Machine Learning Street Talk (MLST)

Subbarao Kambhampati - Do o1 models search?

28 snips
Jan 23, 2025
In this engaging discussion, Professor Subbarao Kambhampati, an expert in AI reasoning systems, dives into OpenAI's O1 model. He explains how it employs reinforcement learning akin to AlphaGo and introduces the concept of 'fractal intelligence,' where models exhibit unpredictable performance. The conversation contrasts single-model approaches with hybrid systems like Google’s, and addresses the balance between AI as an intelligence amplifier versus an autonomous decision-maker, shedding light on the computational costs associated with advanced reasoning systems.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
ANECDOTE

O1's Costly Experiments

  • Early O1 experiments quickly incurred high costs due to the numerous unseen reasoning tokens.
  • Subbarao needed special permission from his university to get reimbursed.
INSIGHT

Defining Reasoning

  • Reasoning should be defined by formal, sound patterns, not by what humans do.
  • LLMs creating connections without guarantees is not sound reasoning.
ANECDOTE

Monty Python Logic

  • A Monty Python sketch illustrates flawed reasoning by connecting random things to prove someone a witch.
  • This highlights how connections alone do not constitute reasoning.
Get the Snipd Podcast app to discover more snips from this episode
Get the app