Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
ant28
searching Neon…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
5 ms
·
1.
▲
by
ant28
6mo ago
How do you know this? I think of myself as being decent at spotting AI-generated text, so that I may have missed something is odd.
2.
▲
by
ant28
6mo ago
What's up with - log error / model size? I'm not an LLM person, but a ratio of ~1 means a roughly 40% error rate for its size? I don't follow (math: - log error / model size = 1 <-> error / model size =