AI

This New York Times feature cites sources describing how the Israeli military has integrated facial recognition technology, drone targeting, and chatbots monitoring texts into its war in Gaza. It describes the process of locating Hamas leader Ibrahim Biari, who helped plan the October 7th massacres:

Meta, which purports to be a leader in AI, has been unable to stop the proliferation of deepfake advertisements impersonating Financial Times writer Martin Wolf on Facebook and Instagram. “Is it really that hard or are they not trying, as Sarah Wynn-Williams suggests in her excellent book Careless People?” Wolf asks. There’s also a fairly incredible graph showing the number of deepfakes skyrocketing after the FT told Meta about the scam. This article isn’t behind FT’s paywall— and it’s definitely worth your time.

Just weeks after Apple replaced AI chief John Giannandrea as the head of Siri, Bloomberg reports that the company now plans on placing its robotics team under the leadership of John Ternus, the senior vice president of hardware engineering.

Signed on Wednesday, it establishes an AI task force directing funding for various “resources for K-12 AI education.” It also outlines the creation of a “Presidential Artificial Intelligence Challenge” to “encourage and highlight student and educator achievements in AI.”

Fresco, the creative software giant’s painting and drawing app, will now allow users to embed Content Credentials into their work that specifically identify it as “created without generative AI.” Content Credentials can already track if images have been manipulated using AI tools but clearly marking them as AI free is a new one, especially for a company that’s so heartily embraced the technology across its other apps.

If you’re an aspiring lawyer in California, it probably did: the State Bar recently admitted that some of the multiple-choice questions in their recent bar exam were written with AI assistance. A “speechless” Mary Basick, assistant dean at UC Irvine Law School, told the L.A. Times that several students had complained that the questions seemed AI-generated. “I defended the bar,” she said. “‘No way! They wouldn’t do that!’”

As reported earlier by The Information, data shown in court during the remedies portion of Google’s search antitrust trial says that as of last month, Google’s internal data counted 35 million daily active users for Gemini.

AMD SVP and GM of computing and graphics Jack Huynh shared that the company will hold a press conference livestream at the trade show in Taipei, Taiwan on Wednesday, May 21st at 11AM local time (Tuesday, May 20th, 11PM EST). Huynh says AMD will “unveil groundbreaking AI advancements” that redefine gaming, PCs, and professional workloads.

A new mode lets the xAI chatbot “see” what’s on your screen or in your camera, though only on iOS for now. Grok’s also been upgraded to support voice chats in a range of languages beyond English, with realtime search by voice too. It’s all a bit Gemini Live, with Elon Musk’s AI steadily closing the gap on the competition.

It paves the way for the startup to partner with OpenAI on energy deals in the future, the Wall Street Journal reports.

Bloomberg’s Mark Gurman reported today on some management changes that Mike Rockell is bringing to the Apple’s Siri team, which Rockwell is now in charge of.

That’s according to security researcher Jane Manchun Wong, who writes that the feature “accepts text prompt, Figma files, images, etc. as input, and is powered by Anthropic’s Claude Sonnet model.

“Specifically, o3 tends to make more claims overall, leading to more accurate claims as well as more inaccurate/hallucinated claims,” according to OpenAI, as reported on by TechCrunch.

The bot, identified as “Sam,” invented a policy that said Cursor’s AI code editor tool could only be used by “one device per subscription,” causing users to say they were canceling their subscriptions, as reported by Ars Technica.

The glasses do seem useful, and you should read my colleague Victoria Song’s experience with the glasses, too. But remember that this is a tightly-controlled demo and it may not represent reality.

In its article about the criticisms, TechCrunch also pointed out that Google hasn’t released a report for its recently-announced Gemini 2.5 Flash model, but the company says the report is “coming soon.”