AlphaGo architect David Silver outlines plans for new AI start-up

In short: David Silver, known for building DeepMind’s AlphaGo, says his new start-up will focus on AI that can teach itself through experience.

What happened

David Silver, a leading researcher at Google DeepMind and one of the main architects of AlphaGo, is building a new company called Ineffable Intelligence. He announced the start-up in February and it has been valued at about $5bn, according to the Financial Times.

In the interview, Silver said he wants to move beyond today’s most common chatbots, which are trained on large amounts of human-written text from the internet. These systems are often called large language models, which means software that learns patterns in language by reading huge amounts of writing (like a student who learns by reading the whole library).

Silver’s focus is reinforcement learning. This is a method where a computer learns by trying something, seeing if it works, and then trying again (like practicing a video game level until you get better). He previously used this approach in AlphaZero, which learned games like chess and Go largely by playing against itself.

Silver also said the world is changing faster than society can adapt. He criticized the gap between progress in AI and progress in AI ethics. He said competition between companies and countries can push development forward too quickly.

Why it matters

Silver is arguing for a different path for AI, one that learns from its own “experience” rather than mainly recycling human knowledge. If it works, it could speed up research in areas like science and medicine. It could also raise new safety questions, especially if systems become hard for people to understand or control.

Source: Financial Times