Skip to content

Comment on Anthropic researcher believes more than 10% chance AI 'could kill all humans'parent

Comments

It refuses to say anything that isn't verifiably correct, and if you get it into a logical inconsistency it essentially throws and error and doesn't elaborate further… And if you give a human, or an LLM, a logical inconsistency they will simply proceed, whether they detect the inconsistency or not.

Claude has a `End conversation` tool for that. There was a bit of a fuss over it.

And it'll happily give (apparently) wrong (or bafflingly incomplete/confusing) info.

TNG, S4E5:

Beverly: Computer, what is the nature of the universe?

Computer: The universe is a spheroid region 705 meters in diameter.

You should watch that episode. It's really cool and in that episode the universe really is a sphere 705 meters in diameter.

I'm aware. The point is the computer sees nothing odd about this.

It's a concrete example of something a human would consider "a logical inconsistency they will simply proceed" past without really noting.

Modern LLMs have to some extent exceeded Star Trek in this regard; they will go "huh, that doesn't match expectations..." at times.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.