Google Deepmind's Alphafold 3 AI SHOCKS Entire Science Community

Recently, Google’s Deepmind labs unveiled the third iteration of Alphafold, Alphafold 3, and it’s actually insane how this revolutionary AI model can predict the structure and interactions of all life's molecules with unprecedented accuracy.
Today we’ll be talking about the technical details of what this model is all about, capabilites and how it’s better than the previous releases.
Now back in 2020, Google's DeepMind rocked the molecular biology world by using AI to decode how tiny proteins behave. Fast forward to today, with the latest incarnation of AlphaFold, the experts at DeepMind and their partners at Isomorphic Labs have cracked the code for all the molecules that make life tick, even our very own human DNA.
This is HUGE because just a few years back, AlphaFold was all about figuring out protein structures. But now, AlphaFold 3 takes it up a few notches by helping researchers figure out DNA, RNA, ions, modified residues, and even ligands.
According to DeepMind, this upgraded model boasts a whopping 50% improvement in prediction accuracy compared to its predecessors. DeepMind CEO Demis Hassabis, in a chat with reporters, described AlphaFold 3 as another stride forward in AI's journey to crack the secrets of biology, following the groundbreaking milestone of AlphaFold 2.
With AlphaFold 3 comes a library jam-packed with molecular structures. Wondering how it works?
Well, on a high level, researchers just input a list of molecules they're curious about mixing up, and boom, AlphaFold 3 presents them a 3D model of the fresh combo using a diffusion method.
It's the same technique used by AI image generators like Stable Diffusion but we will get into the details of how this particular diffusion architecture works later in this video.
#LLM
#largelanguagemodels
#chatgpt
#AI
#ArtificialIntelligence
#MachineLearning
#DeepLearning
#NeuralNetworks
#Robotics
#datascience

[@vallab19]

AI drug discovery enormously beneficial to human kind. In the no so far future I hope and believe that AI need not has to go though all those millions of compound to find a suitable candidate, AI will able be able to design new chemical compounds suited exactly to the intended target to cure the disease.