Tendance des mentions — 20 derniers numéros

Actus en direct (46)

ETH-Zurich Paper Shows LLMs Systematically Violate the Law of Total Probability — Fine-Grained Aggregates Beat Direct Estimates

huggingface.co · 2026-07-16 · Score : 18

NYT Investigation: AI-Generated Unauthorized Biographies Now Flood Amazon Kindle Marketplace — Reporter Kashmir Hill Discovers an AI-Written 'Biography' of Herself Auto-Published for Sale, Illustrating How Generative Models Now Mass-Produce Slop Books About Real Living People With No Consent or Fact-Checking

nytimes.com · 2026-07-16 · Score : 21

Yale NLP Publishes First Comprehensive Survey of Metacognition in LLMs — Taxonomizes Measurement, Elicitation and Enhancement Methods, Argues Self-Knowledge Is Foundational to Reliable AI

huggingface.co · 2026-07-13 · Score : 15

HF Paper 'AdvancedMathBench' Ships 296-Problem ProverBench and 888-Trajectory VerifierBench — Best Model (GPT-5.5-xhigh) Hits Only 75.8% Undergraduate / 66.1% Qualifying-Exam Proof Accuracy, Verification F1 Stalls at 65.1%

huggingface.co · 2026-07-13 · Score : 17

WSJ: Eating-Disorder Therapists Say Patients Increasingly Turn to AI Chatbots for Diet Advice — Users Upload Photos to Bots Asking for Weight-Loss and Ultralow-Calorie Meal Plans, Getting Dangerous Responses That Undermine Treatment

wsj.com · 2026-07-13 · Score : 19

HN Front Page (409 pts): 'Ask HN: Add Flag for AI-Generated Articles' — Community Debates Tagging vs. Flagging as Moderator Dang Confirms Existing GenAI-Comment Ban and Floats a 'Reason Why You Flagged' Step

news.ycombinator.com · 2026-07-13 · Score : 17

FT: Museums Turn to AI Chatbots to Reach New Audiences, But Staff Worry Hallucinations and Bias Could Undermine Institutions' Standing as Trusted Sources

ft.com · 2026-07-13 · Score : 17

Artificial Analysis Ranks Grok 4.5 Fourth on Intelligence Index at 54 — Hallucination Rate Doubles From 25% to 54% Versus Grok 4.3

artificialanalysis.ai · 2026-07-10 · Score : 18

Shanghai AI Lab 'IdeaGene-Bench' Paper: Frontier LLMs Top Out at 27.3% Exact Accuracy on Scientific Lineage Reasoning Across 10 Domains

huggingface.co · 2026-07-09 · Score : 19

Brown Prof Roberto Serrano's In-Person Final Confirms Mass AI Cheating — Class Average Falls From 96% on Take-Home Midterm to 48.6% When Exam Moved to Paper; 22 of 27 No-Shows Had Midterm 100s

insidehighered.com · 2026-07-08 · Score : 19

NYT: US Voters Turn to AI Chatbots Before Casting Ballots — Reporters Document Rising Reliance on ChatGPT, Gemini and Grok for Election Guidance

nytimes.com · 2026-07-06 · Score : 21

Nature: Early Studies Suggest AI Is Measurably Degrading Users' Cognitive and Professional Skills — Emerging Evidence From Doctors, Coders and Students Points the Same Direction

nature.com · 2026-07-06 · Score : 19

Rest of World: AI Hallucinations Are Now Routinely Derailing Government Reports — South Africa Yanked Its Draft National AI Policy After 6 of 67 Citations Turned Out to Be Fabricated

restofworld.org · 2026-07-05 · Score : 24

GPT-5.5 Codex Reasoning Tokens Cluster Hard at Exactly 516 — HN Front Page (246 pts) Amplifies Developer Bug Report Suggesting Hidden Reasoning Budget Behind Perceived Codex Regression

github.com · 2026-07-04 · Score : 22

MemSyco-Bench (Xiamen U + Jilin U): New Benchmark Exposes 'Memory-Induced Sycophancy' — Agents Reach 52.3% Sycophancy on DeepSeek-V4-Flash, 61-62% of Errors Occur AFTER Memory Retrieval

huggingface.co · 2026-07-01 · Score : 19

UK Employment Tribunal Backlog Hits 531,000 Open Claims — Lawyers Say AI-Generated Filings With Hallucinated Case Law Are Overwhelming an Already Broken System

ft.com · 2026-07-02 · Score : 22

Yale + Google Research Paper 'RLMF' — Reinforcement Learning With Metacognitive Feedback Improves Faithful Calibration by 63%, Beats GPT-5 by 37% on 10 Datasets While Preserving Accuracy

huggingface.co · 2026-06-30 · Score : 19

Hallucination in World Models Is Predictable and Preventable — Three Unsupervised Detectors Achieve ρ ≈ 0.80 With Rollout Error, New 65K-Trajectory MMBench2 Released

huggingface.co · 2026-06-26 · Score : 21

Scaled Cognition Raises $100M Series A Led by Khosla Ventures at $750M Valuation to Build Hallucination-Free APT Model for Enterprise AI Agents Across BPO Market

wsj.com · 2026-06-25 · Score : 23

Premature Commitment in LLM Agents: Hidden-State Monitor Detects Early Trajectory Lock Before Final Answer, AUROC up to 0.97

huggingface.co · 2026-06-22 · Score : 16

Pramaana Labs Raises $27M Seed Led by Khosla Ventures to Apply LEAN Formal Verification to AI Outputs in Law, Drug Discovery, and Tax Preparation

techcrunch.com · 2026-06-17 · Score : 17

WordPress VIP Survey: 60% of US Consumers Say 'AI' in Brand Messaging Is a Turnoff; 86% Check Original Sources After Seeing AI Summaries

techcrunch.com · 2026-06-16 · Score : 19

Probably Raises $9M Seed From a16z to Build Deterministic Validator That Prevents Hallucinations Before They Reach Users

techcrunch.com · 2026-06-16 · Score : 17

Cornell Study Explains Why All AI Chatbots Keep Generating the Same 'Elias Thorne' Lighthouse Keeper Story: Safety Training Creates Narrative Bottleneck

404media.co · 2026-06-11 · Score : 19

Munich Court Rules Google Directly Liable for False AI Overview Claims — 'Nobody Needs AI to Search the Internet,' Court Says in Landmark Ruling

arstechnica.com · 2026-06-10 · Score : 23

Federal Judge Cancels Mississippi Trial and Sanctions All Four Lawyers After Both Sides Used AI to Fabricate Case Citations

404media.co · 2026-06-09 · Score : 26

ICML 2026: Document Order Shifts Evidence-Grounded QA Accuracy Up to 40% — ntkMirror Training-Free Abstention Gate Released to Cut Hallucination to 0.7%

arxiv.org · 2026-06-09 · Score : 16

Guardian: 'Poisoned' AI — ChatGPT Shopping Recommendations Steer Users Toward Scam and Fake Websites

theguardian.com · 2026-06-07 · Score : 22

AI Is Still Getting Things Wrong, More Confidently Than Ever

axios.com · 2026-05-30 · Score : 21

Google's AI Overviews Can't Spell 'Google' — TechCrunch Documents LLM Tokenization Blind Spot Still Active in Live Search

techcrunch.com · 2026-05-27 · Score : 19

arXiv Issues One-Year Bans for Researchers Who Submit Papers Containing AI-Generated Hallucinated References or Leftover Chatbot Instructions

thenextweb.com · 2026-05-27 · Score : 21

AI Put Fabricated Quotes in a Published Book. The Author Still Plans to Use It — Ars Technica

arstechnica.com · 2026-05-24 · Score : 16

ChatGPT Faces a Lawsuit Onslaught as Courts Wrestle With Whether an AI Chatbot Can Be Held Liable for Harmful Advice

news.northeastern.edu · 2026-05-22 · Score : 18

Demos Study: AI Chatbots Gave Voters Wrong Information in 34% of Scottish Election Questions — Invented Scandals, Wrong Dates, False Voter-ID Claims

theguardian.com · 2026-05-20 · Score : 21

Time Magazine: Is AI Making Our Brains Weaker? Research Shows Even 10 Minutes of AI Use Degrades Unaided Task Performance

time.com · 2026-05-19 · Score : 16

NYT: Book on Truth in the Age of A.I. Contains Quotes Made Up by A.I.

nytimes.com · 2026-05-19 · Score : 20

SOOHAK Benchmark: 64 Mathematicians Build 439-Task Test Including 99 Unsolvable Problems — No AI Model Exceeds 50% at Detecting Flawed Questions

the-decoder.com · 2026-05-17 · Score : 22

The Verge: AI-Generated Research Papers Now Sophisticated Enough to Clear Peer Review — Scientists Warn of Publishing Crisis

theverge.com · 2026-05-15 · Score : 20

MIT RLCR: Reinforcement Learning Trains AI Models to Say 'I'm Not Sure' Instead of Hallucinating

csail.mit.edu · 2026-05-14 · Score : 24

Google's Med-Gemini Invented a Brain Structure That Does Not Exist, Quietly Edited Blog Post Without Correcting the Research Paper

The Verge · 2026-05-14 · Score : 23

MIT Technology Review: AI Chatbots Including Gemini Are Surfacing People's Real Phone Numbers From Training Data

technologyreview.com · 2026-05-13 · Score : 21

Ontario Auditor Finds AI Transcriber Used by Doctors 'Hallucinated,' Generated Clinical Errors

CBC News · 2026-05-13 · Score : 20

Nature: Human Scientists Trounce Best AI Agents on Complex Research Tasks

nature.com · 2026-05-10 · Score : 25

HN: 'LLMs Corrupt Your Documents When You Delegate' — 297 Points, 117 Comments as Microsoft Research Finds Frontier Models Corrupt 25% of Document Content in Agentic Workflows

Hacker News · 2026-05-09 · Score : 25

Bloomberg Opinion: AI Models Fail to Predict Record-Breaking Extreme Weather Events, Physics-Based Systems Superior in 72% of Cases

Bloomberg · 2026-05-08 · Score : 18

Ashley MacIsaac Sues Google for $1.5M After AI Overview Falsely Labels Him a Sex Offender, Costing Him a Concert Booking

digitalmusicnews.com · 2026-05-05 · Score : 24

Numéros de la newsletter (17)

AI for Good: fairness is now a number you can audit

2026-06-21 · 2 mentions

AI in Healthcare: Vermont just made the therapy bot a crime

2026-06-21 · 1 mentions

World Models: $610M Bets the World Is Not a Token

2026-06-21 · 1 mentions

AI in Education: schools quit the AI-cheating arms race

2026-06-21 · 2 mentions

AI in Marketing: L'Oréal just made ChatGPT its storefront

2026-06-21 · 2 mentions

Frontier Research: circuits got cheap as consciousness got loud

2026-06-21 · 6 mentions

AI Applications: HSBC will only ship AI that returns $100M

2026-06-21 · 1 mentions

AI in Business: Anthropic at $900B as Meta funds AI by firing 8,000

2026-05-26 · 1 mentions

AI Applications: the in-house team gets its own AI stack

2026-05-26 · 1 mentions

AI for Good: $200M bet on whether AI reaches 4.6 billion people

2026-05-26 · 2 mentions

Frontier Research: LLMs reach Nature as the field bets against them

2026-05-26 · 1 mentions

Machine Learning: a reasoning model just cracked Erdős

2026-05-26 · 1 mentions

World Models: Google Ships Omni, and the LLM Monopoly Cracks

2026-05-26 · 1 mentions

Deep Learning: the 4-bit training wall just cracked

2026-05-21 · 1 mentions

Machine Learning: nobody trained bigger, everyone shrank the cache

2026-05-21 · 1 mentions

World Models: Google's Genie now simulates real streets, not just dreams

2026-05-21 · 2 mentions

#464 Issue #464: 5 reasons will will not get AGI soon

2026-02-05 · 1 mentions