“There may be an LLM… there'll be a policy network… there'll be some symbolic components, meaning human written and human understandable components for processing and evaluating plans”
Prof G Markets · Cal Newport: The AI Industry Is Betting On The Wrong Kind Of AI
Play from 41:13
41:13
Audio streams from the publisher’s own feed. Verbatim does not host or copy it.
What was described
Proposed a modular agent architecture that combines an LLM, a policy network, and human-written symbolic components for planning and safety
Said nearby on the same subject
“…that's why Shopify have been so successful keeping their thing as a monolith… because they spend a lot of time building their monolithic application as a modular domain-driven architecture”
“One of the hottest companies coming out of Y Combinator is AI agent evaluation… to simulate all of those conversations, those edge cases, put them through and ensure consistency.”
“One of the big things that I teach... a series of coding patterns... you can ask the AI agent, how do you propose that we build this? ... you tell it the patterns it’s allowed to use”