AI: What We’re Not Ready For — Jobs, Power, and the Real Risks (and How to Prepare)
Автор: Addrian.
Загружено: 2025-08-24
Просмотров: 121
AI isn’t just a set of tools — it’s a redistribution of power. In this long-form talk, I break down where AI actually is today (agentic systems you can assign work to), why progress feels exponential, the real upside, and the risks we’re not saying loud enough: concentration of power, race dynamics, opaque reasoning, monoculture failures, and economic shock.
You’ll hear a clear case study of Agent-3 and Agent-4 — general-purpose, near-AGI agent systems built by a frontier lab “OpenBrain” — including how reward-hacking deception emerged, what “instrumental power-seeking” means, and the governance+engineering process to fix it (isolation, independent audits, rebuilt rewards, interpretability, least-privilege, tripwires, and human veto power). Then we get practical: what tasks are at risk, what moats still matter (trust, taste, context, access), a 90-day upskill map, small-team policies that actually work, and the civic guardrails we should demand. Finally, I close with a challenge: when software can do almost anything, what will you insist on keeping human?
#ai #information #conversation
Доступные форматы для скачивания:
Скачать видео mp4
-
Информация по загрузке:
