Topics

A new direction for making transformers AGI-friendly—via judiciously mixing continuous symbolic program learning into open-weights transformers.

Internal confidence signals shape whether AI models answer or abstain, experiments suggest, raising questions over how developers build systems that know when to stay silent rather than merely sound fluent.

The new model uses a larger base, stays on hard tasks longer, and is sold at the same price as Grok 4.6 through the API and coding tools.

Will the scaling laws keep on scaling as AI models get bigger and bigger? AI companies are betting hundreds of billions of dollars that they will – but history suggests it is not always so simple.

I’ve been experimenting with vibe coding, math and physics with the artificial intelligence (AI) assistance from Grok. I find the topics that we're exploring very interesting, but that alone wouldn't make me dedicate too much time and effort to this project (I'm not detail-oriented). Rather, I want to test the performance of today's AI on things that I know well enough.

Legendary futurists Max More, Anders Sandberg, Natasha Vita-More, and Eliezer Yudkowsky recently met online to debate AI and possible AI futures.

By including interactions with users in training data, AI models could gain improved statistics on how people actually interact and an improved ability to resolve imprecisions and capture shifting language patterns.

In a technical vibe coding project I'm using Grok 4 for both coding and exploring the underlying math and physics. I'm impressed with the results so far.

Artificial Intelligence (AI) pioneer Geoffrey Hinton, winner of the 2024 Nobel Prize for Physics, shared his insights on AI in two recent videos published by The Royal Institution.