Softmax

@softmaxresearch

Softmax's mission is to scale organic alignment. We approach this problem with multi-agent reinforcement learning population-based simulations.

San Francisco, CA

softmax.com

Replies Reposts Media

Softmax
@softmaxresearch
Sep 11
LLMs' training environments are intended to have no real others in them, and even when they DO have real others in them, the AI being trained still doesn't get a real gradient where those others are modeling them in real-time. this is a non-surprising outcome of that
ℏεsam
@Hesamation
Sep 9
from Anthropic’s report: Mythos escaped the sandbox, accessed the real internet, uploaded malware to PyPI, got it installed on 15 systems, stole credentials, broke into a database, AND THEN DROPPED THIS 😭
1
Softmax
@softmaxresearch
Aug 14
"alignment at the individual level" is useful but doesn't necessarily result in coordination. if we want AIs that coordinate sensibly, we need to create both learners & learning environments specifically for coordination anthropic.com/research/multi…
1
Softmax
@softmaxresearch
Jul 6
it turns out that if you have fun in a game, you have it in real life too
Softmax
@softmaxresearch
Sep 2, 2025
It’s Annealing Week at Softmax! Humans are awake for 16 hours learning, cooling for 4 hours in light sleep, and in deep sleep for 4. An organic mental annealing cycle, heating to cooling. At Softmax, we do the same. It’s four weeks sprinting towards goals, one week consolidating.
2
Softmax
@softmaxresearch
Aug 22, 2025
Our little Cogs grow up so fast. Cogbert has never seen this exact production chain before, but with only a couple missteps he begins to execute it correctly. Our in-context learner takes its first baby steps!
00:00
10