Can AI self-improvement overcome diminishing returns?

3 pointsposted 14 hours ago
by marojejian

2 Comments

marojejian

14 hours ago

Quality post. It boils down to a lot of trend measuring, which is not reliable. But I can't think of a better method.

I'm assuming this is the same guy who wrote the Nexus series, which was amazing BTW.

verdverm

11 hours ago

> Here’s my take: Given our best current data, the AI self-improvement loop would need to be roughly 5–10× stronger to sustain itself, let alone run away.

Sentences like this are when the back button gets exercised, it's just fantasy hype slop imo