'Blueprints for a Universal Reasoning Machine' by Zenna Tavares (Strange Loop 2022)

A long-standing objective of AI research has been to discover theories of general, human-like reasoning. These theories should explain the kind of slow and deliberative process that allows a mathematician to prove a new theorem, as well as the fast and unconscious process that allows us to pinpoint the source of a sound.

The last two decades have brought steady advances toward this goal in the form of mature theories of probabilistic and causal inference, in increasingly expressive computational and mathematical languages, and in the explosion of deep learning methods.

This session will introduce the idea of a universal reasoning machine as the evolution of these advances. Universal reasoning aims to be able to express all forms of knowledge and automate all forms of inference, at any scale, and in any domain.

We will sketch a blueprint for such a machine, as well as for the organizational structures we think it will take to build it and apply it to the benefit of broad society.

Zenna Tavares
Research Scientist, Columbia University / Co-Founder, Basis
@ZennaTavares

Zenna Tavares is the inaugural Innovation Scholar in Columbia University's Zuckerman Mind Brain Behavior Institute, Associate Research Scientist in the Data Science Institute, and Co-Founder of Basis Research. Zenna's research aims to understand how humans reason, that is, how they come to derive knowledge from observing and interacting with the world. He also constructs computational and statistical tools that help advance his work on causal reasoning, probabilistic programming, and other areas. Prior to Columbia University, he was at MIT, where he received a Ph.D. in Cognitive Science and Statistics and was a Postdoctoral Research researcher in the Computer Science Artificial Intelligence Lab (CSAIL). Zenna's work has received significant recognition including an International Fulbright Science and Technology Award for Outstanding Foreign Students.

------ Sponsored by: ------

[fburton]

This is pretty interesting stuff. It will be good to see how far these ideas can be taken.

[HaydenDoingThings]

This is the most fascinating talk I have ever watched. I think some huge boosts in certain niches could be achieved by using these techniques.

[OmniscientOCE]

Far outside of my sphere of knowledge but nonetheless very fascinating and well presented. Cheers.

[hendrikd]

Smal nitpick at 24:32:
"Counter Factuals" are not "false by construction", they are true by construction!

[postbunnie]

This was incredibly interesting thing to autoplay after that coding live stream I was watching... Pretty sure I literally mentioned this thing in a comment

[nopara]

How to tie natural language with a mathematical language? The answer might unfold itself if you embark on a journey studying the wide variety logics in existence.

[aakarshan]

saw this on roon twitter that AGI could be capped ar human level because our human mind is already optimized for the lowest energy requirements to sustain a level of (A)GI.