AI News – May 12
Grok 3.5 is delayed for at least a week Google is looking to take shopping to the next dimension Higsfield is blowing people’s minds and there’s a cool demo of real-time image recognition Here’s today’s AI news. Take a deep breath. There’s no Grok 3.5 yet and it’s not likely going to drop this week.
Elon Musk replied to a post about Grok 3.5 timelines by saying “3.5 is still too rough around the edges, needs another week or so.” It’s hard to decipher the true nature of the delay, but perhaps the latest Update to Gemini 2.5 Pro made it too difficult to launch a model that would top the coding charts.
There’s a lot of expectation on Grok 3.5 after Grok 3 performed so well, releasing a model that didn’t hit number one on a bunch of benchmarks would feel like a fail. Google are looking to help enhance the whole online shopping experience by making it far simpler to create 3D product visualisations.
In a new concept, sellers can add as few as three images of their product, and then Google’s generative video model Veo is used to produce an interactive 3D model for the user to explore. This will put 3D models into the hands of even the most modest of sellers without requiring expensive, specialised equipment.
With the amount Higgsfield ships, it’s honestly hard to keep up. Today, Higgsfield has released Effects Pack 3, extending their library of cinematic visual effects by a further nine. These include ‘Morph Skin’, ‘Glow Shift’ and the Creepy ‘Skin Surge’ There’s even one effect that they’ve had to censor on their promo video called ‘Head Explosion’ I guess it’s easy to guess why it had to be censored.
And finally, an X user posted a demonstration of a camera application where the app is capable of describing what it sees in real time. It’s quite incredible to see the accuracy and speed of the translations, but even more incredible is is the fact that it’s all running locally on a MacBook M3.
The application was built using SmolVLM, a small yet efficient vision language model that’s available on Huggingface. This is another one of those moments where you really start to see the power of AI outside of its typical text prompt inputs.