Latest AI News

Many people have called what we’re living through an epistemological crisis, and the most famous of them is probably Barack Obama, who told the Atlantic in 2020, “If we do not have the capacity to distinguish what’s true from what’s false, then by definition the marketplace of ideas doesn’t work. And by definition our democracy doesn’t work. We are entering into an epistemological crisis.” Maybe Obama is overstating things a tad when it comes to the implie

Abdulrahman Alfozan seemed to be something of a rising star in Silicon Valley. Identified as one of the few Saudi Arabian nationals working in Silicon Valley and perhaps the only one at OpenAI, he reportedly shmoozed with top Saudi officials and was even recently set to give a speech at Stanford, according to the Financial Times. In an interview, he had identified Elon Musk as his role model, and it seemed like he could have the chops to chart a similar pa

The trade secrets battle between OpenAI and Apple just took another turn. In a new federal court filing, OpenAI accused a former employee of transferring more than 100 confidential files related to an unreleased device to their personal accounts, according to Reuters. The employee later allegedly presented information about the product to senior Apple executives. The allegations stem from OpenAI’s ongoing investigation into a lawsuit Apple filed against th

The artificial intelligence industry is growing its influence in the public sphere as the technology increasingly enters every facet of life, from the workplace and academia to entertainment, healthcare, and even grocery shopping. But women are being left out of key decisions regarding artificial intelligence development and deployment, Harvard researchers argue in a newly published report. In a survey of more than 5,000 American adults, researchers found

Life, it seems, wants to play. Anyone who’s ever spent ten minutes with a five-year-old knows this perfectly well: very young children are masters at make-believe. And it’s not just people—play comes instinctively across the animal kingdom. Rats are fond of rough-and-tumble play, especially after periods of isolation. Octopuses will toy with Legos in much the same way that a newborn will play with blocks. Far from being a useless pastime, play appears to b

If there is one thing this moment of techno-negativity—from the rejection of Flock Security cameras to the distrust of AI companies gobbling up data—tells us, it’s that people hate feeling watched. You could put the brightest minds of our generation in a room and task them with coming up with the worst possible idea to meet this moment, and they wouldn’t be able to come up with a better idea than the startup Outlit, which promises an AI agent over every pe

AI companies keep pushing their models as useful tools that can boost productivity at work or even serve as personal assistants that book appointments or shop on behalf of users. But it seems like every week there’s a new story of AI screwing something up. This time, the producers of the Swedish version of MTV’s reality competition show The Challenge are dealing with a major technical blunder that some sources are blaming on AI. Swedish newspaper Aftonblad

AI giant Anthropic announced earlier this year its decision to drop a splashy initial public offering, and multiple reports have the widely anticipated market debut scheduled for before Thanksgiving. But according to a series of reports this week, the IPO, which would be the first market debut of a frontier lab purely focused on AI, is sparking less excitement and more worry than previously anticipated. First on the list of unsettled parties is the market.

OpenAI has been expanding its advertising business, recently introducing visual promos in addition to text ads that get injected into ChatGPT conversations. But how does OpenAI choose what ads to display? The decision is more personal than you might think. 404 Media ran two advertising campaigns in ChatGPT and found it allows advertisers to target users based on personally linked accounts and information, including a person’s email address, phone number, o

U.S. political ads have officially entered their AI slop era. Political campaigns and groups have spent more than $137 million airing ads created at least partly with AI tools ahead of the 2026 midterm elections, according to new data from the Wesleyan Media Project. That’s a massive jump from the 2024 election cycle. “AI is making its way into the 2026 midterms in a big way this year. For comparison, we identified less than $50,000 in spending on AI-gener

Some U.S. banks are, thankfully, blocking their clients from letting Meta’s Muse AI bot root around in their accounts. Even more thankfully, it appears relatively few people have been dumb enough to try. Speaking to Bloomberg TV on Friday, Charles Schwab CEO Rick Wurster said the brokerage is actively blocking Muse from moving money because it poses unacceptable risks to the firm and its clients. “We don’t want someone just giving their password to Muse, a

People’s trust in AI is taking a dangerous new turn. Researchers are saying that AI-generated photos and summaries are successfully distorting people’s own memories. MIT scientist Pattie Maes spoke to the Washington Post about her team’s research. Participants viewed a series of photographs, and after a break, some of the participants were shown the exact same photographs while others were shown versions altered by AI. When facing a memory test later, the

Media Library in the default skin, with Cyrillic in tags displaying without issuesClassic on the outside, modern on the insideI tried to make everything that was in Winamp 2.95 look and work just as it did there. I keep screenshots of the original next to the code and compare them against it. Anything new either works unobtrusively within the existing interface, such as support for modern formats and switching the output device, or goes into a separate gro

Version 2026.3 ships with Air alongside the IDE: connect agents, assign tasks, monitor sessions, and review changes. I’ve already tried out Air, and I really like the underlying logic: the agent performs the task, while the IDE helps you monitor its work and check the result. And thanks to ACP for making it possible to choose an agent independently of the editor. The protocol is a legend in its own right.The Zed team introduced the Agent Client Protocol on

Pretraining: 31 billion tokens2.2 continues the pretraining of base version 2.1: I added another approximately 19 billion tokens to its 11.8 billion, for a total of roughly 31 billion. Mixture:Wikipedia in three languages: 23%;CulturaX (web) in three languages: 13%;Python code: 16%, web code: 5%;Cosmopedia and Zyda-2: 5% each;poetry 4%, facts 5%, translations 5%;TAG26@@synthetic text-processing data (spelling, word analysis, CAPS): 4%;TAG30@@synthetic code

It contains everything you need to start the working day: which rule applies, what is worth remembering, what is overdue, what needs review, and why a particular course of action was chosen over another. There is no need to explain any of it again.I’ll say something separately about the top section. It contains established project rules. You can store rules discovered through work there and link to the entries from which they emerged. For example, since th