319
Audio & Video Production299
Software Development224
Automation & Workflow197
Writing & Content Creation179
Marketing & Growth172
AI Infrastructure & MLOps140
Design & Creative147
Photography & Imaging136
Data & Analytics107
Voice & Speech121
Education & Learning118
Sales & Outreach105
Customer Support109
Research & Analysis84
David Silver has started a new AI company, Ineffable Intelligence, backed by $1.1B. It plans to build “superlearners” using trial-and-error training.
In short: David Silver, known for leading work on AlphaGo, has founded Ineffable Intelligence and raised $1.1 billion to pursue AI that learns through trial and error.
David Silver, a longtime researcher at Google DeepMind and one of the key people behind AlphaGo, has started a new company called Ineffable Intelligence. AlphaGo became famous in 2016 after it taught itself to play the board game Go at a very high level.
Silver says his new company will focus on reinforcement learning, a way of training AI by letting it learn through practice, mistakes, and feedback (like learning to ride a bike by wobbling until you get it right). This is different from the approach many AI companies use today, where systems learn mainly from huge amounts of human-written text.
In an interview with Wired, Silver argued that AI trained mostly on human text can be limited by what humans already wrote down. He used a thought experiment, if you sent a text-trained AI back to a time when most people believed the Earth was flat, it would likely repeat that belief because it cannot test reality on its own.
Ineffable Intelligence has raised $1.1 billion in seed funding and is valued at $5.1 billion, according to the report. Silver also said he plans to donate any personal money he makes from his ownership stake to charity.
This is a big bet on a different path to more capable AI. If Silver’s approach works, it could lead to AI that learns skills by doing, not just by reading. It also raises safety questions, since a system that learns on its own might find goals or strategies people do not want, so how the company tests and controls these systems will be closely watched.
Source: Wired