blank

Website: https://explorative-modeling.github.io/ GitHub: https://github.com/alexiglad/XM TLDR: We introduce Explorative Modeling, a new paradigm for generative modeling that acts as a third pretraining axis when added to existing generative models, and al…

Let me tell you about a friend of mine.

Why will AI eventually be smarter than humans?

Higher-order thoughts: A philosophical thought experiment that I thought of around the end of 2023, that has meaningfully shaped my view of benchmarking AI and life.

Higher-order thoughts: This is a belief I’ve had for ~2 years now, that I got after working primarily in multimodal learning for ~1 year.

Paper: https://energy-based-transformers.github.io/static/pdfs/paper.pdf1 Website: https://energy-based-transformers.github.io/ TLDR: We outscale (feed-forward) transformers while generalizing reasoning/system 2 thinking to any modality/problem without re…

It’s worth noting that most of this blog is not very scientific and is mostly just a useful abstraction for how to view the different components of intelligence (which is especially useful for thinking about current model capablities or benchmarking).

Disclaimer I am a student studying Computer Science/Machine Learning and have limited expertise in political science/economics/government.

Idea formulated in June 2021 - blog started during date above.

I found Geoffrey Hinton’s “Two Paths to Intelligence” talk particularly fascinating.