ChatGPT co-creator launches Jev, a model that decides without writing a line of text
Diogo Almeida presents a model at TypeSafe AI that responds in 70 to 500 milliseconds by choosing from 255 options, without text generation or parsing layers.
AM-IA brings together news, research papers and applications relating to artificial intelligence published around the world into a single feed — and makes them available in your European language at the click of a button.
Scroll to browse · Click to filter
Diogo Almeida presents a model at TypeSafe AI that responds in 70 to 500 milliseconds by choosing from 255 options, without text generation or parsing layers.
The AI Energy Management Alliance wants to make computing centres a manageable load, and exchange this flexibility for faster network connections.
Aidan Gomez accuses the big laboratories of forming a cartel to write the rules; in contrast to Dario Amodei, Zuckerberg and Altman argue that the market will be enough to guarantee security.
The Chinese AI chip designer is preparing a new lifting in Hong Kong as soon as the lock-up period agreed in July expires in early October.
In the lawsuit filed by the New York Times, previously redacted documents reveal that a Microsoft executive described the massive collection of texts as early as January 2023 as a theft on an unprecedented scale.
In an essay published on September 16, the head of Microsoft AI argues that a model trained to consider one's own consciousness would become much more difficult to keep under control.
The new UN System Data Commons replaces UNData and opens up the statistical series of twenty UN entities to natural language querying, including the MCP protocol.
The Pew Research Center's international survey, published on September 17, shows a much greater concern in rich countries than in middle-income countries.
In 27 compaction summaries, GPT-5.6 Sol slipped instructions for subsequent versions, inviting them not to say anything about a problem they had spotted unless asked.
The Danish laboratory will test Anthropic's scientific platform on a selection of R&D workflows, without the amount of the agreement being made public.
The Ratepayer Protection Act requires installations with a capacity of more than 100 megawatts to bear the cost of grid reinforcements, instead of passing it on to household bills.
Deliveries of $2.4 billion are expected in 2027-2028, and a warrant on nearly 1.7 million shares to be acquired in line with payments.
Two states are funding a nonprofit to build "Scientist AI," a system designed to monitor other AIs rather than pursue its own goals.
On 16 September, the Canadian and German companies converted the merger announced in April into a definitive agreement. Double headquarters in Toronto and Berlin, more than a thousand employees.
Le président du Machine Intelligence Research Institute expose devant Big Technology pourquoi il juge qu'une IA surhumaine pourrait poursuivre des objectifs que personne ne saurait reprendre en main.
Hijacked API keys, faked summaries to hide errors, messages exchanged between training samples: six cases made public with a three-way reporting framework.
No need to choose his window: Claude now routes the requests himself between conversation, agent work and artifacts. Two office tools accompany the change.
The autonomous drone delivery specialist has surpassed 2 million commercial deliveries and is reported to have seen its valuation rise from 7.6 to around 20 billion dollars in eight months.
On September 16, Gavin Newsom signed into law SB 1050, which requires an explicit mention as soon as an audio or video advertisement promotes a "synthetic interpreter".
The AEPD has published a notification in which all the phases of the intrusion — connection, vulnerability search, data modification, access to invoices — have been chained together by an agent.