Matthew Green jokes: the most productive AI alignment task right now is launching a URL shortener
matthew_d_green · x · 2026-10-15
Johns Hopkins cryptography professor Matthew Green (@matthewdgreen) quipped: "I think the most productive task in AI alignment right now is just to launch a URL shortener."
A satirical jab at AI safety discourse — implying that the real, rampant security risk in the LLM ecosystem is prompt injection (agents blindly trust and fetch URL content), so fixing that concrete attack surface beats abstract alignment theorizing.
More from Fun
- AI-built MOBA hits 13.1K plays and 11.3K players in its first week — TAbrodi · 2026-10-15
- Zuckerberg fumbled a pint on camera and Guinness won millions in free PR doing nothing — yungcontent · 2026-10-15
- 'Lovable Is a Scam' Post vs Sarcastic Excel Rebuttal Ignites AI Builder-Tool Debate — gabriel1 · 2026-10-15
- AI engineer lists YC founders' wildest pitches: tailgating him into his apartment, hitting him with a car — jxnlco · 2026-10-15
- Musk praises Pichai's AI demo: 'the creativity in the prompting and the model is awesome' — elonmusk · 2026-10-15
- 23-year-old faked his way to a $280k Head of AI role with just a Claude subscription — Aizkmusic · 2026-10-15