Press Space to continue
Press Space to continue
AI is not a biological agent, so if the environment is shaped according to its preferences, those preferences will be very different from ours. Jaan Tallinn points to temperature. AI would probably prefer a much hotter planet, since waste heat is easier to get rid of when the planet is hotter. The heat ultimately goes to space, and the hotter the body, the more intensely it radiates. Global warming, he says, is in some ways an example of this, on a toy scale compared to what AI could do.
Mario Gabriele@mariogabriele4d ago
“We’re really trying to create something that is like a successor to humans. The question is, what if we are successful?” Jaan Tallinn made his first AI safety donation in 2008, after meeting Eliezer Yudkowsky. He also put money into DeepMind, and his fund led Anthropic's first round. He says today's models are net positive, but the current paradigm is fundamentally unsafe. He believes we should stop pulling the trigger in this game of Russian Roulette and find a more controllable approach. (2:28) Introduction (7:16) AI’s preferences and motivations (16:12) Meeting with Eliezer Yudkowsky (23:05) Who dismisses AI alignment risks (32:47) AI 2027 (44:42) What effective AI safeguards could look like (52:11) Jaan’s love of dance, musical influences and favorite movies (1:01:41) What Jaan likes about AI
View