995: End-to-End Foundation Models for the Energy Industry, with Jazmia Henry

25:39Jon Krohn:been pretty cool well i wish i could do another one now because i really i didn't know i'm not i'm not going to i'm just saying i theoretically wish i could i mean i i guess never say never i have no plans to do a second one. But my point is that when, when you've been in industry already, and then you go to do a PhD, you have so much appreciation for how these things will be used in the real world. Whereas like, I had just been living in a bubble of textbooks and academic papers until I went into industry and you, you don't appreciate how special that time is. And And I think if I had been in industry first, I probably, you know, not to say that I didn't make good use of the PhD, like published a dozen academic papers.

26:25Jon Krohn:And, you know, so from that perspective of like academic success, it was successful. But if I had industry experience, I think I genuinely 10x in terms of the impact and the synergies that I could have noticed, you know, using the time well on things that will matter in industry later on, maybe even, you know, making impactful strides in the real world while doing the PhD. Yeah, as opposed to kind of just being in the bubble. And yeah, so I think that you, you know, you had a great path there. And, yeah, I also now see that I, if I clicked on the ellipses on the dot, dot, dot on your bio, it does say left to focus on industry.

27:10Yeah. Oh, it's totally fine. I mean, I, I will say though, to, to your point, I think that it is a interesting time right now in AI. I, if I had went straight to my PhD after Columbia, like if I went straight in, I think I would have just finished it. because it's like you know also oxford has a the european phd and the us phd is his own other i'm sure podcast where we could have a discussion on the differences of it but you know three years versus six years is a very different oh totally when it comes to but also it requires you to have a different what's the word understanding of what you want to do immediately like you start Oxford and people already know this is my advisor.

28:02This is my focus. Where's the US to get more time because of the timeline to do it? And so I can definitely see on one end, if I would have done it immediately, I would have just checked it out. And then, you know, because AI wasn't moving as fast at the time, we were, you know, like most in actual English processing, even though the the attention is all you need paper came out in 2017. Most of the industry was still using natural language, like the NLTK toolkit in Python, pretty much up until XactoT came out. And so you could still kind of get away with, you know, doing that and taking more time.

28:41I think it's because I started the PhD in 2023. And then all of these things are taking off at the same time. I might have applied in 22 before, right before Chaggy BC came out, started it. And then like, that was like, you know, on the run and I was at Microsoft. So I was right in the middle of it. I think that that is a lot of why I was like, okay, if I want to do research, that's cutting edge right now, it's an industry, but I'm sure the tide will change and we'll be right back to academia and I'll go ahead and finish the PhD and make my professor father very happy.

29:24Jon Krohn:Yeah, he'll appreciate it if you come back for sure. But regardless, am I correct in understanding based on your LinkedIn profile that basically throughout the PhD, you were also working full time at Microsoft and then later ISO? That is wild to me. Especially with that kind of European three-year program. I mean, yeah, you've got to be really productive. That's wild, Jasmine. Yeah, you got to be productive. You have to have, and this is, you know, for anybody who might be listening to this and thinking about taking that route, you have to be open and honest with everybody. So I was on a Microsoft research team when I first started at Project Bonsai over at Microsoft.

30:12And then I moved over to Microsoft AI. With my original manager, I had to have a conversation with him about what this PhD was going to be, what types of things I would have to publish. I had to, even during my application process, I applied with them knowing that I would be working full time and got advisors who had other students who were doing the same thing. You know, one of my advisors, Chris, Professor Chris Russell, one of his other people he was advising was at Google DeepMind. So, you know, having that understanding that like we're going to work on this together. But then, you know, throughout the time, having open and honest conversations, both with he and my other advisor, Brent Middlestat, and being like, hey, here's what's going on in industry.

31:00Here's what I'm seeing. Here's what I'm able to do. And here's what I'm not able to do. And even with me deciding to take a step back and go back to industry, that was a group decision. That was me coming to them with a paper I had worked on and a patent that I had worked on and them being like, oh, great. Where is this patent taking you? And I was like, to a company. And they were like, oh, if you're going to do a startup, you should probably, you know, consider because, you know, Professor Chris Russell, he's, you know, had his fair share doing that type of work. So I think that you have to have the right type of team around you.

31:37If you don't, it doesn't matter how much you want to do it. It's going to be very hard to, you know, get something like that going.

31:45Jon Krohn:For sure. Well, congrats on, you know, getting a couple of years into it. And yeah, hopefully, I don't know. I mean, not hopefully, like whatever happens, happens. You're going to make the right decision. But yeah, one way or another, it's going to be interesting. And regardless of whether you formally finished the PhD or not, you're doing frontier AI research. Let's dig into some of that. So at Collide, that oil and gas AI company that you're at right now, you are building RIGS, not oil RIGS, but R-I-G-G, lowercase s. So RIGS, it's a model name, I guess. You can tell us if RIGS stands for something.

32:21Jon Krohn:But my understanding from our research is that it is a set of petroleum engineering foundation models for the oil and gas industries. And so I guess that ties into a lot of the stuff that you've been talking about already in this episode. But as part of RIGS, you're tackling one of the most intractable challenges with large language models, and that's evaluation. So a few weeks ago, you dropped a new paper that identifies four systematic failures that make current evaluation frameworks structurally inadequate for assessing deployed agentic systems. And as a consequence of those failures, reward hacking is a predictable consequence.

33:04Jon Krohn:So tell us about these systematic failures, what reward hacking is, why it's a problem, and then what your solution is. Yeah. So I kind of was touching on how we use reinforcement learning with our models. And so there is this very interesting thing that has driven me absolutely insane for a very long time in AI, which is regardless of what type of sub genre of AI that you're doing, which is that it takes an insane amount of time for you to figure out if your experiment actually worked. Like you spend time, OK, I'm going to tune these parameters. I'm going to take this pathway. I'm going to do all this.

33:48Great. It's wonderful. You begin moving forward with it. Like, let's say for I'm going to talk about specifically large language models here. But let's say you have your data, you decided your chunk sizes for retrieval. You've done all these things that's like super fancy and amazing. You're like, OK, we're going to do QLaura because that's the that's the greatest thing to do. Great. You begin running training. You even do some checkpointing and stuff like that until your checkpoint is finished. you don't actually know how well your model is performing. So you have to wait for a checkpoint for eval.

34:22That could be anywhere from two hours if you're doing something very quick, fast and dirty with a really small model. Or if you're using a very large model, it can be days before you're able to actually sit here and go, okay, the model is moving in the right direction. Huge waste of time. And again, this isn't just large language models. This was even back when I was doing Monte Carlo simulations. It takes way too long. So my idea was, okay, what happens if we're already kind of spinning up these environments? What happens if we use that Monte Carlo simulation idea? Okay, I'm going to have a model that's training within these certain specs, within this data.

35:04And based off of how it's moved, like the direction of trajectories that is moving in, like within this reward model, because I can quickly pull out the calculations for the reward as it's going before it moves to a checkpoint. I'm going to do Monte Carlo simulations on those and see how well with these types of conditions it's going to be by the time it gets to the checkpoint. So that means I want to get my first burst of my first pass of reward that's coming out of that reinforcement learning model. it's probably about an hour or some odd of time, I can very quickly go, okay, based off of these trajectories, it's going to go in this direction versus the other direction.

35:46By the time I've had that do a few times, I'll be able to say like, oh, it's moving in the wrong direction. We need to go ahead and chunk this experiment. And then I can run ablation tests as well. So let's say like I have an environment that's, you know, doing a calculation, so on a return, I can go ahead and say, what happens if I have multiple different small variants of this experiment and I change just one small thing? So I'm going to weight the data more heavily towards equations for this set, or I'll weight the data more heavily towards oil and gas. And I can do the same thing of like, I can run multiple experiments at the same time, do simulations on all of them, see which one is going to be the best, and then move forward with the winner.

36:31And so that basic idea was like, okay, this is great because it means that we don't have to wait for days. I think the worst part, especially when you had a startup. So like back when I was at Morgan Stanley, back when I was at Microsoft, you just ate that time. You had a company that was big enough to where you can eat that time. You can eat that compute cost, whatever, who cares? Microsoft was charging itself to pay itself for the compute. It didn't matter. But like when you're at a startup, you have to be scrappy. you have to identify what can actually work. And also, you know, I might, we might have somebody who sells to our customer on Monday, hey, Riggs, you know, performed well, that X, Y, and Z thing by Friday, they want to go ahead and demo it.

37:18I don't have time to, you know, wait, because if the demo doesn't work out, we might not hit our numbers, we might not hit our funding round. So that was where that came from. Speaking about the failures, though. So when you have reinforcement learning, what ends up happening is models are very smart in that they are able to identify the least that it has to be able to do in order to get a good reward. Because over time, it's seeing the same things over and over again. It starts to go like, okay, fine, I'll just do X, Y, and Z. And that's another reason why these types of things, these way of evaluating can be so powerful because in that first hour, it hasn't seen the same thing over and over again.

38:04And so when I do a simulation on how well it's going to do by the checkpoint, if it doesn't look like it's going to be good, but then it comes out on a checkpoint and it did amazing, I already know what reward had because it went ahead and identified of like, oh, she's asking me these questions that talk about math. And I don't have to know the definition of something if I just guess at the equation or, oh, we're doing retrieval and I'm going to be judged based off of my retrieval at K and she's going to do three and five and judge me based off of how much I'd be getting three and five. So I'm just going to guess the five most common equations that are used in this set.

38:44One of them will be right. And then I can move to the next bit. And then suddenly if I give it away more equations, it falls apart. So that's another issue. And then another one of the failures that comes with the word hacking is the data that you're using to train is different than the way people are actually going to use the model. So I'm going to ask questions at eval, like there's something called the petroleum engineering exam. And at eval, I might have the model test itself on how well it can do with the petroleum engineering exam. Okay, given that you have this well and you have this, these amount of mouths and you have this tool, what's the hydrostatic pressure?

39:25Okay, great. People are not going to ask a perfectly formulated, created by academics question with multiple choice question and answer about oil and gas. They're going to come in and go, hey, tell me about this well. And it needs to be able to tell them about that well. And then they're going to go, what's the decline curve? Might even say, what's the decline curve? Some people just say decline curve. Some people just say decline. So being able to make sure that the model as well as able to converge to that, that's again where those simulations can come in at. because now suddenly I can begin injecting questions into the environment and see how well it does on those questions at the moment.

40:13So the model might be doing very well to be able to answer those SP &E questions. And I'm like, okay, great. This is exciting. But it might not do well at being able to understand the underlying text and definitions and how things are related.

40:32So so that that way I can make the adjustments as it goes on.

40:36Jon Krohn:Nice, yeah, so to kind of sum up this reward hacking issue, it's the situation with reinforcement learning models in particular where they are, the way that you program the reinforcement learning model to learn is by having some kind of reward function where a simple example that people can easily visualize is if you're training it to be really good at Tetris, then you can use the point score in Tetris as the reward because the higher the point score in the game, the better they're performing. But that's easy in Tetris, which is just a video game in the real world when the reward is more complex.

41:12Jon Krohn:You gave lots of examples there, but like, you know, for a self-driving car, how do you, if you were using reinforcement learning to, so is it, you know, every meter that you travel without an accident, you know, you get some reward. And then if you hit a pedestrian, and you get like negative reward. And so, you know, coming up with what the reward function is, is really complicated in the real world. And you can end up with, you might think that you've crafted a really good reward function, but the reward hacking happens when, as you described, the reinforcement learning model figures out some way of getting to like that Tetris high score or that self-driving car high score, getting a really high score in a way that you didn't imagine it would and that is different than what you how you wanted it to behave yes and that happens a lot with a reinforcement learning algorithm called grpo um which is just like a group policy i don't want my model to just know one thing which is like maybe math right i want to know uh be able to choose the right equation i want it to be able to exist within the physical constraints i wanted to answer or a question about oil and gas well, wanting to retrieve well.

42:27But in the end, I want the final answer that it has to be correct. And maybe the model might learn that, oh, I actually don't have to, like, I don't get as big of a reward if I know the math equation correct, for example. So I'll just not learn the math equation correct. I'll just make sure that I get the final answer correct by looking at some keyword that's in a sentence and know that every time there's keywords in a sentence, the final answer is likely going to be one of these four things. And so when you have that issue, you want to make sure that your model isn't over optimizing just to give a CBO award.

43:09And then by the time it gets in someone's hands, it's completely useless because it's just trying to get the final right answer. And we kind of see this a little bit with different models on different varying levels that use reinforcement learning. Like, you know, people have issues with a lot of models where they're asking a question and it will just like, you know, hey, you're right. This is actually the correct answer. And you're like, no, actually, that's not the right answer. It's like, you're right. My bad. And you're like, okay, like, I need you to be grounded. And essentially, that's what, you know, that paper is for it's try to have a level of groundedness that's deterministic within it.

43:53And also, too, another big portion is, you know, handling how much it takes to train these models with reinforcement learning. Many times you have to have a reference model, which is a whole other conversation that we can have. And so by having these deterministic environments, you don't have to have a reference model. You can simply take the weights of the model and how it was performing at a given time, use that as a simulation and then be able to compare how well your model is doing, which means lower compute, don't have that big old issue with burst, as big of an issue with burstiness and, you know, able to actually create a model that works well on a MacBook as opposed to having to always go for the much larger GPU cluster.

44:38Jon Krohn:Yeah, so great explanation there, Jasmia, thank you. You mentioned grounding in there a few times, but I don't think you mentioned specifically that in the paper, your framework for mitigating this reward hacking is called grounded continuous evaluation, GCE, right? Yeah. Yes. Yeah. Yeah. It sounds pretty cool. So yeah, using simulation-based fine tuning to make everything happen, get around these reward hacking issues and get all those kinds of intermediate steps that you were describing as well, not just the final answer. Yes, exactly. And, you know, injecting some determinism. It's so interesting that I think everyone is just kind of assumed that the non-deterministic nature of large language models is just a foregone conclusion.

45:23It's like, oh, non-deterministic, who knows why it decided that. And I think that it's cool that there is a level of non-determinism. I mean, like, even in this conversation we as humans um we are existing both on a non-deterministic and a deterministic frame like the non-deterministic frame is that i might not know exactly like i might not know exactly what words you might say next i don't even know what words i'm gonna say next yeah but i do know that in the context of this conversation that there's going to be certain guardrails there's going to be certain things that you aren't going to do like you're not going to jump up and down and start doing much of jumping jacks.

46:03Jon Krohn:Because that then would be outside of context. But, you know, like identifying that there are certain contextual cues that kind of inform us on different rules of behaviors that we kind of exist within. And we've all decided to trust each other with these things. Like, and anyone who violates that, we're like, whoa, that's a violation of a norm. And so, and here's me pulling in my undergrad philosophy major here, but, you know, we need to begin to inject those levels of norms and determinism in our models. And by having that GCE, you know, I try my best to explain things without using so much jargon, because not everybody comes to know the same backgrounds, but that's really where that idea comes from.

46:56I want the model to have a sense of this is a space where you have to exist and anything that deviates from that is not going to receive a reward.

47:06Jon Krohn:Really great analogy there, bringing into human deterministic or non-deterministic norms as well. I love that. Let's move on to another paper that, or actually I'm not 100 % sure it was a paper. Definitely it was a NeurIPS presentation. You can fill me in here. So it was a 2023 NeurIPS. You demonstrated that scaling bad data makes models worse, not better. Tell us about that. Okay, great. Yeah, that was a paper that I worked on. So this is where I might sound a little bit cheeky and arrogant here. When I made that paper, everyone thought that I was absolutely ridiculous. So there's this calculation that came out of OpenAI.

47:49It's a paper written by Kaplan et al. Essentially, it's this idea that is the underlying rule that people still kind of exist by, but less so, which is that if you have data, compute, and parameters, you can linearly extend the capabilities of a model as much as you just keep on adding on to it. There's some people who push back and say, well, data is most important. Those people quickly lost. I think there's some people who said, well, no, you know, parameter is more important. Those people also have been to loss over time. Compute has still remained that important thing where we're seeing it now.

48:30We have a compute shortage happening right now where people go like, OK, just throw more compute at it. Like if you just add more data and more compute, then the model is going to be even more amazing. And back in 2023, I was kind of looking at it from a perspective that comes from economics. So there's like a law of diminishing returns in economics. At a certain point, like money is great. For example, we live in a country that is, you know, very much fueled by money. But if our government is going through something like a recession or something like that, and they print like a bajillion dollars, then suddenly the value of money goes down.

49:07There's a level of printing that has to happen, but it comes at a cap, which suddenly the laws of supply and demand end up becoming an issue. And so that was the idea of the paper. The idea was data is great. Bad data is not great for a couple of reasons. One of them, if you have model data that hasn't been deduplicated, which a lot of these models in order to have just that much shared data, not only are they not deduplicated, but it's nearly impossible to deduplicate them because they're huge, you know, huge amounts of data. It begins to make the model get more forgetful over time, less intelligent over time, begin to overfit over time.

49:51It's so interesting because in the large-ingles models, we stopped talking about overfitting, even though that was one of the basic issues of machine learning. And then the same thing from just – I spoke a little bit about compute, but focus more on data. But also from a compute perspective, right? You can't compute your way out of bad data. So it doesn't matter how many GPU clusters I can have. I can have a machine with, you know, 10 ,000, you know, GPU clusters and all of it has really bad data and the model is going to perform worse versus if I have less compute and I have less data, but it's better data.

50:30It's more quality data. So, yeah, those were, you know, really great paper at the time. Again, a lot of people were skeptical. I got accepted into a workshop. I, you know, top 5 % paper, which is pretty exciting because I just assumed everyone was going to boo. You know, it ended up kind of, we're kind of starting to see that happening and at play now where you consistently see a 70 billion or 120 billion parameter model be able to outperform in a much larger, you know, 1 trillion plus parameter models.

51:07Jon Krohn:Sure. Yeah. And I did find that paper while you were speaking. I'm embarrassed that I wasn't sure whether, well, because it did also have a poster presentation, right? Yes. And so I think that's where I got confused. But so the paper title is Scaling Laws or the Laws of Diminishing Returns, How Scaling Law Quality Data Degrades Model Performance. And so I'll have that for people in the show notes, as well as your earlier, well, sorry, the paper we were discussing earlier, that came out only a couple of weeks ago called Beyond Static Snapshots. So I'll have the archive link to that one available as well.

51:45Jon Krohn:Jasmia, this has been a fascinating technical episode, but I do also want to cover a couple of other topics quickly before we wrap up. So you are a champion for linguistic equity. Yes. So dedicated to ensuring that the future of AI includes the voices of the marginalized through your pioneering work in creating the African-American Vernacular English data set. So AAVE. Tell us about that data set and why it's so important. Yeah. So we talked a little bit about, you know, my, my, my major. So my background when I was in college was a philosophy, English, and poli sci, and a focus in poli sci was on economics.

52:25I'm an economics nerd, but the idea came from kind of the way I was raised. My, My dad is an Afro-Latino Caribbean man, and my mom is a West African Black American woman. My mom, you know, Low County, South Carolina, Gullah accent. She just says all types of stuff that I'm like, that's not, I don't think that's fully English.

52:54And my dad, you know, spoke many different languages when he was growing up and has an accent. And I grew up in Atlantic, Georgia, and they have their own vernacular way of saying things, way of doing things. And one of the things that I kept on picking up was, and it started off with me doing my thesis in undergrad on this, which is that African-American vernacular English is like this tie that binds between the African languages and between English. and if languages can have a tie that binds to each other to where I don't necessarily have to understand exactly, like I don't have to understand tweet to kind of get a sense of what somebody is saying because the way that they might say something.

53:44And so I was like, this is very interesting. This is something that can help unlock a lot of richness for language models, for how we communicate with each other, for how we spread information. And so, yeah, I went and was working on that at Stanford because I was like, I think that there can be a way for us to create a corpus that is able to tie these two things together. I ended up not moving forward with it, though, because I began realizing that linguistic equity, yes, is about representation, which is very important. But there's also a beauty, and actually I was told this by one of my peers that works in a lot of African language work that actually worked on a model called AfroBert, which has recently been taken down.

54:37But there's a beauty in people being able to own and have ownership of their language. So yes, we want to have languages and systems and linguistic equity from a sense of language models being able to communicate across different people. But also having an understanding that there are certain individuals who feel unsafe by the powers that be. And so giving those powers access to be able to mimic their language, which gives them a sense of feeling safe. But then being able to use it for things that might be nefarious isn't the best thing to do. And so giving people ownership of their language, acknowledgement, but also ownership.

55:18and being able to say, I am a guest in an AAVE home, even though I identify, obviously I'm African-American, but identifying that with my parents being from different places, there is a certain level of guestness that I have to acknowledge and allow people to have ownership over their own thing. If they decide they want to come out with an AAVE language bot, somebody who's 80, somebody who comes from being an American descendant of a slave, then that's a tooling that they can use. And I've had some people reach out to me to do that. But I can provide the bridge. That doesn't mean that I need to be the person who crosses it, if that makes sense.

56:00Jon Krohn:I see. Yeah, that does make sense. That's a, yeah, it's a nuance that I didn't appreciate as I was asking the question. And yeah, I'm now enlightened on it. I didn't appreciate it as I did the research. I was like, So in that most recent response, you did mention, you know, your family background. And so I just want to have, I've got one last question for you, uh, before, well, technically I have three questions left for you, but I asked my final two questions. I ask every guest those same two, uh, and they're short, quick ones. So this is the last one that's like for you specifically. And it kind of, it goes back even further into your background.

56:36So, you know, we've talked about, you know, you did your undergrad at Tulane.

56:41Jon Krohn:We didn't specifically mention Tulane, but great university. You did your undergrad at Masters of Columbia. You did the technical research at Stanford HAI Lab, which is a super famous lab, you know, that's world leading. The Oxford PhD work. You've worked at Microsoft. You've been a data strategist at Morgan Stanley, a head of machine learning at the Motley Fool. And now you're at Collide at the front here, building end-to-end models for the oil and gas industry and publishing in NeurIPS. I mean, you're doing such amazing things all over the place. And so, I don't know. I don't know if you have kind of insights into how you became like this, that you're just able to achieve so much.

57:32Jon Krohn:You know, like what drives you? And, you know, do you have any tips for our listeners who would like to be, you know, competing at the frontier in so many ways like you have over your career? Sure. I laugh because I'm like, oh, I don't know if it's always a great thing to be like this. I think I've just always been a kid. I already talked about my parents a bit. I, you know, Atlanta is a what they call a chocolate city. It's very black city where people come from all over the world there. But, you know, my specific type of mixture is a more rare combination. And so my sisters and I, we just kind of existed within our own kind of pod.

58:22And then I have cousins who like all of my uncle is married interracially. And so all of my cousins, we all have like different racial identities. And so I kind of grew up knowing what it kind of felt like to like look like everybody, but not feel like everybody and kind of be excluded from some stuff and accepted in other stuff and try to like navigate that world. But also navigate that world with the understanding that we have a country that has multiple layers of privilege. And in some places I had an insane amount of privilege that were different than other people who looked like me and in other places that had less privilege than people who looked like me.

59:00And so I've always wanted to answer why that is. What does that mean? And how can I use that experience for positivity as opposed to walking around and being one of those people who's just like trying to create these strange, like, you know, intra-racial, I guess that would be called, race wars, like, well, Caribbean versus like all that silliness. or, you know, without being somebody who felt like I had to try to over-index by being somebody that I wasn't. Just saying like, okay, well, I'll just, you know, completely abdicate and I'll just only be around kids that were like, you know, white kids or whatever.

59:43Like having a beauty of the fact that I could exist in all spaces. And what would that look like? Like, so that's why I did my, you know, undergraduate the way that I did. But that's also why I took the jobs that I ended up taking, because all of them in some way were kind of helping me answer questions that I had for myself. When I first left college, I was working for the Clinton campaign and it was like, obviously I'm working in politics. So that's answering some questions that I have about what I can do to be able to help people. But then I went to graduate school and graduated and the thing that paid student loans was going to be in finance.

1:00:28But then learned very quickly that finances is a lot of the reasons why people have privilege versus not in this country and learning that a lot of the questions that a lot of things that are separating people might appear racial in some aspects because they are, but also are class interplays. and Atlanta definitely gave me a masterclass in that. And so me just trying to learn how to navigate what that means to be a petite bourgeois versus being somebody who might be, you know, a proletariat or in some other space. And then, you know, Motley Fool, same thing, was the distinction of that, right?

1:01:11This idea of a company that's like, we want to give the average person the same access that everybody gets when they have a financial advisor. and me being like, great, I'll do that work. And then we go into Microsoft and working at first for a reinforcement learning team that was actually in industrial manufacturing. We were building reinforcement learning agents that were powering robotic arms for companies like Shell and PepsiCo and Abbott Nutrition when they had a Similac shortage. And doing that work and going, wow, this is really cool. Like I can actually help single mothers or just any mothers who are having a hard time accessing Simulac during COVID because I can create a robotic arm that can do that work so that people can stay at home and be able to avoid, you know, being too close to each other during those COVID things.

1:02:04So I think my career has always been guided by that passion of trying to figure out where I exist and how I can exist in a way that can give of whatever privilege that I have while also protecting other people who don't have it, while also acknowledging that I don't have all of the privilege in the world, but trying to find ways to express to people who do ways that we can all work together.

1:02:30Jon Krohn:You brought together in that final answer, you know, economics, social aspects, philosophy. In a few minutes, you managed to capture a lot of your influences and a lot of the most important factors driving so much that's happening in our world today or at any given point. And, uh, yeah, so I'm looking forward to see where, seeing where your journey takes you next and how you continue to, to answer questions about the world for yourself and for us through your publications. Really fascinating. So for people who are also excited about what you're up to, what is the, what's the best way to follow you in your work?

1:03:08Yeah. Uh, best way is LinkedIn. Um, occasionally I post articles on medium or towards data science. but I'm most consistent at LinkedIn. So that's the best place.

1:03:21Jon Krohn:And you probably also crossbook when you do something on those, on Medium or towards data science, you probably post better on LinkedIn as well. I do. Great. So one stop shop. Fantastic. And then I apologize that I was supposed to warn you about this last question before we started recording. I've forgotten a couple of times with guests recently. I always ask my guests for a book recommendation. And yeah, it doesn't need to be something in our space. It can be in our space. It can be fiction, nonfiction, whatever you want. Do you have anything for us? Yeah. Uh, values, value systems design by Batya Friedman.

1:03:59She wasn't the only one who wrote it, but she's the head author of it. That book has guided all of my decisions in AI, regardless of what function of AI I was working in. Essentially, the idea of the book is how do you make sure that you're building systems that are rooted in the values that are important to us as other humans? Because the technology that we create is simply an extinction of human values. And so it's absolutely amazing work. I think everybody should read it.

1:04:32Jon Krohn:Nice. Thank you for that recommendation, Jasmia. And thank you for this whole episode, taking time out of your busy day to, yeah, enlighten my listeners on so many topics. Really enjoyed this conversation, Jasmia, and hope to have you on the show again in the future. Yeah, it'd be awesome. Very interesting episode indeed with the brilliant Jasmia Henry in it. She covered her work at Collide where she builds foundation models for the petroleum industry in which 90 % of documents are unstructured and AI's accuracy can be the difference between a routine shift and someone losing a limb. She described her full stack foundation model building as having four stages, first curating and structuring data, then building bespoke tokenizers and embeddings, training the model through continued pre-training and reinforcement learning is the third step, and then finally optimizing for inference at scale.

1:05:25Jon Krohn:She also talked about how reinforcement learning models are bursty because they idle the GPU during reward calculation and then dump enormous loads on it all at once. And finally, that reward hacking happens when a model discovers the laziest path to a high reward rather than actually learning the task we wanted it to. As always, you can get all the show notes, including the transcript for this episode, the video recording, any materials mentioned on the show, the URLs for Jasmia's social media profiles, as well as my own at superdatascience.com slash 995. Thanks, of course, to everyone on the Super Data Science Podcast team, our podcast manager, Sonia Breivich, media editor, Mario Pombo, partnerships manager, Natalie Zajski, researcher, Serge Massis, and our founder, Kirill Aramengo.

1:06:07Jon Krohn:Thanks to all of them for producing another super episode for us today for enabling that super team to create this free podcast for you. we are deeply grateful to our sponsors you can support the show by checking out our sponsors links which are in the show notes and if you'd ever like to sponsor the show yourself you can find out how at johnkrone.com slash podcast otherwise you can help us out by sharing this episode with folks who would like to learn about end-to-end foundation model training reinforcement learning and so on review the episode on your favorite podcasting platform or on youtube subscribe if you're not already a subscriber.

1:06:43Jon Krohn:But most importantly, I just hope you'll keep on tuning in. I'm so grateful to have you listening and I hope I can continue to make episodes you'd love for years and years to come. Till next time, keep on rocking it out there and I'm looking forward to enjoying another round of the Super Data Science Podcast with you very soon.