know, as we get into this further detail, right, one of the things that I'm particularly interested in is understanding the role of some of the new technology coming in, in biotech. But before we get there, I think I'd love to find a few minutes to discuss what has been a lot, been hearing a lot in the market lately, which is how biotech funding has been a challenge over the past several years.

And since you've got a really interesting vantage point being a venture investor as well through Optifund, how would you describe it? What is the vantage point you have around what's happened over the past several years and maybe even why that's happened?

Tom Mather (08:30.018)

All right, well, you do see a significant less participation of the venture of capital funds in the development of of ophthalmic products or all medical products, but particularly ophthalmic products in the maybe from five years ago to a year ago or six months ago. There might be a little thawing on that. And some of that is just the our capital structure and what's going on in the capital markets over that.

interim period of time. But it's also is that I think the venture capital companies really wanted to have a little less risk that they were looking, they had a taste of it and it's a risky proposition. And there are other less risky propositions out there they can invest their money in. so they really were looking for

products that were de-risked for them. it's not, it's it's not, I mean, that's understandable. Somebody wants that, it has a little bit less risk. There's still risk involved, but they wanted a little de-risk. And so that's where they've sort of set back and maybe looking more at products that have been farther on the development process, more in the phase two, phase three than in seed. And just want to take less risk.

So I do think that if we want to have more venture capital involvement than as an industry, we have to do what we can to de-risk those and understand that they wanted a better, they wanted to be less risky for them to invest in that. And I think we have to give some thought to that about how we can de-risk the programs.

extra work here and there, better animal models, better pre-clinical studies, more intuitive ways or more advanced ways of looking at the data. All those will help de-risk that. All of them have a, I wouldn't say that each one of those things that I mentioned is going to completely de-risk the process of the development of the product. But all of them together can make a

Tom Mather (10:56.63)

significant improvement in the risk factor. And then I think we'll, and I think that as the capital markets do better, which everything goes in a cycle, if we put both those together, I think that we'll, that the capital markets will improve, that our investing in VC components of, or VC involvement with the development will come back to a port.

Ram Yalamanchili (11:23.982)

So there's basically these, I guess, two trends which I'm hearing about, right? One is the macro, which is the cycles in the market itself. And then you've got the risk reward, which is involved in early stage development. In terms of just looking at the latter factor, the risk reward, right? Has that meaningfully changed? you have a vantage point on what's happened over the last decade or even longer? And you sort of zoom out.

What's your perspective as being a very early stage investor? And just for our audience, believe Optifund is very early, frequently the first check into companies, is that right?

Tom Mather (12:02.498)

Some seed, we also are advancing a little bit along that process, a little bit in looking for a little less risk. And we grow a little bit larger that we wouldn't pass up any opportunity to do something that has a little less risk involved to it. Just like I'm saying to all the other larger VCs, larger venture capital firms, I understand that process. But yes, somewhere between early on and

Still early-ish.

Ram Yalamanchili (12:35.448)

Got it. yeah, going back to my question, how do you see the risk-reward part of the equation, right? Has that materially changed in the last several years? And what would that be?

Tom Mather (12:53.013)

Well, I actually see in the last several years, because we brought on other avenues you're working in that are fairly risky, like gene therapies, long runway, a lot of money. In some ways, you could say that risk reward has gotten riskier in that period of time. Now, the reward is there. Those drugs and all that are needed and will be valued by patients and will bring

important changes to their life. the reward is still there, but boy, I think that the risk has kept up with the reward and perhaps exceeded it some proportionally. You should look at that. So yes, we still have some work

Ram Yalamanchili (13:39.598)

That's fascinating. I actually find it really interesting in the sense that what you're also saying is the new modalities and new discoveries have actually contributed to an increased risk versus potentially what it was for a period of time when things were potentially much more de-risked modality. Is that actually how we look at it?

Tom Mather (13:58.818)

All right. Well, and I don't want to keep beating. don't want to beat up had gene therapies, but they are. That's a long program. And it's just a good example of a long program. And if you start out early with the gene therapy, to think that you could get to the point where it could be exited or completing the, that that's gosh, that's easily 10 years the way that it's currently working in that. And that's a long time. That's a long time to

access funds. It's a long time to have that work along. It's that's that's a lot. And so

That gives people more pause as they look about trying to fund a gene therapy because it can be a difficult task.

Ram Yalamanchili (14:52.386)

Yeah, I mean, I guess there's significantly higher amount of risk if you're planning to continue to fundraise over a 10 year period. And maybe even then you have to think about the cycle, like the first part of the discussion around macro rate, what happens if there's a market headwind or something like that. So, you know, just fast forwarding, let's say we go five years out. Are you then saying that because we have all this interesting discovery coming out right now, there might be new modalities as well coming out?

Does that mean we're looking into a future where at least this part of the risk equation continues to get more complicated?

Tom Mather (15:31.007)

Well, it could get more complicated, but it could also get simpler. So I think we have to, we as an industry and among people, we have to understand that there are probably ways of us making this simpler and more transparent. Right now at best it's translucent. And a lot of times it's opaque to the venture capital group that are trying to come in and try to understand a product. and there are probably ways that we can help that.

But yeah.

Ram Yalamanchili (16:05.018)

When you say it's hard to understand or opaque, what would make it better? how would you see a venture firm, I guess, getting the right insight to be able to do better?

Tom Mather (16:19.297)

Right. Well, I think that gosh, you way back from the very beginning. I think that there are probably ways of us evaluating a given IP more thoroughly and with the use of machine learning that you can evaluate that potential of the drug. When you put it into nonclinical early studies, the

animal studies, there are probably ways of us again accessing machine learning so that we can do those studies more efficiently and that we get better data and that if there's ways, is there a model, are we really using the best model? We kind of in our own mind think that we know the best model for different entities, but do we know the best model? Are we using that model the most efficiently? And all those will then produce

better early data, more reliable data, and a little bit more backup on all that data that I think will de-risk to some way, certainly make it more attractive for investment because there's more backup than we have these couple animal studies and gee, didn't it do great? Well, for a venture capital firm that doesn't work,

works only part-time or even in all of biotech. It's such a huge area that for them to have the expertise in one certain area to be able to understand all that can be difficult. But if you're able to review those studies and that early study and what the company has proposed as their molecule, their fix, that if you

we have the chance or developing a chance through the use of the machine learning to be able to have more credibility to what we're proposing.

Ram Yalamanchili (18:24.344)

I got it. it's essentially like turning the page, right? As you sort of look at the continuum of data collection, the more data you collect or the more data you could show on the program, you're essentially giving a better vantage point or better perspective on where the risk lies. Is that hard to think about it?

Tom Mather (18:43.232)

think it's better data. don't know if we, you can get more data and certainly machine learning can give you more data. But what we really want is better data. We really want the, a fresh view of is this really, are we doing it correctly? And what does this study, what does this really mean? And we're close to that.

in machine learning. We're very close to being able to do that. Even in the early studies to kind of help us evaluate those early drugs more effectively.

Ram Yalamanchili (19:23.448)

Got it. That makes a lot of sense. And then you sort of get into the paradigm of first in human, let's actually go run a clinical trial. you know, I guess it's the next step of de-risking the asset, right? So thinking about that process, can you tell me a bit more about your journey at MCAL? How did that transition happen? What were the requirements before you were able to fund the asset, fund the company itself? And

Essentially convincing investors to say yes, like there's enough understanding and guess there's reward here to be part of the journey.

Tom Mather (19:58.273)

All right. Yeah, so the company itself would have to prove that it can make the drug, get your CMC right, and then you are then looking for how you can show that in the dog studies that it would be, it's gonna be effective in evaporative dry eye. And then from that, if you have that data, what's nice about

dry eye is that a lot of the people out there in the world and the population, a lot of people are affected by it. And there's a good bit of knowledge on dry eye in the general population. So even in the venture capital firms that there are people that suffer from dry eye and that they can relate to that. So it's an easier catch on than an orphan drug that they may not have had any exposure to in their lifetime. And you're asking them then to make it

decision on it. But back to MCAL, so you have to go through and meticulously show how you can make the drug, that it can be made safely, and then you have to show that it can work in the available appropriate animal studies, at least what we consider to be the most appropriate animal studies at that point, evaluate that, and try to learn from those studies, which is always interesting, to try to learn from that and try to

from that then make the next step the guess, the hypothesis of how it's gonna work in humans and what dosing is gonna work the best and how you're gonna apply that and then create your clinical trial, then make a protocol and envision your first human trial.

Ram Yalamanchili (21:52.846)

Given the process you just mentioned around, certainly it seems like manufacturing and the CMC process is really key. It sounds like that's a big part of the early decision. And once you're past that, then you're talking about your trials and how you actually run the trials and perhaps the investment required to do that. Something which I've been thinking a lot about recently is been hearing and reading a lot about new types of

AI models which are able to predict better protein structures or new discovery of molecules becoming easier and a world where we might have quite a few abundant molecules which are potential cures. How do you see the investment landscape changing if we did have such an abundant set of targets which we now want to take to the next step into the trial space?

Tom Mather (22:50.41)

So once we adapt to that, I mean, you have to kind of like realize that as you start bringing in more machine learning and you're going to propose a drug, well, instead of getting proposed one drug as we currently do it, or one that's associated with something we know in life, that you could get proposed drugs that we've never had any exposure to and that we don't really know how to make. And so...

Once we sort of get used to that, the increased number of drugs are gonna be proposed, increasing molecule is gonna be proposed, then you have to understand how you're gonna be able to make those. And I do think the machine learning can not only propose it, but I think the machine learning will be needed in helping it. Cause you might have 10 candidates that might have a chance to work, but eight of them might be really bedeviling and trying to get through the CMC. And two of them look like it's a little bit could be.

through machine learning, perhaps a little more straightforward way of producing the drug. So when we contain that, mean, certainly we're gonna, and not too distant future, there's gonna be a big flood of proposed drugs, and then we're gonna have to kind of.

corral that and figure out which ones are going to be the best ones to work on, again, through machine learning, that's going to be, have to get used to that. And then from that end, then how we're going to best test that drug in its development and then how you set up your trials from that. so we're going to go through, I would imagine, I mean, I don't, it's, one could see that the very beginning is going to, we're going to be

trying to drink out of a fire hose at the very beginning. But through machine learning, we'll be able to calm that, siphon that down a little bit, decrease the pressure a little bit, and be able to manage the fire hose delivery of that in an effective way. What the end result will be will be better drugs, better drugs, more efficient ways of working through how to develop them.

Tom Mather (25:12.415)

All that will be able to be done.

more drugs, better drugs, more efficiently evaluated, prosecuted, know, proposed work through. And we will be better for it in 10 years ahead. For 10 years from now, it will really have

Maybe transformed is too strong of a word, but it will add a lot to the inherent value because we're going to be able to treat patients better. It will really improve that. also along on all that, that you will have a better process to even on the early drugs to predict their success.

And you'll be able to share that review with money sources, venture capital. And I think that they will become more comfortable at an earlier stage in signing on to undertake the development of the drug.

Ram Yalamanchili (26:32.152)

me, this sounds really fascinating. It sounds like there's a lot of interesting stuff going to be happening in the next couple of years or least the next five to 10 years. that how you feel?

Tom Mather (26:41.289)

Sure, mean, the whole thing, you don't have to look too far back to think that, gosh, chat TPT was just, we've had it three years. We've had chat TPT for three years. And before that, if you mentioned that, or I would say that the penetration to the population was not very much. Now, how many times a day do we all use

chat CPT, perplexity, our own sources of AI, we use it all the time at this point. And that's just three years in our lives because we realized that it's really transforming our life and we're in the very infancy. That's gonna happen with pharmaceutical development. think biotech development is so ripe for all of that because it's so vast.

And all what we do now is so human generated, so many human needed that it's just too much for the humans to be able to go through. if we can, well, when, not if, when we corral that, that yes, it's going to be something great to watch. I'm happy that I will get to see it in 10 years. I mean, if I have any sort of life expectancy, I can't wait to actually witness it.

over the next 10 years, it should be very exciting.

Ram Yalamanchili (28:11.246)

I feel the same way. It's more coming from a tech career, especially in the AI space for two decades. I think this has been the most exciting time to be working in it. I'm so glad that I'm actually professionally involved and working in the space. So I second that. So one of the things I want to talk about is we've spoken quite a bit about the early stage de-risking, mostly through better data and many tools which we are talking about coming out into the next.

So leaders are maybe already here, but there is still the bottleneck which shows up later and I've frequently heard that the clinical trial process could be a fairly high challenge to sort of like get over, right? What's your experience been from that perspective? You you're obviously investing in companies which are going into clinicals and I'm sure you have to evaluate the risk reward from that perspective as well, right? Outside of just CMC and potentially the early stage data being right.

I'd like to kind of get your view on how you're thinking about that and what's the future going to look like if we had what you just spoke about, this fascinating period of great discovery, lots of innovation. What's the next step in terms of what we have to be prepared for or tackled?

Tom Mather (29:30.079)

Right? So my thought is that the currently the biggest bottleneck is running a clinical trial. Financially, it's the hardest bit. get all those pieces together and to make it through that, it's the largest bottleneck. because you're not running one clinical trial, you're running a series of clinical trials. And to develop the

drug all the way through. if you, when you stack up the cost and what it takes in human labor to get that to the end of that, it's staggering.

And I think that again, the machine learning on that process is going to be delightful to watch because so much of that that we do now is hand labor, hand thing, just even even to start. You want to write a protocol, right? So you have to hire a writer. You go through these details just to get through that. Well. That's frustrating.

That should be, I want a clinical trial protocol on A, B, C, and D, press a button. There it is. It's just, and I know we'll have to evaluate it, work on that, kind of work through that. But right now people will look up what a clinical trial has done before. They'll cut and paste pieces here and there, trying to put it all together. And it's onerous to get that. And that's just writing the clinical trial. And then he goes through the process of,

selection of sites and selection of inclusion, exclusion, criteria. Well, those are all based on what we recall from the last study or this, or what this person said about it. And the data is at best soft. And that data needs to be hard data. We need to be able to go through and say, if you want to do a clinical trial,

Tom Mather (31:40.872)

and you're gonna have these inclusion criteria, well, you should be able to toggle down. I'm gonna change this criteria model a little bit. I'm gonna see how that's, is that gonna affect my enrollment? Who's got those patients? How is this all gonna work out? All that. And so what that does then is it should make that process so much more efficient and so much more effective that the ultimate cost, big barrier,

will come down, that's just not that the CRO companies won't be profitable. They'll be able to profit and doing a better job because it's gonna be more efficient and more effective and the product is gonna be better. Going through the clinical trial, it's gonna be way better than what we can now.

Ram Yalamanchili (32:37.432)

Yeah, I can see that. And I think you make a really reasonable and logical sort of like argument around better discovery leads to more number of shots on the goal, more targets discovered. And the next part will make around everything you just spoke about on the clinical process, right? Identifying sites, creating your first protocol, creating your first ICF, your informed consent and...

All this work, I think, has a fairly high activation function. And maybe just like how you're talking about using machine learning to reduce the risk and the burden to generate quality data, if you could do the same thing across various parts of the further continuum, then you sort of overall make the process much more efficient. And I think from that perspective, what do you think or what do you say when...

Frequently I've heard this term that the clinical research industry is a really slow industry in terms of adopting technology and maybe that's broadly applicable even to biotech or pharma in general, right? So What are you seeing? What do you expect to happen in the next five years? You know, Clearly I guess the reason I'm asking this is you brought up chat GPT chat GPT is one of the fastest growing products in history They've gotten from zero Active users, I think just three years ago

probably just about three years ago, it November 2022, the launch time. So from there to today, I believe somewhere in the 800 million active users per month. So that's almost a billion people using these tools to interact with an intelligent model. So I'm just curious about adoption, the reality of how this diffuses into the actual market we're talking about. And of course,

The idea is that how does this finally affect the ultimate reward recipients will be patients and us, Like us as humanity.

Tom Mather (34:41.758)

All right, so medicine in general, pharma development has the tendency to be reasonably conservative and that's a great thing. It is, medicine should be conservative. We should think expansively, press ourselves, but as far as what we're delivering to the patients, we wanna make sure that it is as safe as we possibly can do it. So it should be conservative. And that is often

Those characteristics are often embraced in the personalities of the people that do that. And they will by themselves likely to be a little hesitant to embrace new technology. However, they're smart, adaptive people, and they might be a little bit slow on the uptake a little bit, but once it takes on and gets it,

Nugget, it's a foothold in there. It will take off like wildflower. It will run through that if you are not, if you're not running a trial that is amenable to...

Tom Mather (35:57.296)

machine learning and to go through that. If you're not doing that, you're going to be a distinct disadvantage. And is that three years, maybe five? Probably. That it just has to take it. People will want that. Sponsors will want it. Investors will want that. They will want to know that because again, we get back to the product that we do these

We're gonna do these series of trials and most of those trials have similarity between the first one, the second one and the third one. Well, besides being able to evaluate the first trial while it's ongoing, if there's some aspect of having trouble, but since you're gonna essentially be repeating that nearly that exact same study as you move along, wouldn't you wanna have data, not only the data of the study, but the data about the study?

So are there aspects of that that we should be doing better that we could do more efficiently and effective? And so if you're, if you are collecting the data of how this study actually is run, you can review that data and you can make your second and third trial to be efficient and more efficient. so if you typically have 200 patients in the first trial and 300 and 400 patients as you go along, well, conceivably.

You could do all those studies for the same amount of money because you will have learned efficiencies as you go along and that it wouldn't cost you more money to do the 400 patients because you would have learned what you went on when you did 200. You were paying attention. You watched the putt.

Ram Yalamanchili (37:41.332)

Mm-hmm. In fact, this is something which is the second time you're bringing this up, right? I think you started off with one of the surprises for you was how siloed the market is, how little context is being shared between teams. And in this particular example, between studies, right, you're going from these different studies and perhaps the learning could be better so that you don't need to do the same type of investment or the heavy lift on the other side. So to a certain extent,

Is that the bet here? Is machine learning essentially providing a way to reduce the barrier of sharing? Because I can imagine a machine can share much easier to other machines than us as people can. The amount of context I can potentially get out of you or the knowledge I can gain out of you in this particular podcast itself is much more limited than what a machine could probably do in the same amount of time. And that's one of the properties of these large language models.

things like that, they're able to go through vast amounts of data, create patterns, create understanding, emergent behavior comes out, and then we're finding them to be what we call intelligence. So I'm just curious, is that one of the learnings or bets you're trying to make here? It almost seems really like you came into this industry with a certain assumption only to be proven potentially otherwise. And maybe right now you're seeing some ways to of like solve for that.

Tom Mather (39:08.669)

Right. So, I mean, maybe it's my own wishful thinking that this is how it's all going to play out, but it makes sense to me. It does make sense to me that you will be able to those different components of all the people that produce the drug, that you will be able to together more coherently. The machines that you use will be able to work more coherently with the other machines that someone else is uses.

Ram Yalamanchili (39:13.157)

Ha

Tom Mather (39:38.213)

And I know we have to be careful with a certain amount of confidentiality, but that can be built into the product. And I think machines can be perhaps more confidential than humans. I think you can control them. You can get that to where they're more confidential and more protective than humans have a tendency to be.

Ram Yalamanchili (40:02.83)

Yeah, I'm glad we're sort of agreeing on that because I will say my own experience building AI-based systems, one of the advantages, I think the key advantage of the systems have is they are able to share and do things which would otherwise be impossible for large groups of people to do to manage. And I think in a world where you're thinking about multiple different AI teammates working with each other, sort of like the more

these teammates come in, you actually end up getting better productivity because there's so much better sharing and planning and getting to the next step. It's almost like the exact opposite of that saying too many cooks in the kitchen. You've got this dynamic where I think if you have too many people, it actually might be counterproductive because you're splitting the context across multiple people and it's siloed.

I find it fascinating that you sort of started and identified that to be one of the key problems in this whole industry. And I think that's also to me a very fascinating place because I'm seeing the exact opposite happening when you go into the domain of intelligence, especially like model-based intelligence, right? So it's interesting perspective. I have a couple more questions. I think the first one would be if I've listened to the...

to your reference of what I call artificial intelligence or AI, I've noticed that you're deliberately saying machine learning. Is that indeed deliberate or am I just hearing it differently?

Tom Mather (41:38.749)

but

Tom Mather (41:47.358)

Well, I'm not a fan of artificial intelligence, the term artificial intelligence, because in my mind, there's nothing artificial about it. It is, if we're, I categorize it a little bit differently. There's primary intelligence and secondary intelligence, right? Primary intelligence is something that you saw and witnessed and you develop that intelligence, let me say as a human, because you primarily witnessed that.

And we all have certain amount of primary intelligence other than that. And also we have secondary intelligence. We've read about it in the book. Somebody's told us and that's how we've developed our intelligence. So sure, the machine, a computer is going to learn and they're going to learn not that differently than what us humans learn. And we're going to collate that material that you there from primary sources, secondaries. So

If you want to say that I would say that maybe machines don't have the opportunity to have primary intelligence or primary knowledge, but they certainly have the opportunity to have secondary intelligence, just like we do. And so I don't think that if you're trying to tear it apart and say that it's, that the machine, that there's somehow.

quantifiable difference in what comes up out of that than what comes out of human. There are some difference about that, but it certainly is not artificial. It's real. Hold on to that. It's real.

Ram Yalamanchili (43:30.03)

It's another fascinating take from Tom. That's really interesting. And I think the last question I have, so tell us what was your fastest marathon time again?

Tom Mather (43:44.285)

Two hours, two hours and 21 minutes. That's right.

Ram Yalamanchili (43:49.422)

And I am assuming this was not recent.

Tom Mather (43:55.621)

No, that was, gosh. I probably my early thirties and I'm 68. So I guess that's do the math.

Ram Yalamanchili (44:06.798)

So my last question, Tom, is are we going to see you run a two-hour marathon maybe in the next 10 years? Are we seeing a world where that might happen? No, there's no machine learning which can solve for that, you think?

Tom Mather (44:16.118)

No. No, they're not.

Tom Mather (44:23.012)

No machine learning that can do that. That is just not going to happen. know that's not going to happen. Yeah, it's not.

Ram Yalamanchili (44:31.47)

You know, I might take the bed on the other side. We'll see.

Tom Mather (44:36.988)

I don't know, there are certain limitations on the body. Your computer has a hardcover outside and it might last and be the same, but there are certain limitations on the inner workings of the human body as far as physical performance. And I'm on that tail end of that.

Ram Yalamanchili (45:05.326)

Well, Tom, it's been a pleasure talking to you. Thanks for sharing. I had a great time talking about the perspectives and how you're seeing the next five to 10 years. It's great to have you here. Thank you.

Tom Mather (45:18.426)

Well, yeah, I tell you, it's going to be so exciting. It gives me reason to get up out of bed every morning and to, and to continue to work along in it because it's going to change and it's going to be a lot of fun to, well, a lot of extra work, a little bit more fascinations, a little bit of more effort, doing things different. But the end product of what we're going to be doing in five years is really going to be exciting. And.

I think that the biotech changes is just, I have to think that it's going to be one, if not the most, one of the most industries impacted, sectors impacted by machine learning.

Ram Yalamanchili (46:06.03)

I also, just thinking about that statement, yes, it will be one of the most impacted, but I also think it will have the most impact, potentially the most impact on all of us. So it's got this force multiplier, which is amazing. It's one of the best times to be perhaps working in machine learning and in the biotech space, or at least looking forward to doing that and adopting more of it.

Tom Mather (46:16.284)

And which is, yeah.

Tom Mather (46:26.62)

I know you look ahead, we would have, again, not just more drugs, but better drugs, more therapeutics, but better therapeutics, more bespoke ones, more for intricate parts of that. And you think about how many disease entities that we think that we know how to treat, but there's a certain pace of those populations that just aren't treated by that, what we're currently doing.

80 % of the people won't be treated a certain way, but there's 20 % that don't respond. And if we can unlock the key to that and you look ahead, well, that's amazing. That really could be amazing. yeah, we're living in a great time. This is gonna be so much fun. Yeah.

Ram Yalamanchili (47:17.002)

I can see it. I can literally see it in your face right now. So I believe you and it's amazing. I completely agree. Well, thanks for your time and I appreciate you sharing.

Tom Mather (47:30.79)

Thanks, Ram. Always a pleasure.