🔴 Bobak Soltani

A neural network, made out of proteins and built in mammalian cells. "we combined de novo–designed protein heterodimers and engineered viral proteases to implement a synthetic protein circuit that performs winner-take-all neural network classification..." Congrats @ElowitzLab.

Introducing ASAL: Automating the Search for Artificial Life with Foundation Models sakana.ai/asal/ Artificial Life (ALife) research holds key insights that can transform and accelerate progress in AI. By speeding up ALife discovery with AI, we accelerate our understanding of emergence, evolution, and intelligence–core principles that can inspire the next generation of AI systems! We proudly collaborated with MIT, OpenAI, Swiss AI Lab IDSIA, and Ken Stanley on this exciting project. Full Paper (Website): pub.sakana.ai/asal/ Full Paper (arxiv): asal.sakana.ai/paper/ Code: github.com/SakanaAI/asal/ In this work, we propose a new algorithm called Automated Search for Artificial Life (“ASAL”) to automate the discovery of artificial life using vision-language foundation models. Instead of tediously hand-designing every tiny rule of an Alife simulation, simply describe the space of simulations to search over, and ASAL will automatically discover the most interesting and open-ended artificial lifeforms! Because of the generality of foundation models, ASAL can discover new lifeforms across a diverse range of seminal ALife simulations, including Boids, Particle Life, Game of Life, Lenia, and Neural Cellular Automata. ASAL even discovered novel cellular automata rules that are more open-ended and expressive than the original Conway’s Game of Life. We believe this new paradigm may reignite ALife research by overcoming the bottleneck of manually designed simulations, thus advancing beyond the limits of human ingenuity.

ICML2024 Tutorial on "Neural Operators & Machine Learning on Function Spaces" is now out. #NeuralOperators #AInScience @nvidia @NVIDIAAI youtube.com/watch?v=_j7bce… Looking forward to building the future of this field with the bigger community, all together.

It's unbelievable how this was written in 1619-1655. Talks about Exoplanets and then consider Earth's habitable condition as a pure chance. wow "A Voyage to the Moon" by Cyrano de Bergerac, 1619-1655 gutenberg.org/files/46547/46…

Another experiment with Gaussian Painters ✨🎨 By optimizing 3D Gaussian Splattings over separate images at several viewpoints, it is possible to get a Steganography effect! Three paintings are hidden in those gaussian splats

High-dimensional asymptotics via the replica method continued. This time applied to denoising autoencoders: arxiv.org/abs/2305.11041 Kudos to Hugo! I was surprised by how few theory works on denoising autoencoders we found. If you know of some we do not cite yet, please do reply.

Well, it makes a lot of sense for sure. 😅

This paper shows t-SNE projections of data after each layer(+ some gif) which shows how each layer in NN will separate datapoints into clusters more and more. * would like to see the effect of BN, Residual connections, poolings, attention, ... on this. deepai.org/publication/pr…

Sparse Nonlinear Models for Fluid Dynamics with Machine Learning and Optimization youtube.com/watch?v=z_CZ_V…

If you know anything about Conway's Game of Life, this is so fucking impressive. oimo.io/works/life/

@rileytomasek @hubermanlab @OpenAI @mortezagk

I've learned so much from @hubermanlab, but podcasts make recall and discovery challenging. Huberman AI uses the latest models from @openai to search every episode with Google-level accuracy and summarize the results. huberman.rile.yt

We can do photosynthesis in mammalian cells why are people talking about chatbots?! nature.com/articles/s4158…

@mpezeshki91 @JeffHumpherys @tj_jarvis may i have your thoughts on this please?

@mpezeshki91 Is it just me or does this "layer-wise update/train" is somehow similar to using BP and just make a model one layer at a time? we can also freeze the previous layers to only train the last one (the newly added). this has been done before: ar5iv.labs.arxiv.org/html/1706.02480

At this year's NeurIPS, @geoffreyhinton presented the forward-forward (FF) algorithm, an alternative to backprop. I took a stab at implementing FF in PyTorch, and here it is:github.com/mohammadpz/pyt… Below is my understanding of the idea (with a grain of salt): 1/5

@mpezeshki91 by the way thanks for the code.

pretty print tensors and much more. useful and elegant 👌 github.com/xl0/lovely-ten…

/researches in the field. One can not see this pattern of concentration of computational capabilities and knowledgeable labor in bunch of companies owned by billionaires and think something good will come out of it for all!

The trend of making bigger and bigger models has interesting aspects but also projects the inequalities of the world from economical to computational and knowledge. A handful of wealthy players backed by billionaires will gain more power and influence to shape the future of/

Will #GPT4 be BIG? Yes. Will GPT-4 change the course of history? No.