AlphaFold, every mention
59 scenes, the whole family · ← back to AlphaFold
tap a year for its mentions
every year anyone Gabriele Corso 16Matt McPartlon 13Jeremy Wohlwend 7RJ Haneke 6Heather Kulik 6Bo Wang 3Pim de Witte 2Sergei Yudinov 1Priscilla Chan 1Pranav Reddy 1
Verbatim, from the transcripts: the passages where AlphaFold comes up
🔬 Why Transformers Hit a Wall the Moment Physics Shows Up — Anima Anandkumar, Caltech
- ▶ 44:47 unnamed speaker So from, you know, my own domain is probably closer to computational biology, but, you know, alpha fold is the obvious, uh, like really exciting development in the community.
🔬They Thought the Model Was Broken — Matt McPartlon & Neil Patil, Chai Discovery
- ▶ 2:15 Matt McPartlon So this is like alpha fold one days, um, and it was in the field during alpha fold two.
- ▶ 2:15 Matt McPartlon So this is like alpha fold one days, um, and it was in the field during alpha fold two.
- ▶ 6:07 Matt McPartlon Outfold one was like, and outfold two was like this huge breakthrough, but then like outfold two multimer came out like a year later.
- ▶ 6:07 Matt McPartlon Outfold one was like, and outfold two was like this huge breakthrough, but then like outfold two multimer came out like a year later.
- ▶ 17:54 Matt McPartlon And then kind of like that was right when outfold three came out and we were, we'd like been talking about like, man, we really need like an MSA pipeline. 2 times in the scene
- ▶ 27:11 Neil Patil Like, I mean, CHI-II alpha-fold, very useful because you can, you know, you can at least intuit and reason about the structure and, and see what you're looking at,
- ▶ 30:29 Matt McPartlon What was done at the time is like, we kind of came up with a bunch of metrics and like alpha fold, it really is what enabled this.
- ▶ 53:57 Matt McPartlon And like a lot of people think outfold to like solve structure prediction. 3 times in the scene
- ▶ 1:04:31 Matt McPartlon In that case, we're following a fold two, three architecture.
- ▶ 1:09:03 Matt McPartlon For example, I think like outfold three, I might get this number wrong, but I think it was like 23 sub modules. 2 times in the scene
- ▶ 1:09:47 unnamed speaker But the only way you can accomplish that is, I mean, the reason AlphaFold II and AlphaFold III worked, they were small models, relatively speaking, they were very compute intensive, but they were very data efficient.
- ▶ 1:11:37 unnamed speaker There was the, there was that apple paper where they distilled on the alpha fold, which it was actually really cool that you could distill on a very large data set and you could get, you know, good signal, but you know, it didn't… 2 times in the scene
🔬Causal Models Need Causal Data - Xaira’s X-Cell model (Bo Wang & Ci Chu)
- ▶ 9:40 Ci Chu That ushered in this revolution in the printing design and alpha fold and other folding models.
- ▶ 27:30 unnamed speaker That trained alpha-fold models, right?
- ▶ 1:12:56 Bo Wang Like Alphafold, for example.
- ▶ 1:20:33 Bo Wang Because of the availability of open source data sets in PDBs, therefore we have models such as AlphaFold, RosettaFold. 2 times in the scene
🔬 RL with Verifiable Rewards, but the Verifier is a Lab — Lila Sciences
🔬 "The Most Innovative Diffusion Research Is Happening in Drug Discovery, Not Image Generation"
- ▶ 13:14 Brandon Anderson AlphaFold three, and then some of the, and then Bolts and, um, OpenFold three and so on have also implemented some of these ideas in their own way.
- ▶ 24:24 Evan Feinberg There is this, ah, a few papers that came out, one was in Cell, I think last year, which showed that for all of the claims about AlphaFold-solving drug discovery, people try to take AlphaFold-produced protein structures, use them for…
- ▶ 35:58 Sergei Yudinov I keep joking about that when Nobel Prize was given for AlphaFold III, a lot of people thought that, like, drug discovery is solved.
🔬 The Limits of AI in Science - Why We Need Self-Driving Labs — Joseph Krause, Radical AI
- ▶ 46:25 unnamed speaker Where she said that there is no alpha fold for materials. 3 times in the scene
🔬 The Bitter Lesson is Coming for Proteins - Alex Rives, BioHub
- ▶ 25:27 RJ Haneke So, I mean, this is very much in contrast to something like AlphaFold, right?
- ▶ 30:33 unnamed speaker One of the common reasons, you know, for this is if you're in the alpha fold paradigm,
- ▶ 32:01 RJ Haneke Using MSA's multistimus alignments, which was one of the, or maybe the critical insight that allowed AlphaFold to work really well.
- ▶ 32:10 RJ Haneke And the fact that you didn't need that in order to make it work basically as well as AlphaFold III is really exciting to me because that means that your thesis of let's, let's cover the space of possible proteins and as well as we can and…
- ▶ 39:21 Alex Rives You know, you can think of ESM as, as being, you know, first generation, AlphaFold as being a first generation of those kinds of approaches.
🔬There Is No AlphaFold for Materials — AI for Materials Discovery with Heather Kulik
- ▶ 18:00 unnamed speaker And people have been working on that for a while, and this led to AlphaFold, like, kind of, without CASP, AlphaFold probably wouldn't exist. 4 times in the scene
- ▶ 24:26 Heather Kulik The challenge here is that, um, what AlphaFold has done really well is, is predict structures of globular proteins, primarily with 20, um, natural amino acids. 6 times in the scene
🔬Generating Molecules, Not Just Models
- ▶ 1:07 unnamed speaker Uh, I guess we're maybe, what, six years post AlphaFold II right now, which was, like, kind of a big moment. 4 times in the scene
- ▶ 1:43 Gabriele Corso So what AlphaFold, so maybe first as a, kind of, introduction for the ones in the audience and not structural biologists. 2 times in the scene
- ▶ 4:11 Jeremy Wohlwend Our, our group at the time was, um, working along, like, small molecules already, and, um, I think AlphaFold is kind of what triggered, I think, this shift to, like, working on, on biologics. 2 times in the scene
- ▶ 6:45 unnamed speaker Going back to the AlphaFold II moment, like, um, I remember this very well. 2 times in the scene
- ▶ 15:40 Jeremy Wohlwend There's this nice line in the, um, I think it's in the AlphaFold II manuscript.
- ▶ 17:38 unnamed speaker So, um, yeah, that like, and that there were many, many, many millions of computational hours spent trying to solve this problem before alpha fold.
- ▶ 20:04 Gabriele Corso One interesting, uh, explanation about how Half-Fold Three works that I think it's quite insightful, of course doesn't cover kind of the entirety of, of what Half-Fold does. 2 times in the scene
- ▶ 20:04 Gabriele Corso One interesting, uh, explanation about how Half-Fold Three works that I think it's quite insightful, of course doesn't cover kind of the entirety of, of what Half-Fold does.
- ▶ 21:28 unnamed speaker You mentioned AlphaFold III. 4 times in the scene
- ▶ 21:33 unnamed speaker AlphaFold II came out and it was like, I think, fairly groundbreaking for this field. 3 times in the scene
- ▶ 24:37 Gabriele Corso Yeah, so one, uh, critical one that was not necessarily just unique to AlphaFoldt-III, but there were actually, um, a few other teams, including ours in the field that proposed this, was moving from, you know, modeling structure prediction… 2 times in the scene
- ▶ 29:20 Jeremy Wohlwend The other part, I think that's enough for three, uh, is sort of
- ▶ 34:57 Gabriele Corso Um, now, you know, models like AlphaVault II and AlphaVault III are, you know, still very large models, but at the same time, 4 times in the scene
- ▶ 34:57 Gabriele Corso Um, now, you know, models like AlphaVault II and AlphaVault III are, you know, still very large models, but at the same time, 2 times in the scene
- ▶ 38:44 unnamed speaker Yeah, so you, you know, uh, AlphaFold II, really cool. 2 times in the scene
- ▶ 38:47 unnamed speaker AlphaFold III, really cool. 5 times in the scene
- ▶ 39:32 Gabriele Corso And now, uh, both, you know, we were in the field and, you know, building on top of models like AlphaFold, and so now we no longer had, you know, kind of the base starting point, uh, to build on top.
- ▶ 43:39 unnamed speaker Both one, how did that compare to alpha fold three? 5 times in the scene
- ▶ 47:32 Gabriele Corso And so, you know, the example of, of, uh, doctrine was, you know, we, um, published this initial, uh, model called DiffDoc, um, in my first year of PhD, which was sort of like, you know, one of the early, um, models to try to predict, uh,…
- ▶ 54:08 Jeremy Wohlwend Between AlphaFold III coming out and the end of the PhD, like, the, um, number of people that would, like, reach out just for, like, us to, like, run AlphaFold III for them, you know, or things like that, just because, like, um, or both in… 2 times in the scene
- ▶ 58:24 Jeremy Wohlwend I think like some of the interesting ones, like, I mean, we had, you know, this one individual who like wrote like a complex GPU kernel, you know, for part of the architecture, um, on, on a piece of, the funny thing is like that piece of…
- ▶ 1:07:09 Gabriele Corso having these two supervision signals, you know, one discrete, one continuous, that somewhat, you know, don't interact well together, we sort of, like, build kind of, like, an encoding of, you know, sequences and structures that allows us…
- ▶ 1:28:51 unnamed speaker Because, you know, as you were saying earlier, like, where AlphaFold-style models are really good at, let's say, monomeric, you know, proteins where you have, you know, coevolution data.
🔬 From Red Teaming GPT-4 to Automating Drug Discovery: The Future of AI in Science — Andrew White
- ▶ 0:58 Andrew White And when AlphaFold came out, and it's like, you can do it in Google CoLab, you know, or on a GPU or desktop, it was so mind-blowing.
- ▶ 43:07 RJ Haneke What about somewhere like the machine learning stuff like AlphaFold and, and. 3 times in the scene
World Models & General Intuition: Khosla's largest bet since LLMs & OpenAI
- ▶ 51:42 Pim de Witte Um, as I mentioned, also, the lab is named after the- Demis de Sabeza code from AlphaFold, which is, wouldn't it be amazing if we could mimic the intuition of these gamers, who are, by the way, only amateur biologists, um, on his path to,… 2 times in the scene
Priscilla Chan and Mark Zuckerberg: Frontier AI + Virtual Biology To Solve All Diseases
- ▶ 15:11 Priscilla Chan Similarly, if you look at AlphaFold, they, they, they built off publicly available data that was collected for 30 years prior, right?
- ▶ 23:59 Mark Zuckerberg So, I mean, you mentioned, um, the work that DeepMind did on AlphaFold, which is great.
The State of AI Startups in 2024 [LS Live @ NeurIPS]
- ▶ 7:44 Pranav Reddy We're lucky to work with the folks at Chai Discovery, um, who just released Chai One, which is open source model that outperforms Alpha Fold Three.