What’s changed in AlphaFold 3?

The new AlphaFold paper is here: www.nature.com/articles/s4158...
Previous AlphaFold video: • How AlphaFold solves p...
The research I referenced that OpenAI did on multimodal models is here: openai.com/index/clip/

[@videotimebaby90]

Your last videos got me curious and started looking at Alphafold and am now seriously considering making it my goal to work in that field. Thank you 🎉

[@LookingGlassUniverse]

That's awesome! I'd love to hear more if you do go down that path! My email is on my "about" page

[@aniksamiurrahman6365]

AI or Biology?

[@shadowskullG]

​@@aniksamiurrahman6365 for me it's ai, I've been working my way for the past 2 years, gained a lot of knowledge and theory , now I want to get a degree

[@Phillyosopher]

I wouldn't if I were you, there will be no jobs thanks to alphafold itself.

[@aniksamiurrahman6365]

@@Phillyosopher Its exactly the other way. More things are possible for Biotech industries now, and thus more opportunity.

[@igNights77]

Good stuff Mithuna keep it coming!

[@ItAintNecessarilySo]

Thank you so much for sharing your insights into this! You have a real gift for seeing and explaining things

[@TedToal_TedToal]

Another great video on alpha fold. Thanks!

[@insidejazzguitar8112]

Fascinating. I really learned something, which says something about your ability to teach, because this is not my wheelhouse

[@codeawareness]

We live in some very interesting times, indeed. Thank you for this video! Just a few notes, and maybe you can address them in a next video. Apparently it's not capable of working with any ligands, just a subset provided by the AI vendor. Also, it seems a ligand with many rotary bonds might be overwhelming for pair representation matrix. Still, many people think this new version will most likely help make important leaps for the drug design industry.

[@CosmicCells]

Great video, you have a great way of explaining complicated concepts!

[@Pradowpradow]

Not related to the video but I really love your handwriting

[@tafazziReadChannelDescription]

Yeah I can see how once you do away with the "triangles" or any other modeling of the relatedness of neighboring amino acids, the data about the evolutionary history of the molecule would not be as important as it was before. It's crazy that you can get atom-by-atom predictions in molecules as big as protein complexes in reasonable computational time. If I had to guess, a goal for the next decade in protein structure algorhythms would be to turn an input that's like "I want a protein that catalyzes this reaction under these conditions" and it spits out a hybrid protein sequence that does that, predicting the speed of the reaction.

[@rachel_rexxx]

Concise and understandable. Thanks!

[@tcaDNAp]

I guess this channel will always be my favorite! I just learned about pair representations on NanoRooms and now I know an application and a prediction about the future of AI tools 🤯

[@tcaDNAp]

Maybe it can fold proteins, but AI can never predict the connections that Mithuna will come up with next!

[@f.andersen3824]

I just read a Book of Manfred Spitzer about AI. AlphaFold was also mentioned but not fully explained in detail. He mentioned those triangels forming the proteins backbone. But not how they are arranged and that there be so many. 😅

[@VirtuousSaint]

that chair looks really comfortable

[@sanemmanueljames2947]

You are so awesome. Well explained

[@Micro-bit]

Great stuff - Im so excited about AF3. I think new ML techincs will boost whole science and inventions.

[@dking7985]

awesome explanation!

[@SzTz100]

You got me interested in this subject, although I don't know why.

[@tanchienhao]

what a time to be alive!

[@LookingGlassUniverse]

Two Minute Papers? Love that channel

[@tanchienhao]

@@LookingGlassUniverse yesss!!!

[@hitoshiyamauchi]

Thanks a lot! 😀

[@ThePowerLover]

The simplicity is expected, GPT-4o seems to be simpler than other GPT-4 variants, (less disintegrated) more integrated.

[@dogslife4831]

Alpha Fold 3 looking good

[@joehopfield]

Perfect overview - just the right level of background and detail. I suspect that non-drug protein / molecule design will have a much bigger impact than novel drugs only for the wealthiest few % of all humans. (Drugs aren't going to fix rapacious capitalism and inequality)

[@tafazziReadChannelDescription]

Drugs fix illnesses, not politics

[@tomholroyd7519]

"merely" astonishing

[@finalfan321]

i loved this!

[@ApprendreSansNecessite]

I feel very uncomfortable regarding AI, especially generative AI but your recent videos have changed my feelings somewhat.

[@aniksamiurrahman6365]

The only discomfort I feel is, due to these AIs, we might never discover the underlying physics. Researchers might just stop working on them.

[@ashishpatel350]

Going to be fun times in the future. At least for drug companies

[@BooleanDisorder]

Bah! I am waiting for SigmaFold 😎 /s

[@DistortedV12]

Since when does she cover AI content, what happened to the physics from her Ph.D?

[@miskee11]

Physics has been stuck for a hundred years. AI and crypto are what actually matter now.

[@aniksamiurrahman6365]

​@@miskee11Crypto? It went nowhere at least 5 yrs ago. Soon might lose it's usefulness in speculation as well.

[@MedusaSkirt]

@@miskee11 you may be interested in a video by physicist Angela Collier titled "a physicist responds: physics has done very little for like 70 years"

[@tafazziReadChannelDescription]

@@miskee11 But sir le graphene! Le superconductors at room temperstures! Le heckin nuclear fusion?! We need more billions of units of currency to make those things real (for real this time, promise)

[@miskee11]

@@tafazziReadChannelDescription The Industrial Revolution and its consequences have been a disaster for the human race.

[@larsyt80]

First?!

[@kirthankamble95]

Ok