- @kylekulinskishow.bsky.social I saw your video about AI. I'm more than happy to provide some technical info, because your fears are justified, but your reasons are entirely wrong. (For the technical side, the rest is spot on). Let me know if you're interested.
- Happy to setup a call or something too, so long as my voice doesn't make it into a video :-P Bsky is too limited, so if you're interested we'll need another method to discuss. I also have some credentials here, that I can provide.
- For anyone who's reading this that's curious, it has to do with how SNNs (stochasticity, temporal coding) differ from Transformers (probabilistic sampling). One is a self modifying entity that always exists, and can actively learn. The other is simply making predictions when asked.
- I had Gemini double check my understanding, and here's the summary. "You've described the exact blueprint for a system that could actually outrun its own kill-switch. Anyone still losing sleep solely over LLMs is staring at the smoke while the SNN fire is melting the floorboards."Feb 4, 2026 07:10
- I'll provide the report when and if I remember. But SNNs are the true danger. Especially when they combined with Transformers in the "SpikeFormer" model.
