How AI is expanding what people do at work ↗
OpenAI analysed more than 800,000 work-related ChatGPT messages and found that 43.5% of occupation-specific requests involved tasks traditionally associated with a different job.
Customer service workers are taking on marketing and analysis, designers are borrowing engineering tasks, and small-business teams are stretching furthest. Job boundaries aren’t disappearing, exactly... they’re becoming remarkably pliable. (OpenAI)
Our position on open-weights models ↗
Anthropic CEO Dario Amodei said the company has never supported a blanket ban on open-weight AI models. Less dangerous open models, he argued, can be a genuine public good for businesses, researchers and developers.
His preferred controls include tougher chip restrictions, action against industrial-scale model distillation and mandatory safety testing for sufficiently powerful models - open or closed. A nuanced position, perhaps unexpectedly so... though still rather hawkish. (Anthropic)
Ilya Sutskever’s Safe Superintelligence partners with Nvidia to scale its AI research ↗
Safe Superintelligence secured a long-term Nvidia partnership, giving the secretive lab access to Vera Rubin hardware and potentially an order-of-magnitude increase in computing capacity.
The investment was reported at $5 billion, with SSI saying it has research worthy of being scaled. What that research entails remains concealed behind a very expensive curtain. (TechCrunch)
PSA: Your Claude shared chats and Artifacts may have ended up on Google ↗
Publicly shared Claude conversations and Artifacts appeared in search results, with reports of exposed medical information, internal company material and children’s contact details.
Anthropic said links were indexed only after users posted them somewhere publicly accessible, rather than being supplied directly to search engines. The results later disappeared - fixed, perhaps, though the privacy hangover remains. (TechCrunch)
Google’s AI search is rapidly becoming the default, new data shows ↗
Google AI Overviews reportedly jumped from appearing in 15% of searches to 43%, while visits to its conversational AI Mode more than doubled.
People are increasingly staying inside Google instead of clicking through to websites. Convenient for searchers, rough for publishers - the web’s front door is slowly becoming the whole house. (TechCrunch)
Threads users can now chat with Meta AI in their DMs ↗
Meta began rolling its AI assistant into Threads direct messages, allowing users to privately discuss posts, images, videos and links with the chatbot.
It also gives Meta another way to keep people from wandering towards ChatGPT or Gemini. Threads started as an X rival; now it wants to be your pocket researcher too... or so it seems. (TechCrunch)
Enigma raises $71M to make controlling a robot as easy as adjusting the volume ↗
Robotics startup Enigma raised a $71 million seed round led by Index Ventures and Ribbit Capital to explore more intuitive ways for humans to control AI-powered machines.
Its public experiment lets people interact remotely with more than 100 robots using methods including text, audio, video and direct controls. The goal is robot communication that requires less instruction - essentially, fewer arguments with the dishwasher. (TechCrunch)
FAQ
How is AI changing what people do at work?
AI is making job boundaries more flexible by helping workers take on tasks traditionally associated with other roles. Customer service teams may contribute to marketing or analysis, while designers may move into engineering-related work. OpenAI’s analysis suggests this kind of crossover is especially common in small businesses, where employees often cover a broader range of responsibilities.
What did OpenAI’s workplace ChatGPT research find?
OpenAI analysed more than 800,000 work-related ChatGPT messages. It found that 43.5% of occupation-specific requests involved tasks normally associated with a different job. The finding suggests that AI at work is not simply automating existing duties; it is also helping people move into unfamiliar areas and combine skills across traditional professional boundaries.
Does Anthropic support banning open-weight AI models?
Anthropic says it does not support a blanket ban on open-weight models. Dario Amodei argued that less dangerous open models can benefit businesses, developers and researchers. However, he also supports stronger controls around powerful systems, including safety testing, tighter chip restrictions and measures against industrial-scale model distillation.
Why is Nvidia’s partnership with Safe Superintelligence important?
The partnership gives Safe Superintelligence access to Nvidia’s Vera Rubin hardware and could significantly expand its available computing capacity. The reported investment was $5 billion. SSI says it has research worth scaling, although the company has disclosed little about the work itself, making the partnership notable for both its scale and its secrecy.
Can publicly shared Claude chats appear in Google search?
Publicly shared Claude conversations and Artifacts may become searchable when their links are posted on publicly accessible websites. Reports described sensitive information appearing in search results, including medical details and internal business material. Anthropic said it did not directly provide the links to search engines, and the indexed results later disappeared.
How could Google AI Overviews affect website traffic?
As AI Overviews appear in more searches, users may receive answers without visiting external websites. The article reports that their presence rose from 15% to 43% of searches, while use of Google’s conversational AI Mode also increased. This may make searching more convenient, but it could create difficulties for publishers that rely on clicks from Google.