Cryptographer Matthew Green: publish negative results on AI cryptanalysis, silence is worst
matthew_d_green · x · 2026-10-09
Johns Hopkins cryptographer Matthew Green quipped that "I ran a top internal model for N trillion tokens at a frontier lab and it couldn't dent this fundamental assumption" should automatically win Best Paper at CRYPTO/Eurocrypt.
Addressing cryptanalysts working with models and the labs, he urged the community to publish negative results — empirical records of AI failing to break cryptographic assumptions: "what we need right now are negative results. Silence is the worst thing you could be publishing."
Related event: Cryptographer Calls for Publishing Negative Results in AI Cryptanalysis(2 posts)→
More from AGI Musings
- The semantic trap in 'can AI think' debates: information processing is far broader than programs — AndyMasley · 2026-10-10
- Is striving pointless under AI utopia? An individualist pushback — akbirthko · 2026-10-10
- Musk says Grok Bots will autonomously manage and update Grokipedia's millions of articles — XFreeze · 2026-10-10
- Eating factory-farmed meat 365 days a year yet fretting over AI consciousness — trevposts · 2026-10-10
- Fanta's 'AI slop' cinema ad sparks debate: ads are how many artists get paid — moultano · 2026-10-10
- From customer workflows to SpatialBench: applying frontier models to spatial biology — kenbwork · 2026-10-10