Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
protagonist_hn
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
4 ms
·
1.
▲
LLM Watermarking from Scratch (Gist)
(gist.github.com)
1 points
by
protagonist_hn
2mo ago
|
1 comments
2.
▲
by
protagonist_hn
2mo ago
a tiny implementation of SynthID-style watermarking with Tournament Sampling
3.
▲
by
protagonist_hn
7mo ago
And if people don’t have understanding they won’t even know the questions to ask
4.
▲
by
protagonist_hn
7mo ago
Unless RL unleashed models come up with their own primitives. But even then if we don’t have anyone left who understands the fundamentals we are up sh*t creek anyway
5.
▲
by
protagonist_hn
7mo ago
Yep mine too. They’re like that meme of the chubby dog vs the units from back in the day
6.
▲
Swamp Dirt and AI Safety
(safeenough.substack.com)
1 points
by
protagonist_hn
8mo ago
|
0 comments
7.
▲
Rounding up
(joshs.bearblog.dev)
3 points
by
protagonist_hn
1y ago
|
1 comments
8.
▲
Exploration, Exploitation, and Thinking
(joshs.bearblog.dev)
3 points
by
protagonist_hn
1y ago
|
0 comments
9.
▲
Being good isn't enough
(joshs.bearblog.dev)
207 points
by
protagonist_hn
1y ago
|
79 comments
10.
▲
Exploration, Exploitation, and Thinking
(joshs.bearblog.dev)
1 points
by
protagonist_hn
1y ago
|
0 comments