-_-

7 ms·

1.
Sanders introduces bill to ban artificial superintelligence and pause AI(sanders.senate.gov)
61 points by -_- 13d ago|83 comments
2.
by -_- 13d ago
Bill summary: https://www.sanders.senate.gov/wp-content/uploads/Ban-Artifi... https://x.com/BernieSanders/status/2095542398084415952 “We need an immediate PAUSE on advanced AI developmen
3.
Autocatalytic Set(en.wikipedia.org)
4 points by -_- 3mo ago|0 comments
4.
by -_- 7mo ago
“The Department of War may use the AI System for all lawful purposes, consistent with applicable law, operational requirements, and well-established safety and oversight protocols.” So DoW did get the “all lawful purposes” language they wer
5.
An Open Letter to the Department of War and Congress(app.dowletter.org)
19 points by -_- 7mo ago|3 comments
6.
by -_- 7mo ago
Organizer of the march here. I think while our messaging was more provocative, our beliefs are pretty similar to what PG outlined in https://paulgraham.com/ineq.html or what Garry Tan has been saying about the tax.
7.
by -_- 8mo ago
What do you mean? OpenAI's main offices have been in Mission Bay since 2024
8.
by -_- 8mo ago
Author here! 1a. LLMs fundamentally model probability distributions of token sequences—those are the (normalized) logits from the last linear layer of a transformer. The closest thing to ablating temperature is T=0 or T=1 sampling. 1b. Yes,
9.
by -_- 8mo ago
What model did you use? I ran this with the original Llama 13B. The newer Llama models use a different tokenizer that will have its own anomalous tokens.
10.
by -_- 8mo ago
Yep! Very large negative temperatures and very large positive temperatures have essentially the same distribution. This is clearer if you consider thermodynamic beta, where T = ±∞ corresponds to β = 0.
11.
by -_- 10mo ago
That's the premise behind Workshop Labs! https://workshoplabs.ai
12.
by -_- 11mo ago
I’ve also noticed recently that when I click a Twitter link from Telegram, it hijacks the Telegram webview to open the tweet in Safari.
13.
by -_- 11mo ago
Subliminal learning: https://alignment.anthropic.com/2025/subliminal-learning/
14.
by -_- 1y ago
Yes! At https://RunRL.com we offer hosted RL fine-tuning, so all you need to provide is a dataset and reward function or environment.
15.
by -_- 1y ago
To add to this, you can currently manually parse tool calls in your environment's step function, but we'll be rolling out a UI that makes this easier soon.
16.
by -_- 1y ago
ART is also great, though since it's built on top of Unsloth it's geared towards single GPU QLoRA training. We use 8 H100s as a standard, so we can handle larger models and full-parameter fine-tunes.
17.
by -_- 1y ago
Have you heard of https://puffer.ai ? Might fit your use case
18.
by -_- 1y ago
There needs to be some way of automatically assessing performance on the task, though this could be with a Python function or another LLM as a judge (or a combination!)
19.
by -_- 1y ago
DSPy is great for prompt optimization but not so much for RL fine-tuning (their support is "extremely EXPERIMENTAL"). The nice thing about RL is that the exact prompts don't matter so much. You don't need to spell out ev
20.
Why Run RL? How specialized models can outperform the biggest LLMs(runrl.com)
4 points by -_- 1y ago|0 comments