For a while now, companies like OpenAI and Google have been touting advanced "reasoning" capabilities as the next big step in their latest artificial intelligence models. Now, though, a new study from six Apple engineers shows that the mathematical "reasoning" displayed by advanced large language models can be extremely brittle and unreliable in the face of seemingly trivial changes to common benchmark problems.
The fragility highlighted in these new results helps support previous research suggesting that LLMs use of probabilistic pattern matching is missing the formal understanding of underlying concepts needed for truly reliable mathematical reasoning capabilities. "Current LLMs are not capable of genuine logical reasoning," the researchers hypothesize based on these results. "Instead, they attempt to replicate the reasoning steps observed in their training data."
At the FIDO Alliance's Authenticate conference in Carlsbad, California, on Monday, researchers are announcing two projects that will make passkeys easier for organizations to offer—and easier for everyone to use. One is a new technical specification called Credential Exchange Protocol (CXP) that will make passkeys portable between digital ecosystems, a feature that users have increasingly demanded. The other is a website, called Passkey Central, where developers and system administrators can find resources like metrics and implementation guides that make it easier to add support for passkeys on existing digital platforms.
Adobe is kicking off its annual Adobe Max conference today with the launch of new AI-powered features across its Creative Cloud apps. New AI features for Photoshop, like automatic background distraction removal and a more powerful Firefly generative AI model, are the biggest announcements, with Illustrator, InDesign, and Premiere Pro also getting new features that can help to speed up traditionally labor-intensive design tasks.
Adobe is making the jump into generative AI video. The company’s Firefly Video Model, which has been teased since earlier this year, is launching today across a handful of new tools, including some right inside Premiere Pro that will allow creatives to extend footage and generate video from still images and text prompts.
The update adds the ability to flag and reject photos and apply a one- to five-star rating. Then, with filters based on flags, rejects, and star ratings, it’s easy to navigate among images to determine which to keep. The process is aided by extensive single-key shortcuts, too.
The update, available now on the App Store, adds smarter recipe import from websites, new Control Center widgets, Foodnoms AI enhancements, and more.
Clicks Keyboard for iPhone is back, and they’re pushing all the right buttons—literally. I had the chance to get hands-on with the latest model, and it’s clear that Clicks has done their homework. From ergonomics to extra features, here’s everything that stood out in my first look with their version 2 product.
Rebecca Ferguson‘s Juliette Nichols is out to unravel a web of lies about a toxic and deadly world that threatens the last people living on Earth in the trailer for season two of Apple TV+’s Silo.
A flaw in Safari’s link-sharing feature allows user-added text to look like a real quote or headline from a trusted source.
Apple is preparing to begin supporting digital car keys in the Wallet app for certain Volvo, Polestar, and Audi vehicles, based on code changes discovered by MacRumors in Apple's Wallet app backend.
I still have a whole bunch of CDs -- Mac OS X, Developer, MacAddict -- lying around that I don't know what I want to do with them.
:-)
~
Thanks for reading.