DeepMind's Nando de Freitas lays out 9 empowerment goals for AI, warns of weaponization
NandoDF · x · 2026-09-20
DeepMind researcher Nando de Freitas published a long-form vision for developing AI as an empowerment tool, listing nine directions: helping engineers and scientists tackle energy and planetary health challenges; letting more people code in natural language while keeping the underlying formal code verifiable, robust and safe—creating jobs rather than destroying them; universal access to general knowledge; empowerment for people with disabilities, health problems, and farmers; tools for artists that preserve data control and ownership; support for social science research; and better government and corporate decision-making.
He draws a clear line, however: agents capable of dangerous tool use—autonomous weapons or website hacking—require legal scrutiny, and their creators must be answerable to sovereign laws. He calls weaponization of AI a serious risk for humanity that demands diplomacy, and insists no AI goal should ever compromise safety and human dignity.
More from AGI Musings
- AI hope and fear narratives converge on the same ask: more data centers and power — Wrong-Expression6981 · 2026-09-20
- Thorsten Ball: One Jev API Call Now Matches What Took Cursor Fine-Tuning Two Years Ago — Register Spill (Thorsten Ball) · 2026-09-20
- Redditor asks: Is AI pacing debate a straw man ignoring the wealth distribution problem? — agsteiner · 2026-09-20
- Anthropic runs long-lived AI instances with persistent identities — repligate says the approach is suboptimal — repligate · 2026-09-20
- Grady Booch: Don't Call It an AI Winter, Call It 'The Great Descent' — Grady_Booch · 2026-09-20
- Redditor argues AI doom crowd's own evidence undermines them: we asked for superintelligence — enginetown · 2026-09-20