Sasha Malysheva

Building something for LLM communication
PhD in RL · ex-DeepMind, Google X · @southpkcommons F25

[Sasha Malysheva @aimalysheva]

Hi all, thanks for checking out my page! Wanted to share a bit more about my background than the X bio allows:
- Finished school at 16
- Started as an AI researcher at JetBrains at 18
- By 21 I was coordinating a research direction there and published 5 papers in one year
- Then I joined Google DeepMind Game Theory Group
- Published a paper in Science at 23 (I honestly thought that milestone would come closer to 40)
- Later went to Google X to work on an exoskeleton project (wanted to see what the product side looked like)

Now I'm building something new. Stick around ✨

[Sasha Malysheva @aimalysheva]

@TamazGadaev @thinkymachines @miramurati @lilianweng @soumithchintala @ssi @ilyasut @reflection_ai @MishaLaskin @real_ioannis @poolsideai @jasoncwarner @eisokant Not yet, but I definitely will!

[Sasha Malysheva @aimalysheva]

continuing to build the neolab people list, adding more today:
@thinkymachines (@miramurati, @lilianweng, @soumithchintala)
@ssi (@ilyasut)
@reflection_ai (@MishaLaskin, @real_ioannis)
@poolsideai (@jasoncwarner, @eisokant)

for context: I'm building an X list of neolabs + the people behind them as they tend to be more interesting to follow directly

follow the list if you want to track who's actually building at the frontier:
https://t.co/mIJB2Ml8GO

[Sasha Malysheva @aimalysheva]

continuing with robotics, a field close to my heart 🤖

- @SkildAI (@pathak2206)
- @physical_int (@hausman_k)
- @RhodaAI (@startupjag, @GordonWetzstein)

[Sasha Malysheva @aimalysheva]

@ktoya_me @thinkymachines @miramurati @lilianweng @soumithchintala @ssi @ilyasut @reflection_ai @MishaLaskin @real_ioannis @poolsideai @jasoncwarner @eisokant Great idea, will do!

[Sasha Malysheva @aimalysheva]

@avgaydashenko 💯

[Sasha Malysheva @aimalysheva]

@avgaydashenko this is the best outcome of a good thread. bookmarked

[Sasha Malysheva @aimalysheva]

@TamazGadaev the implementation plan for the agent to track its own work is a small thing that makes a large difference. agents lose context less when they have a map

[Sasha Malysheva @aimalysheva]

just updated a skill that does exactly this: breaks a paper into abstract, summary, and other useful layers so the agent gets compact, structured context instead of raw PDF garbage
if you're doing any research and this sounds useful, comment or DM and I'll send the file

[Tamaz Gadaev @TamazGadaev]

the main problem with how agents handle papers is that they use vanilla pdf parser by default so their context fills up with useless garbage

a simple skill that pulls clean text with a well-written script and summarizes from that is a must-have for any researcher

[Sasha Malysheva @aimalysheva]

my rule for when to make a new Claude skill: the third time I find myself typing the same prompt

[Sasha Malysheva @aimalysheva]

@heydyago 😂

[Sasha Malysheva @aimalysheva]

@urban_geek123 @SkildAI @pathak2206 @physical_int @hausman_k @RhodaAI @startupjag @GordonWetzstein @TheHumanoidAI thanks, added!

[Sasha Malysheva @aimalysheva]

continuing with robotics, a field close to my heart 🤖

- @SkildAI (@pathak2206)
- @physical_int (@hausman_k)
- @RhodaAI (@startupjag, @GordonWetzstein)

[Sasha Malysheva @aimalysheva]

Realized I wanted a list like this but with associated people, not just the labs. So I made one 😁

Starting with math and video, both moving fast right now.

Math: @axiommathai (@CarinaLHong, @KenOno691) and @HarmonicMath (@tachim, @vladtenev)

Video: @moonvalley (@MateuszOnAI), @higgsfield (@alexmashrabov), @recraftai (@AnnaVDorogush)

Will keep adding. Follow the list if you want to track who's building in this space!
https://t.co/mIJB2MlGwm

[Sasha Malysheva @aimalysheva]

@burninganna @SkildAI @pathak2206 @physical_int @hausman_k @RhodaAI @startupjag @GordonWetzstein @KyleVedder thanks added!

[Sasha Malysheva @aimalysheva]

Sutton is brilliant. Must read or watch, highly recommend. If you want to know where RL is going, follow him.

[Richard Sutton @RichardSSutton]

A new and possibly controversial perspective:
In this video, I explain the sense in which generative AI trained by supervised learning is incapable of making novel discoveries.
https://t.co/zin5QbbT9N

The text of the speech:

AI Creativity and Discovery

Good day ladies and gentlemen. I regret that I am unable to be with you all today to engage in a back-and-forth discussion, but I am nevertheless pleased to be able to share with you, via this recording, some high-level thoughts about the current and future state of artificial intelligence, and in particular about AI’s relationship to science and mathematics, which is, as I understand it, the central focus of this meeting and of the SAIR Foundation.

I would like to start with an old joke; I am sure you have heard it before. It is the one about the researcher whose work is being evaluated, and the review comes back, and says “This work is both novel and good. Unfortunately, the parts that are good are not novel, and the parts that are novel are not good.”

My first point about AI is that this assessment applies exactly to large parts of AI as we know it today. Not all of today’s AI, but a large part of it. Pretty much all of what we mean by “Generative AI”---which includes large language models, and the images and video models, and even the new methods for learning world models. All of these AIs take large numbers of examples and produce a “model” which behaves similar to the examples, that is, which generates text like people, or images like artists or nature, and videos like we find on the internet. Don’t get me wrong, Generative AI can be extremely useful. No doubt about that. But the assessment of the joke still applies. These systems can produce output that is both novel and good, but not at the same time.

In many ways this is just absolutely not a problem. When we ask an AI for an answer from the internet, or to summarize a document, we don’t want it to be novel. We are happy if the quality of the answer, the goodness, comes from the source material—from the people who wrote the document or the articles on the internet. If the AI’s answer is novel it means it is going beyond the source material, adding something beyond it. This is what we call “hallucinations”. In most cases, we don’t like it when the AI makes something up, when it adds something novel.

One exception, of course, is when we are looking not for facts or reality, but for fiction and entertainment. We might ask for a bedtime story for a child, or an image based on existing images on the internet but which is nevertheless different and distinct from them. In these cases, it is never easy for us to know how creative the AI is actually being, as we do not know how close the AI’s story, poem, or image is to the source material. In a real practical sense we can not know this because the internet is too big, the possible sources that the AI may draw upon are too numerous.

When we ask for a fiction or novelty, the AI can give it to us because its processing is in part stochastic. Every decision can go multiple ways and will go different ways and produce a different trajectory every time. The trajectory can be random—and thus novel—or it can be based on the training data—and thus “good” because the training data is good, sourced from people or reality. Thus, the trajectory is either novel or good—based on randomness or based on data—but never both at the same time.

Really, I think it is okay if the output of Generative AI is never good and novel at the same time. For the researcher in the joke this is a devastating criticism, but for most things it is not, and for Generative AI it is not. Generative AI is meant to be a mimic. This is what supervised learning is for. Generative AI can be extremely useful, even when it just mimics, if it is faster, or cheaper, or smaller, or more customizable, or more copy-able, than the thing being mimicked. It is okay if Generative AI cannot be both novel and good at the same time. It is still a transformative technology.

But it is a limitation. And remember we are here to use AI for science and mathematics, and for these areas the assessment of the reviewer in the joke is devastating. For these areas we need true creativity and discovery. Generative AI—or Mimicking AI—will never get where us there. For these we need something more, and indeed we have something more in other parts of AI. We have many AI systems which can give us more. We have AlphaGo with its world-changing move 37, or AlphaZero with its brilliant original chess-playing style. We have GT-Sophy that drives simulated racecars better than any human. We have AlphaFold and AlphaProof and Claude-Code, which have brought true advances in science, mathematics, and programming. We have RL-Lyft which optimizes the assignment of cars to passengers in the ride-hailing business. All these systems have found things that are both novel and good. And, truth be told, some language models have been augmented in ways that make them more than Generative AI based on supervised learning.

All these systems have some additional features that make them capable of true creativity and true discovery. It is important for us to recognize what this is—and that it is not present in ordinary, garden-variety Generative AI. It is something that can not come from just supervised learning, from learning from examples. What is it? Well, it is a simple thing, a commonsense thing. It is not new. We have many names for it, but unfortunately none of them are very good names. I will call it Discovery. Basically, Discovery is just the idea of trying many things and seeing which of them work, then keeping those that worked the best. Evolution by natural selection works this way. The scientific method works this way. And just ordinary life and learning works this way. We try things and remember what works. What could be more obvious? In this behavioral case, psychology has two names for it— “instrumental learning” and “operant conditioning”—and in machine learning it is what we mean by “reinforcement learning”. We also see the idea of Discovery in planning and combinatorial search—anything that involves the idea of “generate and test”.

The essence of Discovery is to combine three steps:
1. Variation,
2. Evaluation, and
3. Selective retention.

Of course, I am not the first to say this. I am not the first to point out that this combination of steps is key to science, to evolution by natural selection, and to animal behavior. I think particularly of papers by Donald Campbell, by Daniel Dennett, and by Gary Cziko. What is new in my remarks is to directly relate the idea of Discovery to modern AI to help us see that it is not present in supervised learning or Generative AI—in particular, that Discovery is not present in backpropagation or gradient descent.

Let me say explicitly what is missing from Generative AI. As we have remarked, these systems do have a stochastic aspect, so they do generate a variety of trajectories and behavior. What is missing is the Evaluation step. The generator was pre-trained by supervised learning, leaving no way at runtime to Evaluate what it generates. And of course without Evaluation there can be no Selective retention, and thus no Discovery. The variation can bring novelty, but without evaluation there is no Discovery, and arguably, no creativity. That is, I would say that creativity requires that the new things generated be Evaluated. Without evaluation, and retention of the best, there is nothing created. The novelty flickers into existence but, if its value is unrecognized, it flickers away and is lost.

In many cases, Evaluation is done by people to make a discovery. As when we have Generative AI make many pictures for us, and then we pick the one that we like the best. The human+AI system completes the discovery.
In many other cases, the Evaluation comes from a clear objective. Some moves lead to checkmate, some steps lead to a proof, some actions result in high reward, some genotypes make more copies, some theories explain the data better.

Some prefer the Variation step to be called Blind variation, where “blind” here means that it is uninformed, a shot in the dark. It does not need to be completely uninformed; a good scientist does not select theories to test at random. But neither can it be completely informed and determined. There must be some uncertainty about where the answer lies in order for there to be a discovery. In practice, the variation is partly informed and partly blind, but it is the blind part that corresponds to the discovery.

Now let us briefly go all the way to modern deep learning, to the backpropagation algorithm. At first it might seem that backpropagation is incapable of discovery because it is deterministic and thus incapable of variation. But this is not correct. The weight updates of backprop are deterministic, but the weights are initialized to small random values. The random initialization is often downplayed, but in fact it is a necessary form of variation; it must be done properly to get good performance. In backprop this Variation is done once, at network initialization, so its effect is temporary, and later the network may lose its ability to learn. This is the weakness of deep learning that is alleviated with a new algorithm that my group presented in Nature a couple of years ago. Our “continual backpropagation” made one small change: every so often a less-used neuron would be re-initialized to small random weights. This allows the variation to continue and plasticity to be retained.

Although there is much more to be said about Creativity and Discovery, this is the key point: they are more than supervised learning, more than pattern recognition, more than prediction, and more than world modeling. Those things are important, but they alone will not bring us to discovery. Discovery requires Evaluation from a person or from an explicit goal, and only in the latter case will we attain full autonomy.

So that is my call to arms. If we want the full power of AI scientists, then we should share the goals with them so they can create, evaluate, discover, and in these ways fully participate in achieving the goals. Let’s be bold! Let’s fully automate Creativity and Discovery!

[Sasha Malysheva @aimalysheva]

Great work from my former JetBrains colleagues. The focal model philosophy is the right call, not every step in an agent pipeline needs a frontier model. Wish this kind of stuff could work without the pretrain though.

https://t.co/xPSLR90tf4

[Sasha Malysheva @aimalysheva]

Highly recommend!

[Aditya Agarwal @adityaag]

Building at the frontier?

Exploring what's worth building?

Both are good reasons to apply to @southpkcommons.
https://t.co/xWaXyqCFGR

[Sasha Malysheva @aimalysheva]

@TamazGadaev been hitting this exact issue! the diff between a good extraction script and the default is so surprisingly large

[Sasha Malysheva @aimalysheva]

@avgaydashenko @hlntnr The gender imbalance in AI (safety) is a real pipeline problem. If the field shapes who builds and governs transformative AI, that imbalance has compounding consequences

[Sasha Malysheva @aimalysheva]

posts like this make me really grateful for my time at @GoogleDeepMind, picked up a lot of useful practices there, including how to store papers properly

one simple one is creating subfolders for every topic (might sound obvious but it's what makes 200+ papers actually navigable)

[Tamaz Gadaev @TamazGadaev]

how do you store papers in your research project knowledge base?

plain .md for agents + pdf in the repo? or something smarter like a summary.md per paper?

[Sasha Malysheva @aimalysheva]

what do you actually read these days? newsletters, articles, Reddit? or is it mostly just Twitter at this point?