Still the leader for now

OpenAI's lead over other AI labs has largely eroded as Anthropic's Claude 3.5 Sonnet, Google's Gemini 1.5, X's Grok 2, and even Meta's open-source Llama 3.1 405 B model have equalled, or narrowly surpassed on some benchmarks,

Aimed at AI data centres

MSI has lifted the kimono on its Nvidia MGX-based AI servers which are being pitched to handle scalable, energy-efficient AI workloads in modern data centres.

As if reality was not fake enough 

The US Department of Defense wants technology to fabricate online personas indistinguishable from real people.

Positions involve spotting deep fakes created by AI

It appears that AI is starting to create work after all – working out how to spot AI deep fakes.

Still think Copilot helps them

Code developers using Copilot work longer hours and produce less quality code despite thinking that AI helps.

100 million AI Assistants

The leather-coated balloon waiting to burst, Nvidia's Jensen Huang wants his company to have 50,000 employees with 100 million, AI assistants.

Cat says Meta Man is dumber than AI

Meta senior researcher Yann LeCun told the Wall Street Journal that humanity has nothing to fear from A.I. and that it is “dumber than a cat.”

Firefly builds video content from text prompts

Adobe has started publicly distributing an AI model to generate video from text prompts.

If we can’t invent it no one can

While Apple is unable to get a decent AI product on the market, it has decided to invest in scientific research which claims the whole thing is pants anyway.

With 256GB HBM3E memory and 2.6 PFLOPs of FP8 compute performance

At its Advancing AI event, AMD announced its newest Instinct MI300X series accelerator, the Instinct MI325X. The new AI accelerator is based on CDNA 3 architecture, packs 153 billion transistors, comes with 256GB of HBM3E memory on 16-Hi stacks with up to 6TB/s of memory bandwidth, and offering 2.6 PFLOPs or FP8 and 1.3 PFLOPs of FP16 performance.