On the Impossible Safety of Large AI Models · HackerPapers discussions

On the Impossible Safety of Large AI Models — mentioned by westurner

Mon, 28 Sep 2026 15:47:48 GMT

What if there were a zero trust sandbox with costed opcodes? Then would there be alignment in AI systems as compared with non-AI systems? Even with a perfect sandbox, an agent can slip in vulnerable code intentionally or accidentally. How different of a problem is that from hiring and HR is that? Idk. Perhaps easier to terminate a model. "On the Impossible Safety of Large AI Models" (2022) https://arxiv.org/abs/2209.15259 "LLMs + Security = Trouble" (2026) https://arxiv.org/abs/2602.08422

Paper and discussion history. A mention is not an endorsement.

On the Impossible Safety of Large AI Models — HN submission

Sat, 15 Jul 2023 11:26:24 GMT

On the Impossible Safety of Large AI Models

3 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

On the Impossible Safety of Large AI Models — HN submission

Sun, 14 May 2023 19:13:31 GMT

On the Impossible Safety of Large AI Models

4 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

On the Impossible Safety of Large AI Models — mentioned by uhryks

Tue, 01 Nov 2022 12:40:20 GMT

> This video debunks 4 dangerously misleading scientific papers, all with a Google co-author, that denigrate the urgent need for massive audits and investments in security and ethics. Terrifyingly, the scientific community has scientifically validated and massively propagated what appears to be intentional disinformation. Script and sources: https://docs.google.com/document/d/1MZinEbqdCyvssBWHsX2Qjrv9... French researcher and youtuber Lê Nguyên Hoang tries in this video essay to demonstrate that the ML research community is plagued by unfounded claims of algorithms being privacy-preserving…

Paper and discussion history. A mention is not an endorsement.