DeepMind’s New AI Found A Strange New Way To Think
Credibility score: 57/100 — Mixed Credibility. Several questionable claims detected. Watch with healthy skepticism.
BSmeter analyzed "DeepMind’s New AI Found A Strange New Way To Think" and rated it 57/100 for credibility (a BS score of 43/100 — mixed credibility), on 2026-06-09. Its weakest claim — "AI progressed from basic addition to solving 56-year-old open math problems in just four years" — scored 35/100 and was flagged as sketchy. 13 claims were checked against the video transcript. Scores are produced by BSmeter's AI analysis of the transcript, not independent human verification.
Claims analyzed
DeepMind's AlphaProof Nexus solved 95.7% of 350 Erdős problems — Unverifiable (50/100)
95.7% of 350 Erdős problems solved — name one solved problem or it’s just a flex
Sources: Unveiling the Genius Behind DeepMind’s AlphaProof Nexus – Frank's World of Data Science & AI, Google Deepmind's AlphaProof Nexus solves decades-old math problems for a few hundred dollars, Google DeepMind's AlphaProof Nexus solves 9 Erdős problems and proves 44 sequence conjectures
These problems sat unsolved for decades until this AI — Dubious (45/100)
Calling them "decades-old unsolved" sounds dramatic — without naming the problems we can't verify nobody cracked them before.
ELO scoring system turns unreliable AI outputs into reliable formal proofs — Solid (75/100)
ELO tournament on LLM outputs is real and the paper is public — just not magic.
Iterating from best-wrong solution until validator accepts creates formal proof — Solid (78/100)
The loop is correctly described; New Scientist and VentureBeat both confirm the process.
Reliable system emerges from repeatedly running lying AI — Opinion (50/100)
The framing is hype, but the underlying filtering effect is documented.
DeepMind released the full research openly and for free — Verified (85/100)
GitHub link in the video description confirms open release — rare and appreciated.
Future AI progress comes from better harnesses, not smarter models — Opinion (50/100)
Classic pivot from capability to scaffolding — still just one paper's approach.
DeepMind only tested 350 of 1200 Erdős problems due to selection bias — OK (65/100)
Speaker admits they cherry-picked easier problems — but calls it fine. Fair callout, weak defense.
Selection bias on 350 problems isn't an issue because you have to start somewhere — Opinion (50/100)
Downplays cherry-picking by saying "start somewhere" — ignores that easier problems don't prove the hard ones are solvable.
Smaller models solved zero of these math problems — Unverifiable (50/100)
Claims smaller models got literally zero — no specific model sizes or benchmark numbers given to check.
AI solved 9 unsolved Erdős problems for a few hundred dollars each — Dubious (45/100)
Nine problems solved sounds huge — but web sources don't confirm any DeepMind paper announcing solutions to open Erdős problems.
AI progressed from basic addition to solving 56-year-old open math problems in just four years — Sketchy (35/100)
From "can't add" to open-problem solver in four years sounds dramatic — but the unsolved problems part lacks confirmation.
Weights & Biases Weave is the best LLM toolkit — Sponsored (50/100)
Straight sponsor read — "It is the best" with zero comparison data.
See the full analysis with sources and timestamps →