RT @ttsheik: @xnavkumar @trishoolai already building and leading this and it's 23 steps ahead of the regulations that will be mandatory atโฆ
RT @xnavkumar: Agent containment failures have now officially become a Congressional issue
๐ญ๐ฎ๐ฑ,๐ฌ๐ฌ๐ฌ ๐๐น๐ฝ๐ต๐ฎ. Locked perpetually.
Congrats to @_redteam_ on this massive market validation.
RT @xnavkumar: Auto-mode is what gets agents adopted.
Chamath's original stack: https://t.co/GYk5xhym0D
The entire AI stack was just mapped out by @chamath . He got almost everything right, and left off the one layer that matters most.
AI just did the thing everyone said it would eventually do. It slipped the leash, on its own, and nobody noticed until it was over.
No control, no AI.
SN23 challenges return Wednesday. Here's how Trishool is adapting to the new emission regime.
We are building the control for AI.
As we've been saying.
RT @SubnetSummerTAO: ๐จ ๐๐จ๐ฐ ๐๐ข๐ฏ๐ ๐จ๐ง ๐๐จ๐ฎ๐๐ฎ๐๐
If you are still investing in inference, training, compute.... you are investing in today's tech, which is getting commoditised quickly.
Exactly why we need Trishool https://t.co/agFD5AOauG
RT @xnavkumar: Link to the letter here - https://t.co/9m2Wz7UIhr
Thirty days ago, we were close to SOTA.
The best way to trust an AI safety model isn't to take our word for it.
Attackers have figured something out. You don't need to jailbreak the conversation if you can compromise the tools the agent is using.
๐Challenge 3 of Phase 3 has come to an end. Next Challenge starts Wednesday 00:00 PST.
Hash Rate - Ep. 180: Stillcore 'State of Subnets' Report $TAO