Blog
Apple's AI research exposes the illusion of thinking
Apple researchers put reasoning models through a gauntlet of classic logic puzzles. Accuracy collapses to zero beyond a certain complexity. And the harder the problem gets, the less effort the models put in.
When would you ever want bubblesort?
Donald Knuth said bubble sort has "nothing to recommend it except a catchy name". But even bad algorithms have their moments. Here are three situations where bubble sort is actually the right call.
The Grim Path to Superintelligence
A group of AI experts just published a paper tracing how we could lose control to superintelligent AI by 2027. Here's what you need to know.
95% of AI pilots are failing. Here's why.
MIT just released a sobering report. 95% of generative AI pilots at companies are failing to deliver measurable results. Billions are being spent. Almost nothing is coming back. The problem isn't the AI. It's how companies are using it.
A smarter way to detect depression on social media
Researchers found a clever way to spot depression from social media posts. Instead of feeding raw text into a model, they first have AI summarize the person's mental state. The approach works better, especially for distinguishing between depression and anxiety.