Demis Hassabis clarifies that while DeepMind is exploring powerful AI systems, the goal is not to trigger an uncontrolled intelligence explosion. Instead, his team is experimenting with combining evolutionary programming and foundation models to explore how self-improving systems might evolve—carefully and responsibly.

chat, are we cooked?

If you're not learning AI in 2025, you're falling behind. Join 1,000,000+ early adopters reading and learn AI in just 5 minutes a day (for free).

“Are you trying to create an explosion?” “No! Well… not an uncontrolled one…”

sounds like covid: let's create a deadly virus to see how we'd control if, were it to randomly appear

🤔 are these alphazero self-improvement loops a rare thing?

I feel like just now we are seeing the alphazero style self-play and the "general intelligence" of LLMs begin to merge. I spoke to two ex-Googlers about this in an interview on the @svicpodcast, here's the relevant parts: (the paper I mention is the "Absolute Zero Reasoner")

thanks for providing value Wes

Release the Kraken!

Self-improvement loop with MCP Agentic AI.

A new foundational approach to self-improvement is needed—recursive, gradient-aligned, and culturally transmissible. Explore the drift:

I am so on it :-)