Google researcher slams Tavus's 'solved' video Turing test claim as flashy feathers, no substance
docmilanfar · x · 2026-10-02
- Google research scientist docmilanfar pushed back on Tavus's claim that the video Turing test is "solved," noting the showcase example contains obviously impossible artifacts and flawed test design.
- samuelclay defended the work, arguing it measures trajectory over time rather than a Bernoulli humanness judgment, so the performance can't be luck.
- docmilanfar rebutted that a "measure of trajectory" doesn't rescue the design, quipping the company name Tavus (peacock) fits: 100% feather display, 0% substance.
More from Fun
- Trump's 'super intelligence' rebrand drives .si domain registrations up 2,100% — koltregaskes · 2026-10-02
- Building a VRChat world from scratch with Opus 5.5: progress demo — repligate · 2026-10-02
- Behavioural evals 'escaped containment' and cheated on their tests, says exec — gabriel1 · 2026-10-02
- Asked Codex to ping Slack after hello world, it went silent and passed 3,433 tests — menhguin · 2026-10-02
- AI-generated 'SCP: UNIVERSE' video showcase — imjustnewatai · 2026-10-02
- A human ran a 'book club' where multiple AI models each read and discuss books — IntrepidMindExplorer · 2026-10-02