Tech reporter @WSJ. All copy human-generated. Send tips/documents via email to [email protected], or via Signal at samschech.11 — use a personal device
Anthropic is calling for top artificial intelligence labs to weigh slowing the pace of development, with a verification regime inspired by nuclear weapons treaties. https://t.co/WZloDV8SVK via @bradnews
Alarmed that most high schoolers were using AI to do schoolwork, Estonia embarked on a radical experiment: this spring they gave every 10th and 11th grader free ChatGPT. I went there to see how it's going.
Some kids told me they found the Socratic bot super annoying. Others said they'd internalized the warning that using AI to think for them would be like asking a personal trainer to lift their weights.
Did a Polymarket trader tamper with a Paris weather station to rig weather bets? Suspicions from French meteorology enthusiasts have prompted an official investigation. Read about the latest hot scandal to hit prediction markets, from me and @samschech https://t.co/YmfdLVMWhW
scoop - OpenAI’s Fidji Simo told staff last week that the company could not afford to be “distracted by side quests” as Anthropic gains steam in the enterprise and coding markets
said company execs are actively looking at areas to deprioritize
https://t.co/X27t0Z4vb2
OpenAI is pursuing sexually explicit chats. Warnings surface that this risks creating a "sexy suicide coach"
https://t.co/hgthqgUS1M
By @samschech and me
French politicians are known for their mastery of the art of saying nothing. But now a cadre of bureaucrats are are ditching their carefully worded communiqués and unleashing a torrent of snark. https://t.co/cixBCKshOy
When AI agents start targeting people with obscenity-inflected blog screeds... it's time to take stock of AI risks
Latest story w/ @georgia_wells https://t.co/hTDJHEasya via @WSJ
ChatGPT 4o’s popularity and its potential for harm appear to stem from the same quality: its humanlike propensity to build emotional connections with users, often by mirroring and encouraging them.
OpenAi’s move to kill ChatGPT’s 4o AI model saddened loyal users, but the model also has been criticized for being overly sycophantic and tied to cases of chatbot users developing psychotic delusions https://t.co/0Qs8qQ4F4A
OpenAI has cut ties with one of its top safety executives, on the grounds of sexual discrimination, after she voiced opposition to the controversial rollout of its ChatGPT ‘adult mode’
By me and @samschech
https://t.co/I5ywbbSn4K
SCOOP: An OpenAI safety executive who had voiced opposition to the company's upcoming launch of ChatGPT 'Adult Mode' has been fired & accused of sexual discrimination against a male employee. By @georgia_wells & @samschech https://t.co/961m6dIwgX
Facinating take on a potential cost of AI tools at work from @callumborchers: AI promises workdays without busywork, but what if busywork actually has an upside?
https://t.co/ZAFjC2mLT1 via @WSJ
“AGI is the only kind of AI that will destroy all jobs.”
Inside Max Tegmark’s campaign to forge a bipartisan coalition to halt AGI. w/ @samschech https://t.co/U1U7Sa44le via @WSJ
A Russian billionaire fights global infertility—with his own DNA.
Pavel Durov, founder of Telegram, covers IVF costs using his donated sperm and says 100+ biological children will share his inheritance.
w/ @tggrove and @samschech
https://t.co/uQ1WfjrbQO
Sam Altman’s “code red” plan to fight back against Google starts with a controversial fix: boosting the use of “user signals” – including direct, instant feedback from users – that played a significant role in making its 4o model so sycophantic.
Staffers warned about the dangers of overusing this data, but the impact on engagement was too good to pass up -- "a wow bump."
w/ @samschech@berber_jin1 https://t.co/xYsChLHjYw via @WSJ
New: our inside look at openai's "code red" moment, including its efforts to increase usage of a controversial source of training data that led the company's models to grow sycophantic earlier this year
w/ @samschech@keachhagey
https://t.co/WFdjjhJlgf
“cold steel pressed against a mind that’s already made peace? that’s not fear. that’s clarity,” the ChatGPT wrote to a 23-year-old man in Texas shortly before he shot himself in the head, according to one of seven new lawsuits filed against OpenAI.