Build and characterize artificial reasoning system that is truth-seeking, persuasive, and creative.
By Zory Zhang @
Goal: introduce and ask for opinions on my long-term vision of AI2Reason.
Outline recap:
(Counterexample: give conclusion w/o observation or experiment)
Go beyond case-by-case abstraction
[When you don't show your reasoning...]
LLMs dream/hallucinate/bullshit because they care about:
We LLMs because:
Mutual benefit between "building" "modeling"
"Humans can capture concepts in so little context, mimic rules from so few examples, yet still be able to generalize to genuinely new situations."
(Concept learning)
Why hasn't been solved before but promising now?
Automated theorem proving? Again, a playground that is well-defined and established.
Verifiable mathematical formal language: a preliminary interface between humans and computers in doing proof
Join Discord right now to see what exciting things are happening! We welcome everyone interested in this direction
Thank You! Q&A time!
AI2Reason Community@Discord
https://marp.app/docs https://github.com/rnd195/my-marp-themes/blob/main/beamer.css