Potential state control of near-superintelligent open AI if it causes harm
The user says they love open models, run a Kimi K3 fine-tune on Tinker and aren't advocating a ban. They describe state control as a prediction, not a preferred outcome.
TLDR
A user warns that freely distributing open-source AI models approaching superintelligence may cause harm, arguing that “the offense dominates the defense.” If that happens, they predict China and the US will try to exert control. In a follow-up reply, they say this isn't their preferred outcome and that they aren't trying to ban open models, while expressing hope that open-source labs act prudently.
Combined views
34K
2 Sources, first seen 18d ago