AI ALIGNMENT FORUM
AF

Sam Brown
Ω6110
Message
Dialogue
Subscribe

Posts

Sorted by New

Wikitag Contributions

Comments

Sorted by
Newest
Risks from Learned Optimization: Introduction
Sam Brown3y10

This might be unwelcome nit-picking, but I find it kind of jarring to read "meta is Greek for above, mesa is Greek for below." That's not quite right, μετα is more like 'after' in "turn right after the bridge" and μεσα is more like 'within' (μεσο is like 'middle', as in 'mesoscale'). Above/below could be something like άνω/κάτω (like anode/cathode).

I think the meta/mesa has nice symmetry, and the name is now well-known, but maybe this particular sentence could be made less wrong :p

Also the bibliography link #7 for "What is the opposite of meta?" seems broken for me.

Reply
No wikitag contributions to display.
5Auto-Enhance: Developing a meta-benchmark to measure LLM agents’ ability to improve other agents
1y
0
35[Paper] AI Sandbagging: Language Models can Strategically Underperform on Evaluations
1y
0
18Tall Tales at Different Scales: Evaluating Scaling Trends For Deception In Language Models
2y
0