Before AI & Biology Take an Interest in You: A Curated Thread on AI and Biology

🔬🎯 "Before AI & Biology Take an Interest in You..." A thread that every curious mind should bookmark: 10 down, 90 to go, from AlphaFold to explainable drug discovery and DeepMind's greatest hits.

AI and Biology

I guess it might be cliché to paraphrase a well known saying about politics, but here goes "take an interest in AI and Biology before AI and Biology takes an interest in you!"

Superb Resource for Continuous Learning

So far I have gotten through ten papers that I selected at random. They are top picks by the author of this thread. Major props on creating this evolving resource. Peeps should bookmark it and revisit it during those train rides or airport waits.

The papers span everything from explainable drug discovery to single-cell omics and DeepMind's AlphaFold enabled ligand design.

Get Familiar with the DeepMind Legacy

A short side note for the curious: DeepMind Technologies Limited, trading as Google DeepMind or simply DeepMind, is the British–American artificial intelligence research laboratory which serves as a subsidiary of Alphabet Inc. the parent company of Google, who own several other companies: YouTube, Waze, DoubleClick, Nest, Looker, Fitbit, and Mandiant.

DeepMind's most well-known programs and initiatives are all worth a look for enthusiasts and many have accompanying fascinating documentaries such as AlphaGo by Netflix, released in 2017, directed by: Greg Kohs and follows the journey of DeepMind's AlphaGo AI as it challenges and ultimately defeats Lee Sedol, one of the greatest Go players of all time. This is considered a pivotal moment in AI history. Their other highlights include:

• AlphaGo – First AI to defeat a world champion at the game of Go.
• AlphaZero – General-purpose reinforcement learning agent that mastered chess, shogi, and Go from scratch.
• AlphaFold – AI system that predicts 3D protein structures with high accuracy.
• AlphaStar – AI that achieved Grandmaster level in the real-time strategy game StarCraft II.
• Gato – A single generalist agent capable of performing hundreds of diverse tasks across modalities.
• Gemini – Multimodal large language model family positioned as DeepMind's flagship response to GPT-style models.
• MuZero – Model-based RL agent that learns game rules purely from observations, mastering Atari, Go, chess, and shogi.
• IMPALA – Scalable distributed agent for learning from massive parallel environments.
• WaveNet – Deep generative model of raw audio waveforms, used for high-fidelity speech synthesis.
• Sonnet – DeepMind's flexible neural network library built on TensorFlow.
• Sparrow – Dialogue agent designed for safe and useful question answering with real-time search grounding.
• Perceiver/Perceiver IO – General-purpose architecture for processing arbitrary input modalities (images, audio, video, etc.).
• Chinchilla – Efficient large language model trained with optimal compute-to-data scaling.
• Flaxformer – DeepMind's open-source library for transformer architectures, built on JAX.
• DM Control Suite – Benchmark suite for continuous control using physics-based simulations.
• Graph Networks – Framework for relational reasoning using neural message passing on graphs.
• BigGAN – High-fidelity generative adversarial network for image synthesis.
• DNC (Differentiable Neural Computer) – Neural network with external memory, enabling algorithmic reasoning.
• RLHF (Reinforcement Learning from Human Feedback) – Used in various DeepMind agents to align behavior with human intent.
• PonderNet is a model that learns when to stop thinking with adaptive computation time in neural nets.

Curation of the AI and Biology Resource

The curation has a balance between foundational theory and practical impact. I intend to keep working through the rest. Already though this is a standout resource for anyone interested in AI and Biology.