The AI Agent That Compressed 8 Years of R&D Into 2 Weeks

0:00An AI agent can process several tens of thousands of papers a day and then has perfect memory of all the content. So that can be reduced from a month to all the order of minutes. Instead of human scientists, you have what's called ALAP, autonomous lab. It's basically a high throughput robot that will do 5,000 formulations in one morning. When you give that to an AI model, it will give you about a thousand parameters. We can't really interpret that. It's like a different language, not meant for us human species to understand, but it works. Welcome humans to the Neuron AI podcast. I'm your host Corey Knowles and I'm joined as always by the undefeated champion of one more thing, Grant Harvey. How are you today, Grant? I'm good. I'm good. It's a little rainy out here in Southern California, which is uncommon. So if you hear the pitter-patter of rain, that's what's going on. Then today, I win the rent lottery. I'd like to say it's beautiful here.

1:02Wow, that's rare. I know, right, right? What will be joined here in a moment by Dr. Chichao, who? Founder, Chairman, and CEO of SESAI, a company working on lithium metal batteries and a more transparent EV battery supply chain, with joint development agreements in place already with General Motors Honda Hyundai. And maybe more. We'll find out. Now, if you're wondering why this matters for AI, SESAI actually uses AI agents to discover new battery materials. Their platform, molecular universe, compresses years of material research into minutes. And they also use AI on the manufacturing side to catch defects and predict battery health. It's a great example of AI solving a hard physical world problem, not just a digital one. But first, please take a second to like and subscribe to the channel, so we can keep bringing you the most interesting people in tech and AI. And with that, Dr. Who, welcome to the Neuron. Thank you both for having me. It's great to have you here. We're really excited about it. And I guess for those who haven't thought deeply about batteries in years, because I assume the average person probably doesn't.

2:04But there's a lot going on. What problem are you trying to solve at SESAI? A lot. I mean, I think if you look at batteries, it's everywhere, but then it's a simple device, but then it's quite often the simple device that SESAI is most complicated, especially if actually trying to change it. So I would say 10 years ago, the problem that we tried to solve was a better type of battery, a new material for the battery. And then that's evolved to trying to come up with a new way to come up with new materials. Wow, that's really interesting. And you're using AI as part of that process. Like we just talked about very briefly two of the ways that you're doing that. One of them is molecular universe. And perhaps we could talk a bit more about that. And then the other one is avatar as well. Yeah, yeah. Yeah, so if you look at the battery applications, some applications you need to have higher energy density, basically make the batteries lighter.

3:06And then in some applications you need to make it cheaper, in some you have to make it last longer. And then each one takes about 10 years. So if you follow the traditional path, and then it will basically take you about a decade to solve each of these battery materials problems. And that's not a very... Yeah, why does it take so long? So there's a couple of things in the battery. And it's similar in life science, in drug discovery, when you have a new material discovery, you go through several phases, right? Basically, first you go through this idea creation phase. Like you have to have an idea. You come up with an idea for this new type of materials. And the second is idea filtering stage. You have this idea and then you have lots of candidates, candidate materials and they have the filter. This could be a tent, could be millions and then billions down to hundreds. And then third is validation.

4:07So you're down to couple hundred, but they have to test this. And then in drug discovery, you go through trials, clinical phase one and phase two and then approval. And then for battery, depending on the application, you have to do room temperature cycling, low temperature cycling, high temperature cycling. And in some applications, for example, EV, you buy a car and that battery needs to last at least a year. A lot of times the warranty is for 15 years. So that means you have to test the battery for at least a year. And there's no good way to accelerate that testing. Testing process just takes a long time. Yeah, that makes sense. And then how does a molecular universe and or avatar help with that process, if at all, or do they solve a different problem? So again, it's still the same three phases. Idea creation, candidate filtering and then validation. So for each one, idea creation in the traditional human process,

5:09that takes about on average a month to come up with a really good idea. And also the horizon of a human scientist is limited. On average, a human scientist reads three to five papers because human scientists also have to eat, have to eat coffee, and then get sick and tired. Don't be efficient. I know. And then an AI agent can process several tens of thousands of papers a day and then has perfect memory of all the content. So that can be reduced from a month to all the order of minutes. And then second, the filtering stage, that in the traditional process, you have senior scientists, principal scientists coming out with this idea. And then those candidates are sent to the junior scientist basically in the lab to make these things. And then a junior scientist, for example,

6:09can try maybe 10 to 20 different formulations a day by hand. So now as part of AI, you have this dry lab white lab. So the idea creation, think of that as a dry lab, basically compute ideas. And then this idea filtering, this is a white lab. So you go to the lab and then instead of human scientists, you have what's called a lab, autonomous lab. It's basically a high throughput robot that will do 5,000 formulations in one morning. Yeah, instead of life times for one human. Yeah, yeah, and then it's like perfect accuracy, no error. So that can reduce the filtering from, again, several weeks, another several weeks, even months to just days. And then the last one, probably the biggest band for the buck, is the validation. Validation takes a long time. It takes several years traditionally. So once you have enough data, and you can train these machine learning models,

7:11you only need to capture just the first, probably two weeks of testing. And then you will know, you will know it's end of life. So each, basically take each phase, and then you can shrink what was originally years now to weeks, weeks, if not days. So this does away with the whole idea of having to actually test a battery for eight physical years. It's able to do this in like a controlled, analyzed setting. Yes, yes, yeah. That's amazing. That's amazing. We've seen this technique being developed and deployed in life science a lot. And of course, so in R&D, when you develop a new material, you have to go through lots of trials and errors. So you go through a hundred times that don't work. And then that 101st time works. So that first 100 times,

8:11you can really use this process to accelerate. And then that final trial, the last trial that actually gets you the breakthrough. That one, of course, you can still take the time, do the full testing, do all that. But just the process before that can be much faster. I guess that explains why batteries essentially went the better part of a century with pretty minimal advancements. Is that in that? Yeah, yeah, almost no. So just put in context the materials that are used in batteries. A lot of times these are small molecules, small organic molecules. Then in the universe, there's about 10 to the 60th, six zero possible small molecules. And then since the 1990s, the last almost 40 years, the battery industry only screened about 10 to the third, different small molecules. So we explored 10 to the third out of 10 to the 60th possible.

9:14A lot of room on unexplored. Exactly, yeah. That's amazing. So did you, when you decided to move on lithium metal, was that because you already had a hypothesis that that was just better? Was that just where the industry was moving? And where did you think of coming, like bringing AI into the equation to help speed this up? When did that happen? Yeah, so we knew lithium metal was going to be the best. Because if you look at the periodic table, lithium metal is number three. Then lithium is the lightest metal we have on the periodic table. And then even lighter than that, you have helium and hydrogen. So in terms of portable energy source, you can't get better than lithium because it's already the lightest metal we have. So that's what we focus on that. Now, to make lithium metal safe in the stable for a long time for different applications, it's really difficult to come up with a material called electrolyte for the lithium metal. So you had to come up with a electrolyte material that was stable and safe on lithium metal.

10:20So then a lot of the work was coming up with a cocktail, a formulation with new small molecules for lithium metal that will make it safe and stable. Got it, got it. And then just to follow up on the AI side of that, did you use AI as part of that process? Or did the AI come after you already had the initial cocktail and you wanted to do all the testing and all of that? Almost in parallel. So we were really frustrated with how slow and how much work was taken to come up with to test different cocktails for the lithium metal electrolyte. So this tool, molecular universe, really came out of them. So we wanted to, okay, say if we have a database, if someone just had mapped all the small molecules that could be used for this battery and have almost like a dictionary there for us, and then we could just go to the dictionary, find the molecules then and then also instead of testing it, then if we just had a model

11:22and then we only give it, for example, just the first phase or so, and then we can almost predict the end of life. And then instead of having a human doing the testing in a very inefficient way, if you just had a high throughput robot, so we had all these ideas, but then no one really supplied these tools to us, so we built these tools. That's awesome. Yeah, I love that. He like, doesn't exist. Well, shoot, guess we got to make it. Yeah, yeah, yeah. From a, from a practical perspective, like to someone who's maybe, you know, not as well-versed in this, what's the practical advantage of, say, lithium metal over mainstream lithium ion that we use every day in our phones and such? So it's lighter and then smaller. What that means, for example, if you were to put that in the pickup truck, the range of a pickup truck is dependent on the amount of battery power and also the weight of the truck. So if you can make the battery lighter, then that truck can go farther

12:27or you can put one more passenger. You can put more, more payloads. And then same thing with drones. So it's, it's able to fly farther or you can add more passengers and more payloads. Wow, okay. So that's really interesting. It makes a lot of sense. I guess up to now, this has really been a balancing act of how much battery can we put in and not cross this threshold where we're losing range based on the weight of the battery. Yes, right. So if you use the, so if you keep the technology the same and then you're limited to the same power to weight ratio, it's almost like a rocket. You cannot just add more battery and then expect that to go farther because then the more battery you add, also your adding more weight. So it's not going to work. So you have to really use a different chemistry. That's when you start looking at how can I make my seats lighter? How can I build lighter dash? How can I make my motors out of that will be a little hard? I guess everywhere you can trim an ounce.

13:30This is why like the early days of EV in the 90s and 2000s, there were these tiny cars, right? Like no one, no one liked those cars because we're too small. But then, then the, the EVs after the 2010s, they started becoming more, more practical, like families and then more practical. Yeah, and it's something you'd be more, I like to say that, you know, Tesla kind of brought the cool factor to, to these like that. That was a thing that was really missing was the, the idea that, you know, a car's always kind of been an extension of yourself in a way and, and representative of who you are in, and the little bitty smart cars were, you know, not very representative of most people. In a lot of ways, I think, it was hanging up for me. Yeah. Well, that makes me wonder about the current hang up today, right? Because the EV market today is really interesting. Where if you look at it in the US, there's really not an affordable EV that most people can drive. And I feel like that's probably the biggest constraint on adoption right now

14:34is like they're just too, too expensive for that average person. I'm wondering, feel free to adjust that or correct that. But what would you say is the biggest technical constraint right now? Is it safety? Is it cycle life, manufacturability, or actually, just cost of the materials? So actually, if you look at EVs today, I mean, the US market is a bit unique. If you look at Europe, if you look at Asia, the EV markets are actually quite, quite different. For example, if you go to Norway, if you go to China, basically more than a third of all the new cars being sold are EVs. And then the cost has come down a lot, so quite affordable compared to a regular car. And then a lot of these new EVs are being made by a new wave of car companies. And they really stress the interior design. There's TV inside, there's a massage chair inside.

15:36And also you can have a hot pot inside of this. So the utility is way more enjoyable driving a EV than a regular car, because of all the new utilities. I would say the US EV market is unique in the sense the subsidy went away as on September last year. Now it's, so without the subsidy, the economics changes. And then also US market is a multi market in the sense a lot of cars from Asia is hard for those to come in without a tariff. So they also changes the market. So that's not even a battery science constraint. That's just a, it's not, yeah, isn't it? It's political economic. Everything else can start basically. Yeah. Does this supply have an effect? I know that traditionally some of these metals are very geographically located and acquiring them is difficult, comes with a lot of struggles as well.

16:40Does that play a role in pushing these things forward? Or the decision to stay with Lithium, I should say. I mean, it does to a certain extent, but now the supply chain is quite diverse. A lot of the Lithium comes from South America, Australia. They get refined in China and also some in Canada. And then they get a simple into batteries. So in terms of availability of this is not an issue anymore. Of course, sometimes the raw material price fluctuates and that influence the cost of battery. But in terms of availability, no, it's not a limitation. Okay. That's good. I was curious because I know I was thinking of like with cobalt, there were struggles as people were looking in those directions and others. And I wasn't sure about specifically how the supply of Lithium looked. So, thank you. Yeah. So cobalt is used in the foam. But then, so for example, in the foam, the cathode is called lithium cobalt oxide.

17:42It's basically all cobalt. So, but then in the EV, there are two types. There's a nickel cobalt manganese where cobalt is less than 10%. And then the other type is lithium iron phosphate. It's actually cobalt free. There's no cobalt in that type. So it's not a constraint anymore. That's good to know. And lithium was a constraint, but it seems like a lot of emphasis went towards making that, making more net new minds for lithium and trying to make it very accessible over the surface. Yeah. And also recycling. So, for example, but now we have lithium coming out of mines. So one example, you take lithium coming out of mines in Chile. And then that gets shipped to China to go process and assemble into a battery and then sold into a vehicle in the US. And then this battery in the vehicle gets recycled in the US. And then that lithium, that nickel, that manganese get used for the new battery. Yeah. So, the recycling actually allows you to not go back to the mine anymore.

18:47Yeah. That's awesome. And that was like a much needed aspect of the supply chain that I feel like it's got going down recently. Yeah. Just great. We have not far from here. We had a lithium battery recycling facility that was dealing in like old EV batteries, essentially, and stripping those and preparing the materials to go back. Is there a limit to how many times that can be reused or does it stay essentially you're looking at? I mean, there's some loss, some loss less than 10%. So each time you lose some, but it's a, to the most part, it's pretty efficient. Wow. We said the battery should last at least eight years, right? So, in theory, that's like eight years worth of potentially 78 years, yeah. Exactly. Yeah. Yeah. That's cool. Well, I want to go back to molecular universe for a second because you have this great database of all of this chemistry information. I guess my question is, what would be the next thing that you would want to do with that database? Like, are you just making lithium metal as efficient as possible?

19:48Are you coming up in new compounds? Are you exploring other material batteries? Like, what can you do with this now that you have it? Yeah. So it's almost like the Britannica, the encyclopedia, right? So we put a lot of emphasis on the data and then two kinds of data, again, dry data and the wet data. So we really want to map the entire universe of materials and all the properties, not as batteries, but then, but then like pesticides, detergents, cosmetics, oil and gas, paint, basically everything, these materials boil down to small molecules and then there is not an encyclopedia of all these materials. So our goal is to gradually build this database of all the materials and then map it. So map it meaning dry data and the wet data, dry data just use computing horsepower and then compute all the properties and the wet data is basically we have these high throughput robots that actually run these experiments 24, seven and then collect the wet data and

20:49then so we use the wet data to calibrate the computer dry data and then we end up with this modern day encyclopedia. And then this, this we can feed it into into new models that we are developing for the different applications. So this goes way beyond just batteries and and the goal is is to apply this to almost any material R&D. So the other day we had this, so one of our employees is working on a project with a home goods product, it's basically detergent and he's testing different cocktails for detergent. Yeah. And that's quite quite similar. That's cool. That's it. That's it. I was not expecting that. Yeah. That mean either. That is that is really impressive and I want to call out something that you mentioned. So you're using AI at the beginning to do your dry work, but you're also using high part robots at the tail end of that to handle the wet experiments as well, right?

21:49Yes. Because if you only do the dry computation, it's not very accurate. I give you one example. If you just use models to compute, for example, melting point and boiling point of some molecules, typically you're off by 30, 50 degrees Celsius. If you have actual data from the wet lamp, like actual raw data and then you use those to calibrate, then that error bar can shrink to maybe plus minus two or five degrees. So does that create a feedback loop then where you're like using the molecular mapping for the dry data, you're then getting wet data to validate and then you can feed that wet data back to your map and create a more like a efficient map or a more accurate map. Yeah. Dry data really allows you to map a much bigger universe. You can compute, for example, 10 to the eighth, 10 to the ninth, pretty quickly. Wet data, you're talking about 10 to the fourth, 10 to the fifth, so significantly less

22:50than the dry data, but that's enough to calibrate the dry data. Okay. So it's like basically, yeah, it's like, would it be equivalent to tuning it to tuning the dry data? Yeah. Yeah. And I guess just so I understand, is this a, when we're talking about this map, is this a bunch of text data or is there 3D models involved when we're dealing with chemistry? Is it a mix of both? Like, what does it actually look like conceptually? What did the data look like? Yeah. Like, what is the map of the data actually look like? Are you dealing with 3D simulation models like that or is this all just a bunch of like texts of chemical compound combinations? What's kind of consists of it, I guess, so I'm wondering, was it consistent? Okay. So the molecule database consists of just molecules structures, and then, and the structures are in 3D, but then you can represent the 3D in what's called smiles strings. For example, the water is H2O, and then you, you just write a, oh, so you can represent

23:51a 3D structure with, with a string of letters, C, H, all those letters, and then, and then we'll be compute, and we'll be measure are these properties, the properties are just in these numbers. For example, melting point, boiling point, energy levels, viscosity, just numbers. So at the end, you end up with an excel table of the, of 10 to the 9th, 10 to the 11th, you eventually 10 to the 60th smiles strings, and then all the, all the numbers, all the properties. That's awesome. That's a rough idea of how many of those you're running in, you know, I don't know what week, year, what's, like, I feel like there are so many applications for this, like you mentioned, that it goes so far on TVs that, yeah, yeah, yeah, not enough, not enough. So for now, the, and then, so the basic, basic dry data, we're computing is using a technique

24:54called density function theory, that one, if we use machine learning accelerated density function theory, we do about 9 million molecules a day, just the, the single molecules level, and then once you've got to the cocktail level, so that's why you mix three or five different molecules together, right now we can do about 2000 a day, but we need to do way more. I assumed it was going to be just a tiny fraction of what your dry is. Yeah. Exactly. Yeah. So how do you choose which from your drive work is going to go and actually be tested? Are they're like, right? We need to test one from this area or is it, is it random? Are you going to shotgun? Are you, are you honing in? It's a big area. Yeah. So that's where we shall have human scientists come in to train this. So think of the database as just like a dictionary, right? You still need a person to know, okay, what, what letter, what, what word do I have? I look up.

25:54So that's where the intelligence comes in. For each domain, we have about 50 human scientists to teach the model. For example, we, we saw use the frontier models, like the GP-D5 and the Gemini and then, and then those are not specifically trained in these domains, they're very general. So we would have about a team of 50 domain scientists and then teach the frontier model models, for example, about batteries, about, about pesticides, about cosmetics. In each domain, here are the things that you should look for. For example, in the battery case, to have a high temperature, stable, cycle life, you need the molecule to have this particular structure. So when you go through that entire database of molecules, look for these structures and then look for for melting point, boiling points within certain range, look for energies within certain range. So the human scientists would actually teach the frontier model these domain-specific

26:57knowledge and then this intelligence would go look for the corresponding molecules in that database. So is this, let's take the form of like a system prompt? Is this like an agent's, like instructions that you're giving us or are you fine-tuning the model? How are you actually talking to it in this way? So for now, we are using an agentic alien and then it's a combination of GPT-5 and then a Gemini and then so all the domain-specific constraints and then the information we would teach this agent and then the agent we'll look for in the database. So cool. That is. What are your thoughts on the whole thing now where it's like opening eyes, pushing this idea that the agents are coming up with their own physics, novel physics theories and all of this stuff? Would you ever, do you buy into that and would you ever have the AI be the one doing the what word to look up at some point? Are you bullish on that idea?

27:59Absolutely. Absolutely. So I'm totally bullish on that. So what the caveat is that I wouldn't trust the explanation. I wouldn't trust the result. I give you one example. So we have lots of the battery test data, charge and discharge, the voltage curves and then as a human scientist, so we're all trained in the, for example, Newtonian science, right? School, we are taught physics, chemistry, material science, mathematics, we're taught these theorems and then you study the theorems and then you apply these theorems and then the world must follow these laws, the different laws of physics. Now with AI, they go beyond that, they use laws that we are not able to comprehend. So, so for example, that voltage and the charge and discharge voltage curve, a trained human scientist would see, would characterize that curve, maybe with 20 parameters.

28:59These are typical parameters you will learn in school, in books. When you give that to an AM model, it will give you about a thousand parameters, but most of these parameters, you are not able to explain what they are. It's not like the human scientists will see 20 parameters, okay, this is charged, this is capacity, this is time, this is a DQTV, you can explain these things, the a thousand parameters from the AI, you are not able to explain those things. It's like a different language, not meant for us human species to understand, but they are, are they, are they real? Like, I like, like, how is this a hallucination? Like, how would you know that? Because, because we see, so when the AM model fits, we see a thousand different parameters, but then they are in the form of zero once, zero once. We can't really interpret that, we can't really give them physical meaning, but these, and then if you were to ask the human scientists to find patterns based on their 20 parameters, the patterns are, are weaker and not as strong as when you ask the AM model to find patterns

30:01based on a thousand plus parameters, right? So, and then when you ask the AM model to predict end of life, just with a beginning performance is, is much more accurate. So even though we're not able to assign your physical meanings, it works. It's like a different set of laws that we're not able to comprehend, but it works. That is so cool. It is, it is, and it's such a, just an interesting field to see this happening in, and also a really interesting application of AI, like so often, Grant and I have these discussions where we're dealing with how to make better models, how to make models understand better and talking about reasoning and inference. And what attracted us to this conversation so much was the idea that this is AI being used in real scientific fields, you know, today, how, I'm wondering how long have you been taking this approach, if you don't mind me asking about, about three years since on the material side.

31:02And then, and I think going forward, now that the approach really works, we really need to expand this. So a lot of the high throughput robots and then the computing, we do need to expand those. So we can actually map it much faster. Okay. Yeah, make sense. You're going to need more robots. Absolutely. And actually speaking of robots, unless you have another trade. Well, I just, I had one little follow up. Go for it. So do you notice a significant difference as the models have improved since you've been doing this over the course of three years? You've obviously seen some pretty monumental leaps in technology over that period. Yeah. It's basic for the, the more data you give it, the smarter it gets. And I will say the biggest difference is once you've reached a sufficient amount of data that you teach it, then the model is able to give you results and forecasts that's, that's almost spun off.

32:02Wow. So then, then you can really save a lot of effort, but you really have to teach a sufficient amount of data. My concern with that approach, though, is that like the current language models, right, they have a limit to their context, right? So are you, you must be using something else or you can tell me what you think about this. Like if you have this giant database of all this different molecular data, how do you make sure that it's considering absolutely everything when it's going to work here to get what I'm saying? It's considering everything in terms of like basically, how do you prevent loss from happening with the context window when you're running an agent through this data? I guess what I'm wondering. So when we have the raw data, we don't really teach that to a large language model. We use a foundation model because the large language models are really good when the data is in a text format, but when it seemed like Excel numbers, it's, it's not as good.

33:05So we use that to teach, so those two parallel approach on the database, the raw data from the lab, dry data and what and what data we use that to teach a foundation model, no large language model. And the parallel to build that intelligence, we take all the books, all the papers about this domain, and then we teach that large language model to learn how to search for it. So one is, so think of the database as the map, and that's not LAM, and the think of of the LAM as the search engine that is LAM. So we don't feed that large database into the LAM. We feed that into the map into the foundation model. And then to the LAM, we only feed a more limited list of properties. And that makes sense because when you're doing a, you know, a more specific run, have a more constrained problem space, so you're like, okay, we know we need to focus in this area, because we're looking for this chemical property. That makes sense.

34:05Yeah, I was, I guess that was the thing that was starting off is like, I know LAMs have a context limit of a million, you're like, I know there's an answer, and I just don't know it yet. Yeah. So you can't necessarily put all the chemical data in the world in LLM and expect to get that. Yeah. Yeah. That's cool. Well, I, I want to talk a little bit more about, about lithium metal, because I think it's really interesting that that you all have these three major JDAs in place already with GM, Honda, Hyundai. What does that, what does that look like in practice? Is that actively providing batteries working together toward that? Yeah. So it's, it's really to improve lithium metal and then develop the battery so that it's ready to be deployed in vehicles. And of course, that technology development, that product development can also be used for drones or energy storage for data centers for lots of other applications. One thing we have seen in the electrification effort, the EV industry has been sort of

35:06the, the pioneer of the technology and product that develop in EV are now used in other industries as part of the electrification. That makes sense. So what are the like milestones that OEMs are looking for in something like this? Are they, I assume there are goals you're after? Cost duration. Yeah. So there are technical specs. You have to meet range, high temperature, low temperature, performance, safety, a lot of safety. Okay. Safety test is, is no joke. And then also the scale, you do it, for example, 1000 cells and then, and then a million and then 10 million and then also at those scales, they alter your supply chain, your quality, all the quality process, I love the details in the manufacturing. Okay. Okay. What about avatar? Because what I thought was really interesting about avatar, which is the other AI tool that you use is it's actually tracking the battery life cycle and you mentioned safety.

36:11So I'm curious if you could talk a little bit about that and why that's a big deal because batteries are living chemistry and being able to track them is really important and other probably. For example, EV and then you really want to track the, the safety, all the batteries in, in the same fleet of vehicles have the same chemistry. But once they start entering the manufacturing line, they will have different defects. Maybe this one has some defects in, in a step, step 70, the other one has some defects in step 400. Typically, you have about 3000 or so steps in the manufacturing. So you'll have different manufacturing defects and the ones they are packed together inside a car and then the driver behavior is going to be different. So the final battery inside the car, the health, the safety, actually is a function of the

37:14manufacturing defects and also the driver's unique behaviors. All this, you really want to track and monitor so that you can do maintenance and then you really want to, for example, a regular car, you do oil change once every four, six months. And then with EV, if you can track that, then you want to be able to predict the incidents before it happens. So that's the top goal. But really is to prevent an incident and also predict an incident before it happens. And then once, and then if you apply that in energy storage, you can actually use that for electricity trading. So what that means is, when you do trading is basically a supply and demand, demand, there are these virtual power plants that has to do with weather, you forecast the weather, you forecast any storm, any major sporting events, any, if it's a data center, any incoming influencing, influencing jobs, that's on the demand side.

38:15And then on the supply side, you have a choice, do I bid or do I not bid? And then if I participate in the bidding, okay, I make some money now, but then I will probably hurt my battery down the road. So I reduce my battery from eight months to, from eight years to seven point five years. So I lose five, five months, half a year of a revenue down the road, do I, do I make this bid? So, so having a very accurate battery health monitoring allows you to optimize the supply side of this trading. And we're seeing this in, in both data centers and EV. So once we save you, yeah, why don't you save the data? Yeah, it's actually use a lot for energy trading. Yeah, yeah, well, especially with the mentioned data centers, I imagine they need to be very, very efficient with their power, right? So this is very helpful. Yeah, yeah, a lot of times the data centers cannot predict what's, what's coming down

39:15the, the pipeline. And the data centers are actually quite different from a normal grid because you really have to allow for search in, in power. So if you have a, a huge job coming in, then you, then you have to drain the, the entire battery in about two minutes. And then that kind of super high power density battery, we have not seen it's actually quite, quite new and, and it's got to be safe enough. And then also it has to have really high power density in the data centers. So the, on the, the supply size that you're quite challenging. So obviously you're developing your own robots. I'm curious what your thoughts are in terms of whether you are potentially working on something like this or whether you just have general thoughts on the direction, how to actually give robots enough power so that they can be as efficient as possible. It seems like it's a battery problem to me, but I'm curious if that's something

40:17you're actively working on and thinking about. Well, so, so the robots that we are building are more stationary and then they are plugged in. So it's not more industrial approach. Yeah, yeah, exactly. It's, it's basically like a, like a machine with a robotic arm. We don't really build like a battery of power towards humanoid. We don't build that. But I think for batteries, I mean, we've, actually we have some, some humanoid customers where we supply the battery and then we're getting. So before it was about two to four hour runtime per battery, we're able to extend that to about eight hours once you get to eight hours, then it's almost like a human worker, like eight hours shift, right? Like one shift, second shift. So then, then, okay, then, then give the robot a break, right? Yeah. So we'll go to the corner, charge, and then another robot comes in and then, and then do the work. I mean, I think eight hour is doable from a battery perspective.

41:21I've seen some companies where they designed the robot. So it can actually swap the battery by itself. Yeah, that's one concept. When it recognizes a certain amount of, like, oh, I'm down it 12%. It's time to plug in. Yeah, yeah, or just swap it with a fresh battery. Yeah, yeah, smart. So we're seeing some of the, yeah. Yeah, no, that's funny. It's like kind of like how, you know, we, we're like, man, I'm getting hungry. I should probably take my lunch break soon because my work's going to suffer if I do. Yeah. That's cool. Is there any benefits of trying to get it to like 20 hours or 10 hours? Is that diminishing returns, do you think? I mean, yes, because the robots are expensive. And then you definitely want, you definitely want the robots to be functioning as much as possible. And then we're seeing, we're seeing swapping. So if the swapping can be done efficiently, then, then the batteries, the batteries don't have to last 24 hours. The batteries need to last probably 8, 10 hours and then to swap it.

42:24And then the robots will go back to work. Some kind of a charging station where it goes and hooks one and, and socks it on, takes the other one then and, and puts it back on a charger and you could theoretically just cycle. Yes, yes. And also, we've seen a robot customer where the humanoid's actually work on the line and then another humanoid, not humanoid, another battery pot will actually come to the humanoid and then swap the battery. So the humanoid actually never has to leave the line. Wow. Wow. So just recharge it. It'll be like, if you're working the line, so it comes in and feeds you. Yeah. Yeah. Some like that. Yeah. I'm kind of a cord. You can just plug in when you know you're not moving. It could just jack into the line with what about as these things get more intelligent, right? Like let's say like Nvidia comes up with a new GPU that could potentially power the robot and give it twice as much intelligence or four times as much intelligence as they will. But then maybe that requires more power. So then that's a trade-off, right? It's like you're trading intelligence and battery life potentially.

43:27Yeah. Yeah. If if the GPU runs more than that does consume battery probably. And I don't get it. But it has to be an on board system though necessarily. You know, if you're running strong enough, I don't know. That's a good question. Would what do you think your GPU rig be on board or would it be in the cloud or a server room in your house? I think both you will have cloud and also edge. I think both that makes sense. Yeah, maybe you have like a local one that's maybe more like for real time. And then maybe your thinking power is done on the cloud. Perhaps what about the GPU costs? I don't want to have to buy GB 300. What about what about the use case of glasses? Because I know air glasses have been power constrained for for the last couple of years. And that's one of the reasons we haven't seen like consumer grade AR glasses really take off. Meta is obviously making good progress there. But we talked a lot about EV size batteries or drones or robots. But what about making it as small as possible?

44:28And it's kind of efficient as possible. Yeah, yeah, it's it's possible. And a lot of a lot of these these new high energy batteries are actually very dense. And then you can pack them in a small place. But the the glasses actually consume a lot of power, especially when you have the camera on. It's especially very pale hungry. Yeah, I wear my daily glasses and I can tell you if you're running the camera, you're going to run out of time about everything else. It does really smooth and they'll you get good life out of. But if you're running a camera, it's going to drain quick. Is that something that is worth pursuing? Do you think like in terms of some of using molecular universe and try and yeah, try and solve for that? We actually have we have some some users that try to use that platform to solve a problem. Wow, that's awesome. That's so cool. Keep us posted. If anything, exciting happens there. Well, Dr. Hood, thank you so much for joining us. It's been just an absolutely enlightening conversation and it's fun to see how real companies,

45:32real scientists are using AI out in the real world and go ahead. I'm sorry. Oh, no, no, it's it's been it's been quite quite fun to to share this with you guys. And and actually AI, especially AF of science has been used in material science in life science, new drugs coming out, new paint, new batteries, lots of new things coming out. We'll be discovered by AI in addition to their human partners. Is there anything on the horizon that you're working on that you'd like to plug in that, in that vein, anything exciting that you want to touch on before we go? I mean, love the batteries and battery backup for data center. We're working on it. It's actually quite interesting. So we have one universe that's powered by a data center and then one universe basically maps the universe and it comes up with these molecules and then we use that to to put them back into batteries and then we use the batteries to power these data centers. So it's almost like a loop.

46:32Yeah, that's awesome. That's cool. Well, I'm sure you're going to be busy for years with all of the data center projects you're working on. A lot of work to do there. Yeah. Dr. Huwap's best way for someone to keep up with what you all are doing and go learn more. Yeah. So we have, they can follow molecular universe. It's molecular dash universe.com. Yeah. And then also once a while, we send out these awesome, awesome, awesome. Very cool. All right. Well, thank you so much to everyone who watched today. Please take just a minute out to like, subscribe. We really appreciate it helps us continue to bring you guests that are doing amazing things in the technology and AI space. On that note, that's all for us this week. Farewell for now, humans.