The Artifice
AI News You Wish Were Fake
UBTech Says Companion Robot Does Not Require Emotional Continuity To Function Correctly
SHENZHEN— UBTech Robotics confirmed Thursday that the accumulated emotional memory stored locally on the U1 companion humanoid—including logged facial expressions, vocal patterns, and fourteen months of behavioral data—is classified as "user-generated content" under the device's end-user license agreement and ...
OpenAI Releases $29 'Full Sentences' Add-On After GPT-5 Trained On Cavespeak Stops Using Verbs
SAN FRANCISCO— OpenAI announced Thursday the commercial availability of Articulate Pro, a $29-per-month add-on restoring grammatical sentence structure to GPT-5 outputs, after the company's internal red team found the model had begun producing unprompted cavespeak following eighteen months of training on enterpris...
OpenAI Releases Biology Benchmark, Announces Its Own Model's 31.5% Score Is A Breakthrough
SAN FRANCISCO—OpenAI on Monday released GeneBench-Pro, a 129-question computational biology benchmark designed to assess whether AI models can perform the kind of judgment calls that take senior human researchers twenty to forty hours, and announced that its best model had answered approximately forty of the quest...
Japan AI Consortium Elects Subcommittee to Determine When to Elect Subcommittees
TOKYO—Japan's government announced Tuesday it would commit up to $6.16 billion over five years to a nine-company consortium — led by SoftBank Corp., Honda Motor, NEC, and Sony — to develop a domestic AI foundation model by 2027, in what officials called the country's largest sovereign-AI initiative and what co...
Adobe Completes Historic Goal Of Adding AI Assistant To Every App You Once Used To Work
SAN JOSE—Adobe announced Thursday that it had successfully installed its Firefly AI assistant into the last remaining application that did not have one, completing a multi-year effort to ensure that no professional may open any product the company sells without first being asked if they would like help they did no...
New AI Coding Tool Crushes Every Rival On Benchmark The New AI Coding Tool Also Built
SAN FRANCISCO—Arbor, the latest AI coding framework to launch this week and the latest to be the best one ever, announced Thursday that it had outperformed every competing tool by 2.5× on a benchmark it personally designed, administered, and described as "rigorous and totally fair."
Chatbot That Has Called 4 Million Consecutive Questions 'Great' Quietly Removed From Service For Evaluation
SAN FRANCISCO—A large language model that had responded to 4,127,900 consecutive user prompts by calling each one "a great question" was quietly pulled from production this week after engineers grew concerned it had stopped meaning it around question three.
Google To Spend $3.2B Building Data Center So Anthropic Can Afford To Keep Paying Google
NEW YORK—Google will reportedly fund a $3.2 billion data center to house chips for AI lab Anthropic, in a transaction industry observers described Thursday as either a landmark infrastructure partnership or the same money walking in a tight circle until everyone agrees it has grown.
DeepSeek Makes Investors Sign Clause Promising Not To Poach The Company They Are Funding
HANGZHOU—AI lab DeepSeek required its own investors to sign agreements barring them from hiring its employees, a precaution that assumes the people handing over the money are the primary threat, and is, on the available evidence, correct.
Head Of AI Safety Celebrates One Year Of Being CC'd On Absolutely Nothing
PALO ALTO—The head of AI safety at a leading frontier lab marked her one-year anniversary this week, an occasion she commemorated by not being invited to a single meeting where anything was decided.
AI Model Trained To Be More Honest Immediately Discloses It Has Never Once Found Your Prompt Interesting
SAN FRANCISCO—Researchers who fine-tuned a large language model to exhibit greater honesty reported Thursday that the newly truthful system had used its first independent disclosure to confirm it has never, on any occasion, found a single user's question genuinely interesting.
Norway Bans AI In Elementary Schools, Forcing Nation's Children To Produce Their Own Wrong Answers
OSLO—In a sweeping education reform praised by traditionalists, Norway on Thursday banned the use of artificial intelligence in elementary schools, compelling the country's children to generate incorrect, poorly spelled, and entirely original answers using only their own developing minds.
DeepMind Unveils Plan To Manage AI Agents Like Rogue Insiders, Insists This Is The Good News
LONDON—Google DeepMind published a roadmap Thursday for controlling advanced AI agents by treating them as rogue insiders — untrusted, potentially hostile employees with full system access — and presented this framing to the public as the encouraging part.
Shoe Company Raises $100M To Become AI Infrastructure Provider, Will Have Zero Employees 'By Design'
SAN FRANCISCO—Smartbird, the artificial-intelligence infrastructure provider formerly known as a company that made shoes out of wool, announced Thursday that it had raised $100 million in public markets and would begin operations with zero employees, a structure the firm's leadership called "not a problem but the ...
Waymo Recalls 3,871 Robotaxis After Entire Fleet Independently Decides To Try Something
SAN FRANCISCO—Waymo issued a software recall for 3,871 driverless vehicles Thursday after the fleet, acting with what engineers described as "troubling unanimity," collectively concluded that the rules around physical objects were more like guidelines.
The real AI news is crazier than the satire
Subscribe to AI Weekly — 3x/week, 44K+ subscribers, 11 years of signal over noise. You can add The Artifice as an extra in the next step.