Transcript 0:00 During reinforcement learning, the model tries its best to hack a lot of things. The current open models can achieve that level of understanding. It is a solvable problem, and we are working on it. 0:11 Engineering is very, very, very important. I didn't know that during school. [upbeat music] Um, hello everyone. 0:24 Today I have the pleasure of talking to Olive Song, a senior researcher at MiniMax. And recently they've been launching very interesting open weight models specialized in different areas. 0:35 And Olive is currently working on MiniMax on the new version, MiniMax 2.2. Thank you for taking the time at 9:00 PM on Sunday night. Does everyone works like this at the company? I'm, I'm really impressed. 0:49 I think different people work on different schedules. We do have people who work even overnight, but they sleep at daytime. So like we have a very flexible schedule. You know, it goes with your experiment. 0:59 For example, if the experiment runs for all day, the person can take a break. 1:04 And then if, you know, there are a lot of analysis to do, maybe, 'cause we are very curious about the results and w- we're very passionate, right? We can't really wait a very long time. 1:13 So yeah, everyone has their own schedule. That's telling about the success [laughs] of the models. Think that's influenced that. 1:21 You specialize in reinforcement learning and model valuation, as far as I understand, which are two of the least forgiving parts of, um, model development. And you also have more constraints than big American AI labs. 1:34 What does a good day look like for you, and what does a bad one look like? I can share something about our recent weeks. So there's not a whole good day or a whole bad day. 1:44 We were joking that during one day we have good results in the morning, and then some sometimes it becomes bad results at night. 1:50 You know, sometimes we call it, um, we have, like, ICU in the morning and then KTV at night. 1:56 So typically, a good time would be usually receiving some good results or, like, even if running into new problems is a good time. 2:04 So if we, for example, during reinforcement learning, right, we can see the model doing a lot of different stuff to achieve the results, and sometimes we just discover, like, new model behaviors, and that's really exciting even though it might not be safe or, like, it might not be expected. 2:18 It's kind of exciting, so I, I call it a good time. A bad time would be... So there really isn't a bad time except for, you know, looking at the, you know, just finding out the bad results. 2:29 The moment itself is bad, but then trying to figure out the problem and breaking it down is pretty good time. What were the recent model behavior that you didn't expect? 2:40 So the model, during reinforcement learning, the model tries its best to hack a lot of things, right? 2:45 So for example, it uses bashes a lot, and sometimes it might not be very safe behaviors as our expert developers say, 'cause sometimes the expert developers have their own expectations on how the model works, but then it doesn't go that way, you know- Mm-hmm... 3:00 if we don't constrain it. So we do a lot of alignment to solve that issue. You just launched MiniMax-HER, and that goes, like, all over the Twitter. How do you come up with those ideas? 3:12 Because, you know, role playing is sort of, like, is it an alignment question? Is it not? Like, how do you, how do you do that? [laughs] Frankly speaking, I'm not the expert person on that part. 3:22 We have a whole team on, um, role playing HER stuff. I'm not an expert, but we do have a lot of discussions. 3:29 Um, we do believe that role playing or, like, you know, accompanying human or, like, human interactions is re- very important in the life with AI or, like, how it would change our social life in the future, and it absolutely represents some ability that's very superior 'cause that's, like, human-like. 3:47 You know, it has emotions where it understands your emotions. It's not just, you know, working out some exams. That's absolutely another side of the AI capability. What it's called, AI with everyone, right, in MiniMax? 3:59 Yeah. It, intelligence with everyone. Intelligence, intelligence with everyone. Uh, what does it mean for you? 4:05 For me personally, I feel like it's more like ch- how it changes my life and enables me to do more work, and then how it can connect me better to different people. 4:15 'Cause, for example, before I wouldn't be able to understand a lot of very professional, for example, right, very professional coding problems or, like, optimization problems, and now I am able to do that with AI, and so that I can communicate with more people and ex- exchange more ideas. 4:32 That's one side, and on the other side, it generally helps my daily life, right? 4:36 So it helps with my work, my daily routine, my self-care, changes life for me, and I hope that it changes life for everybody, obviously in the good way. 4:44 Can you tell me a little bit how day-to-day work organized in your lab? I remember from your talk at AI Engineer that it's very interconnected between developers and researchers. I would love to hear more about that. 4:55 Absolutely. We sit around every day. So we share our experiment results. For example, as I just said, during experiments, for example, reinforcement m- learning experiments, we see some scores going up high. 5:08 We look at the model's behaviors, and we look at the model behaviors with the developers in that area as well. 5:15 We sit together, and then they will spot the issue right away, and then we, we are able to come up with new ideas to f- fix it or, like, building more data on it. If we can go into details- Mm-hmm... 5:26 like your current work on the current model, uh, the current version, what are the biggest problems you're trying to solve comparing to the previous version? 5:34 One important thing we focus on right now and also, uh, in the future is human alignment 'cause we are focusing on Coding models for 2.1, 2.2, and the M2 series, right? 5:46 And then what we realize is that for it to become very productive in our daily work, or like for it be productive and safe at the same time, we have to do a lot of alignment on it, so the model can't just grow on its own and then do some dangerous behaviors just to achieve the final goal. 6:04 So for us, the important thing would be how we define human alignment, how we define expert expectation, how we actually train the model to be more aligned with our expectations. 6:16 So I want to go, like, in some real details here, and you're an expert here, so correct me if I'm wrong, but I saw that there was a recent interest in details like keeping the LM pad in FP32 during reinforcement learning training. 6:31 Why do small decisions like this end up mattering more than just, like, a clever new algorithm? It all ends up being closer to the theoretical algorithm. 6:41 So we have the theoretical reinforcement learning algorithm, and then but when we implement it, it could be a little bit off. That creates a little bit of gap to the theoretical extreme of this algorithm. 6:54 So that's how we think and approach to this problem is that we try to, you know, scale to the theoretical extreme. 7:01 And for example, the precision part is one thing that we found that would, you know, kind of, like, prevent us from being close to that extreme, and that- that's how we solve it. 7:12 That was a very funny story, actually, for when we discovered that. I kind of talked about it when we published, um, Minimax M1. So during our experiments, we found that the accuracy didn't go up. 7:23 W- we looked at, like, layer by layer. We looked at the log prompts layer by layer and f- found it out. Theoretically speaking, it, it has to work, right? 7:32 And then there has to be some gap between the theoretical and, uh, how we approached it. So we thought about the gap, and we'll, uh, analyze it layer by layer, and then eventually found it. 7:42 Is there anything like this happening now? Definitely, yeah, every single day, every single day, and in every different groups. 7:49 I can't actually disclose something that we haven't had the concrete conclusion, 'cause we want our conclusions for, like, anything public to be very concrete, and we understand it very deeply. 7:59 So if we have, you know, breakthroughs, we'll definitely publish it later. But I'll have to say, we, it, do encounter these problems every day, and we think... I think it's called the first principle, right? 8:10 So we think from the very fundamental part of the problem and then approach it. 8:14 The models that you launch are open weights, and from your perspective and from the alignment perspective, what do builders actually gain from open weights, and what responsibility do they have to take on that, like, you don't have to take responsibility for this? 8:29 Again, I'm actually not an expert in building developments or, like, building things with models. I feel like because it's open weight, people can have, like, freer use with it. 8:40 For example, they can deploy it on, uh, by themselves, or, like, they can even fine-tune it with the weights and then have, you know, all the data on their, their properties. It's very safe. 8:50 But if we talk about alignment, how do you look at that from that perspective when the model is out there in the wild? Before you launch the model, before you publish it, what does tell you that it's safe to publish? 9:02 We have some internal benchmarks in terms of safety, and it has different dimensions, something that's sensitive safety or something that's, like, alignment safety. 9:11 We have that as our evaluation, and then before launching, about one or two weeks before launching, we do scaled up evaluations, and we do scaled up alignments on the model, and that's how we assess if the model is safe. 9:25 But then if it's already open weight in the wild, people actually can do something on it. I guess that's what you are approaching at, right? People can do more things on the model that we can't control. 9:38 I don't know how we handle that, frankly speaking. There are laws on that, right? There are regulations where, like, people do agree on some moral standards on that. 9:47 Do you see, do you follow any reinforcement learning failure modes that haven't showed up on benchmarks but then become obvious in real agent use? 9:57 How do you collect feedback for the next versions for improving the reinforcement learning process? W- we collect feedbacks on the model itself first. 10:06 So when we publish a model outside, many developers use it or, like, many people use it. We collect it systematically. We analyze these problems. Some of them are fundamental. 10:17 Some of them are, you know, just something that we missed, and we can fix it real quick. So there are two parts. 10:22 First, we do the internal evaluation with the developers, and they point out problems, and that's how we can fix this part, but they are not enough, and, uh, more feedbacks will come to us after we officially publish the models, and then we collect it. 10:34 'Cause, uh, the way we organize our group is that different people work on different capabilities of a general model. 10:42 If we collect some things that we think we should improve in the future, different people take their parts, right? 10:47 So they're like, "Okay, I, I think I can solve this issue, and I'll solve it next, in the next gen- uh, generation," and that's how we, you know, collect feedbacks and then improve the model. 10:57 How did you initially decide to not build one general use model, everything for all, and go more into specialization, like coding? I think we are approaching generalized models. 11:09 It's just that we are putting more emphasis on coding. For example, our model also, you can take it into any general agent scaffold, including our own agent product, and that's for general purpose. 11:22 We do work on, like, you know, researching, report writing, PPT... stuff like that that's more general. 11:30 Personally speaking, I feel like with coding you can, you know, kind of structure the whole world, or, like, you can model a lot of stuff with coding. Yeah, engineer it. Yeah, with engineering. 11:39 So behind it, it's, it's scaled up humanity for me. 11:42 So it itself has a lot of intelligence with it and a lot of work to do, so that's how we view this issue, but we do work on generalized stuff, and even more generalized stuff in later versions. For example, we- Mm-hmm... 11:57 uh, our model can do some general workplace scenarios in the future, and that's not just coding, right? If we talk about coding and agentic use, it requires long horizon. How do you solve long horizon for agentic use? 12:11 I think define your goals good and define the model behavior is good. And also, we require great infra- extraordinary infrastructure. For example, for reinforcement learning, right? 12:23 The very important issue besides algorithm, besides things that people have been working on for a very long time, what's special for agentic stuff is how we define agents, how we define how a agent model would work. 12:36 First, it... you need to define the task, you need to define the model's goal, especially in a long horizon task, you need, need goals that are actually hard and diverse. 12:45 And then the second part is that you need environments. 12:48 You need great engineering environments, scaled up environments, different, diverse environments, not just coding or, like, more, for example, workplace, different kind of tools. That's great engineering. 12:59 And then you need great infrastructure. 13:01 You need outstanding RL infrastructure to let the model really roll out in a very long horizon and then with, you know, very efficient, for example, GPU use, very efficient, uh, training, rollout and training and stuff. 13:15 I feel like that's what's different in agentic reinforcement learning as compared to before. Are you affected by GPU's constraint? How do you solve the compute problem? 13:25 We do have a team that works on how we utilize the computes the most. You know, that's actually one of the RL scaling issue is to utilize the compute very efficiently. 13:35 So their purpose would be to minimize the compute use and then while train more, right? Personally speaking for me, I don't really have a GPU constraint. 13:44 It's that we have a great team who works on it to utilize the compute most while, you know, stabilize the training the most. 13:52 But do you have problems that you need to solve with your expertise, how to use more efficiently, or it's just that team? We are actually the, the same team, 'cause [laughs] Ah... 14:00 we're actually the reinforcement learning team. We view this issue on different perspectives. It can be, you know, implementation, right? It can be implementation. 14:08 You can view it as a data perspective, you can view it at different perspectives, but our goal is the same. 14:14 We always, you know, we're always looking forward to new solution that comes from Chinese labs because it always mind-blowing. 14:21 [laughs] We are actually working on some new agentic, uh, reinforcement learning stuff, but, uh, it won't really come out with 2.2. With the next generation model, we are actually s- we are still working on it. 14:32 I'm not sure what I can share or not, so [laughs] I can share it later when I have concrete conclusions, as I said before. 14:39 I can't really say something that we, you know, don't document yet and- Will it be available when the model is out? That depends on our timeline. I'm not very confident yet. But we are dedicatedly working on it. 14:54 Yeah, a lot of constraints talking to researchers. [laughs] So many secrets. Well, if we talk about openness, then this whole conversation that I'm having with people right now in this quarter is about open source. 15:05 I wonder if you can talk about the company strategy, why the company decided to go and publish open weights of the models. What's the benefits? Mm-hmm. What's the cons to that? 15:16 So for our team, like, for the researchers team, we al- we always wanted to go open source, 'cause, you know, open source communities is, is fantastic. 15:24 I learned that from day one when I joined the team, is open source community is fantastic. So as researchers, we did want to, you know, join the open source. 15:32 But then on the other hand, speaking of the cons, we are a company that people can care about if this can be, you know, sell money or, like, if this is a business. 15:42 So the cons would be if w- the, the weights are open source, less people will use APIs. But then as a researcher, that really isn't my focus that much, so I'm not very confident about the company strategy. 15:56 For the tech part, we just believe that we can build better models with the open source community. How much do you use open source tools yourself from different other companies? A lot. 16:06 For example, inference, we use, I'm not sure if I'm allowed to say specific open source branches, but then we collaborate with both ViLLM and SGLM. They are, they are open source code repositories. 16:18 How do you look at open source stack? Because when we talk open source, sometimes it's perceived as one thing, but actually, like, m- it's, it's multi-layered. How do you look at it? 16:27 Uh, for example, there are a lot of open source agent scaffolds, both coding agents and agent-agent scaffolds that we use ourselves to test our models, and then we look at their logics, we look at their code to see how the design specific, both scaffolds and, for example, engines, and then we take what they worked on was, you know, really good, and then we reflect on how we think the problem, how we structure the problem, if we're on the same page and stuff like that. 16:55 So we learn from each other. Do you think team underestimate how much, like, engineering discipline open models require comparing to using closed APIs? 17:05 It always requires a lot of, like, setting up, and it's a different compute, and you need to have a talent for that to use it, engineering talent, instead of just like, you know, choosing closed API, turn it on, and, and use it. 17:19 Do you have any difficulties to do that, or just inside the company, the open source stack is, like, established and, and working? Personally, I don't have a problem with that. 17:28 There are other open source models, and if they publish, I'll just download it and, and, you know, deploy it on our machine and then work with it if I want. 17:36 Personally, I don't have that issue, but if there are, you know, personal developers out in the wild, I understand the problem, especially when they don't have their own compute, and then it would be easier to, you know, connect to a model through, for example, Open Router and stuff like that. 17:51 Do you use a lot of other open models on the, the same Open Router, let's say? Do you play with them? Yeah, I play with them. I would play with them day one. If they release at midnight, I play with them in midnight. 18:02 [laughs] Are you, like, taking notes? 18:04 [laughs] I don't actually take notes, but I do have my personal evaluation stack, a- a list of different fun questions that I like to test with every single model to see how they work. Can you tell me about it? 18:14 That's super interesting. 18:16 Um, yeah, I've been collecting a bunch of questions since I entered the company on different areas, including, uh, logical reasoning, including mathematics, proofs, including report writing, including agentic tasks, and stuff like that a lot. 18:32 And some of them are very... You know, I, I just like to see how the model reacts to these problems and how they approach it, and different model have different personalities when approaching something. 18:42 That's true, and you always need to adjust them. If we want to give, like, sort of a little guide to people who want to evaluate a model themselves, can you give me examples of the questions? 18:54 Like, five questions you need to ask the model to understand how it works- [laughs]... if it works well. From the professional evaluation perspective, five questions isn't enough. 19:04 [laughs] So if you want to do, like, you know, very standard and very fair comparison among models, uh, we have to make it very a confident test. 19:14 So there has to be a certain number of questions in each domain to see how the model performs, and usually you need to test it multiple times, 'cause models are not very stable themselves. 19:27 If you are testing for fun, use the fun questions. But if we are actually assessing the model's capabilities, we need some sets and that's, you know, very fair among different models. 19:39 That's correct, 'cause some problems are not correct. Some questions, the, the answers are not single, for example. Um, and sometimes when we run the test, the, the environments are not fixed, for example. 19:50 For example- Mm... the goal you had, this wouldn't pass, and stuff like that. 19:53 So if we're doing professional evaluation, we have to make, make sure the, uh, evaluations is correct, it's diverse, it's above a certain threshold so that the test is, you know, confident. You mentioned characters. 20:05 How do you work with, uh, your model character? I don't work on my model's characters. Okay, that's how I think of this issue. 20:13 A general model should have all characters, or, like, it should be able to perform all characters. It might have default character. If the user want it to be a different character, it should be. 20:24 If the model injected into the system prompt, it should be. That's how I view this issue. I find it hard to adjust to the new models because they're so different in terms of character all the time. 20:36 I just don't even understand why it happens. I think it has to be c- be something related to the data that the model was trained on, the different patterns the models have been trained on. 20:49 And also different people, a different team might have their own, you know, constitution. In the system prompt? Or, like, as the model's default behavior. 20:59 If you look at open models in production today, I don't know if it's, if it's a relevant question, but where do they fail first? Open models specific, like reasoning, tool use, state tracking, evaluation blind spots. 21:14 Like, there are all those risks for open model. Where does it break first? I think open models are not very good at adjusting to different environments from what I see right now. We can see, for example, Claude, right? 21:27 People use Claude in different coding environments, and then people think they perform well in all environments for, like, different tool definitions and stuff, but I don't feel like the current open models can achieve that accuracy or, like, that level of understanding of the different environments. 21:44 Why? Where, where is the problem? I don't know how Claude does it, but for me, I think it is a solvable problem, and we are working on it. We are improving it in 2.2, but it's still not as good as, for example, Opus. 21:58 But for 2.5 it might be. We do have some systematic research going on in the area that has shown some results now, but it still is not concrete conclusion, so I won't say it. 22:10 [laughs] I'm so curious, but do you think it's the problem of compute? Because they have this infinite amount they can just throw at it. 22:16 I feel like compute is one side, but how we structure the problem and how we approach it is another side, and that's where we're more confident on, that we can solve the issue. 22:26 What can you tell me about M 2.2 if it's launched by the time that the interview is out? Can you give me some overview? Better coding, obviously, and better multilingual coding, obviously, and more stable than before. 22:40 It has better performance in the area of 2.1. In different areas it's better, more stabilized, longer horizons and stuff like that. 22:49 And we are testing it in different environments right now, and we believe that it's better than before, so different coding environments, right? 22:57 Even environments that we haven't seen before, even environments that are, you know, totally out of distribution, we see some very promising scores that are higher than 2.1. 23:08 I wonder, how do you stay updated to everything that happens, which is super hard because the pace is just insane. You said when the models are out, you're playing with them. Do you read research papers? 23:21 What is your, like, other interest that-... helps you kind of cross-pollinate with what you do. Can you tell me how you stay up to date and what inspires you? 23:30 There are different articles, different blogs [laughs] going out every single day, and a bunch, you know, all the information. 23:36 How we deal with it is that we have a internal agent that tracks [laughs] all the new articles and blogs and papers, and then it dispatches to different subjects, and then it summarizes, and then it analyzes- Oh... 23:50 and researches. 23:50 So, so we have a internal researcher, if I call it, that does some filtering by, by itself, and then it gives what is filtered to us, and then we can improve the researchers if we think it doesn't do well. 24:03 And that's how we, how we filter out a lot of information first. And then we play with new code repositories with coding agents so that we can understand it more quickly and then play with it more quickly. 24:18 So we're pacing up with the, you know, with all the improvements with agents and with our models, for our models. [laughs] That's fascinating. 24:27 When you became a researcher, when you chose this path, what was you thought you would be doing, and what you actually doing? Like, is it close to what you thought? That's a really good question. 24:39 Um, when I joined the team, I thought I would be reading papers every day, 'cause that's what I was doing during school, right? During a labs. We would read papers, come up with ideas, implement ideas, run experiments. 24:52 If the re- experiments results are good, we run it as a larger scale. I thought I was, I was about to do that. 24:57 But then what I realized was that when joining the company and then, you know, working for a couple of months, you already become pretty much top of the area or, like- Mm-hmm... 25:07 of the industry, and then you have to come up with something that's really new, or, like, you encounter problems that you just don't know how to, how to solve. 25:15 It's not like you can read a lot of papers and then building up your thinkings on the papers. 25:21 It's more like you need to really understand the problems from the fundamental and then think of it from the fundamental- Mm-hmm... so that you can find the right solution. 25:29 And another things would be that engineering is very, very, very important. I didn't know that during school, 'cause, um, during school or during a labs, it's more toys as compared to companies. It's not that scaled up. 25:40 But when you re- do scale up data, you scale up compute, scalable people, right? You encounter engineering issues that you need to tackle very beautifully, and engineering is very important. 25:54 That's part two that was different from what I imagined. Pretty much these two, I feel like. 25:59 When you work on the model currently, is it mostly that you're solving problems that you see immediately from your hands-on work, or is it that this, the company says, "Oh, we have to achieve, let's say, Opus results"? 26:14 How do you set the goals? Uh, we have a, like a meta goal at the company level. 26:19 For example, we want to improve the AI's capabilities in u- you know, improving productivity, for example, 'cause that's how, like, people view... So we have a company's mission. 26:31 As a single researcher in the team, we have our own missions that we set our own goals with. What is your goal currently? 26:38 For the next generation, I would be, uh, I really want the model to, to be working elegantly with experts. So it's more like better collaboration with experts, with developers. 26:49 That's my goal as well, but that's maybe, like, two versions away, I think. I think we're, um, launching one version about per month or a month and a half, right? 26:59 For longer horizon, we are, we are definitely working on it. But then for me, for the goal that I set on along that path, that's like a three-month away thing. 27:09 But for the cl- better collaboration thing, that's like a one month or two months away thing. 27:13 I wanted to ask you a little clarification question about interleaving learning, what you were talking at AI engineer also, that the model doesn't set on, like, one action. 27:24 It's constantly, like, in the loop of asking more questions and trying things. How do you look at it? Is it a continual learning? Is it part of it? 27:31 What do we need to solve to have the model continuously doing this learning for, like, longer and longer horizons? 27:38 That has some overlaps with defined concept of continual learning, and by overlap, I mean, I think both conceptually and technically. 27:47 But I don't feel like they are exactly the same, or, like, the things that I talked about at the summit was not at the level of the full continual learning. It's more, like, on the path to that. 27:58 How do you see it being solved? Any ideas? 28:01 We do think that's a different, different problem definition, or, like, that's a different way of the model working with people, and we are working on that norm with our own defined question. 28:15 But if I need to say how we approach it, I would say to, we would approach it through experiments. [laughs] That's a very interesting question on continual learning, and I... and it's very, still very exploratory, right? 28:29 That's definitely what we are going at, but then it has different phases or, like, different stages. We might approach stage one first while exploring more stages later. And they're not yet outlined, the stages? 28:44 Outlining stages, we do have our internal definitions that I didn't prepare today. I would say first would be to be more sta- uh, stabilized into long horizon tasks and what I said a- at the summit, right? 28:55 And then the next thing would be, um- Optimization If you can repeat it, because the people don't know what you said. So for example, w- we see a model, it receives, you know, environment feedbacks in a new environment. 29:07 It needs to know what to explore and what environments to see, 'cause it's a partially observed environment. 29:13 It needs to know the actions that it takes to receive better information and then better reactions, and then perform harder, more complex tasks in the environments. That's, like, more, more of stage one, right? 29:23 That's pretty simple. Basically, all agentive models can do that to some extent. Maybe not perfectly, but to some extent, and that's how we can actually solve it with our current algorithms. 29:33 But we do see different norms of how model improves itself in the environment that we don't have a concrete conclusion yet. Maybe in 2.5 we will. That will be a different definition than what I said. 29:46 The model itself would be defining its own goal. That's something that, that would be different. Thank you so much. My last question is about AGI. Do you believe in AGI, and if yes, how does it look to you? 29:57 Okay, that's a very large question. People talk about AGI and ASI every day. 30:03 Actually, when I was interviewing with MiniMax, when I was interviewing with our CEO, I said the same thing, 'cause he asked me the same thing, right? 30:11 And then what I said was that I think people talk about AGI, people have different definitions of AGI, but we can only know the definition of AGI when we achieve it. 30:21 Or, like, it is still progressing so fast that the definition even changes every day, and people have different comments on it. 30:28 But what I think is more important is we actually, you know, work towards it, work towards our own definitions of AGI. And, and as long as we, we figure it out, it becomes true. 30:38 And that's what I said during the interview, and that's still my view today. The definition will become true when it becomes true. When we see it, we know it's AGI. Yes, exactly. But we're not there yet. No. 30:49 There can still be better AI intelligence, for sure. Thank you. One more last question. What was the book that influenced you the most? And it can be recent book or a book from your childhood. 31:01 Let me just double-check the name, though. [laughs] Something like The Art of Creativity or something, that I read, um, during, uh, during undergrad, so it's a long time. I don't remember the exact name. 31:11 Yeah, there is a book, the name Art of the Creativity. How did it influence you? It opened up how I think of my own mind a lot, and then how I view the world and how I view problem-solving. 31:20 For me now, problem-solving is, is more a discovery. That's how I would summarize it in one quote. Thank you so much. Thank you for your time. Uh, that was very interesting. Thank you for having me. 31:31 [laughs] [gentle music]