How Intelligence Evolved | A 600 Million Year Story
Art of the Problem
Video follows evolution of intelligence from simple nerve nets to complex neural networks in humans
Made with Max Bennett after reading his book
Sponsored by Jane Street
Support AOP via Patreon
Timeline of key points in the video:
00:00 - Introduction
01:13 - Nerve nets
01:29 - Steering
02:20 - Reinforcement learning
06:23 - Mental simulation
08:50 - 3rd person simulation
11:50 - Language- The brain is made up of billions of neurons forming trillions of electrical circuits.
Brain function doesn't exist in separate regions but as an interplay of many regions.
The brain evolved to create intelligence.
A book called A Brief History of Intelligence takes an evolutionary view based on modern research.
Evolution of intelligence has gone through five key leaps in learning ability.
Before brains, there was a nerve net allowing the simplest animals to sense and react to their world.
Ancestors resembling worms with simple brains evolved to detect chemicals in their environment.
Animals couldn't learn new behaviors within their lifetimes due to genetically encoded brain circuits.
A new learning mechanism based on experience allowed animals to learn from trial and error.
The neocortex evolved, allowing animals to imagine possible futures.- Animals can imagine possible futures like mental movies
Dopamine circuits can be triggered by these simulations
Animals can learn from mental trial and error
Rats simulate going down each pathway in a maze
Primates developed theory of mind by thinking from a third person perspective
Primates can hold accumulated knowledge not based on their own experiences
Language in humans allows communication of imaginations with symbols
Writing allows thoughts to accumulate outside living minds
Language has allowed knowledge to accumulate across generations
Intelligence evolved from learning from own actions to learning from others' imagined actions
AI has exploded due to large language models like GPT which model human language by reading text
Debate in AI field on whether artificial intelligence can shortcut the steps human intelligence took or if it needs to build up.- The video discusses whether artificial intelligence can short circuit the steps that human intelligence has to take and go straight to language, or if it needs to build up from more foundational experience.
It questions what form of experience the AI needs to take and whether the brain needs the body.
Jane Street is sponsoring the video and is a quantitative trading firm that uses machine learning, distributed systems, programmable hardware, and statistics to trade on markets around the world.
Jane Street is looking for intelligent people from all backgrounds who love solving complex problems to join their team, and background in finance is not required to apply.
The video suggests that the stock market is a form of distributed intelligence that no human or computer can predict long-term, and encourages viewers to explore opportunities at Jane Street's website.