🧙🏼 Elon's massive new AI cluster

Howdy, wizards.

Wanna see something strange? Google searches for AI Perfume is growing fast, according to Exploding Topics. Companies like EveryHuman, an “algorithmic perfumery” are working on AI guided scent creation.

Dario’s Picks

The most important news stories in AI this week

1. xAI launches a cluster of 100,000 GPUs. It took just 4 months to build, and Elon says it’ll even double in size soon.

• ‎ Why it matters‎ ‎ The next phase of the AI race is likely to be more impacted by physical construction (huge GPU clusters like this), rather than new, scientific discovery, and xAI is moving at lightning speed setting up the infrastructure to train frontier level AI. Nvidia is, of course, standing in the back rubbing their hands 💰

Continued after the ad… ‎

This issue is brought to you by

Superfilter: Your AI-Powered Email Assistant

Superfilter isn't just another inbox tool - it's your AI executive assistant, dedicated to managing your email. Unlike traditional email clients, Superfilter learns about your email habits and actively manages your inbox. Some of the features users really like:

• It briefs you on critical inbox activity, so you don't miss what's important.

• It tracks unfinished tasks, follow-ups, and invitations, keeping you on top of your commitments.

• It handles repetitive work like scheduling meetings and drafting responses, freeing up your time.

Stop drowning in emails. Let Superfilter handle your inbox, so you can focus on work that matters. Whether you're a founder, manager, or executive, Superfilter adapts to your needs. Reclaim your time and boost your productivity.

Let Superfilter save you time without sacrificing peace of mind today.

2. ‎Stable Diffusion 1.5 gets taken down after claims that it’s trained on child porn. Version 1.5 was the most popular image generator from Stable Diffusion (downloaded more than 6m times just in the last month), but has now been taken down from the Hugging Face AI repository. The model was allegedly trained on a dataset containing thousands of images containing child sexual abuse material (CSAM).

    

   

   Via David Evan Harris on LinkedIn

    

   Photo realistic image models like Stable Diffusion 1.5 are notorious for being trained on porn content (@Levelsio talked about this recently with Lex Fridman), which makes them fine-tuned to generate realistic human anatomy. BUT this model in particular (and probably others) were trained on an old dataset named LAION-5B, for which there is now a new, “safer” version: Re-LAION-5B.

    

   While the model is taken down from Hugging Face, clones/backups are popping up on other model hubs such as Modelscope. On Reddit, some users are backing up the model referring to its “irrevocable copyright license” for doing so.

    

   Non-profits like the Internet Watch Foundation has for some time reported that AI-generated CSAM is becoming more prevalent and severe, including identifiable victims and deepfake videos.

    

   ‎ Why it matters‎ ‎ This case highlights big ethical and legal concerns of AI models being trained on “the whole internet”. The ongoing investment into image generators by mainstream companies shows there’s a big business opportunity – with many legitimate use cases – but that probably also makes it less interesting to scrutinise the origins of these models.

From our partners

Wired's "rocket fuel of AI" label has Wall Street buzzing. Projections skyrocketing to $80 trillion, akin to 41 Amazons, signal a seismic shift. But here's the kicker: astute investors have a shot at riding the wave with a company primed for supremacy. Dive into The Motley Fool's exclusive report for your front-row seat.

Read the report now

3. Asking LLMs to “read the question again” improves their performance. A paper from earlier this year found that simply asking AI to “read the question again” in the prompt – making the LLM process the question twice before answering – enhances its understanding and the output.

• ‎ Why it matters‎ ‎ You can use this to get ChatGPT/Claude/whatever to reason a bit better on challenging problems. Like Ethan Mollick pointed out, it’s fascinating that simple tricks that teachers already use with their students ie. “read the question again” actually works for AI, too.

4. TLDR covers the most interesting tech, science and coding news in just 5 minutes. TLDR covers the best tech, startup, and coding stories in a quick email that takes 5 minutes to read.

    

   No politics, sports, or weather – we promise. And it's read by over 1,250,000 people. Subscribe for free ‎ (sponsored)

Affiliate disclosure: To cover the cost of my email software and the time I spend writing this newsletter, I sometimes link to products and other newsletters. Please assume these are affiliate links. If you choose to subscribe to a newsletter or buy a product through any of my links then THANK YOU – it will make it possible for me to continue to do this.