Totally normal behaviour from GLM-5.1

The headline on that was, if you can believe it, courtesy of an editor called Steve

Thank you for arranging that connection, Mark! (How was the talk?)

What is Mythos and why are experts worried about Anthropic’s AI model The company says Mythos is too dangerous to release publicly. Cybersecurity experts agree the model's capabilities matter, but not all of them are buying the most alarming claims

The release of Mythos scared an awful lot of people - but now the dust has settled a little, how worried should we be? @sciam.bsky.social asked me to take the temperature of the experts www.scientificamerican.com/article/what...

AI isn’t built for all languages and cultures. There’s a push to fix that From Egypt to Indonesia, developers are building their own models to better reflect local languages and cultures.

One corollary of the current local LLM drive is that it puts the power to train and finetune new models into the hands of people outside big tech giants. The result? More representation of global majority cultures and language. My latest for @fastcompany.com www.fastcompany.com/91527477/the...

AI Is Finding More Bugs Than Open-Source Teams Can Fight Off Anthropic’s Mythos and similar AI tools can identify threats and vulnerabilities faster than small teams can fix them, putting the internet at risk.

12 years ago I wrote how the internet is being protected by two guys named Steve. Today, the situation isn't all that different, except those handful of volunteers are being swamped by AI systems finding holes in our digital systems. My essay for @bloomberg.com www.bloomberg.com/news/article...

What is Mythos and why are experts worried about Anthropic’s AI model The company says Mythos is too dangerous to release publicly. Cybersecurity experts agree the model's capabilities matter, but not all of them are buying the most alarming claims

The release of Mythos scared an awful lot of people - but now the dust has settled a little, how worried should we be? @sciam.bsky.social asked me to take the temperature of the experts www.scientificamerican.com/article/what...

Potential new profile pic courtesy of the Tekniska Museum in Stockholm (which is also very flattering on age)

AI Is Finding More Bugs Than Open-Source Teams Can Fight Off Anthropic’s Mythos and similar AI tools can identify threats and vulnerabilities faster than small teams can fix them, putting the internet at risk.

12 years ago I wrote how the internet is being protected by two guys named Steve. Today, the situation isn't all that different, except those handful of volunteers are being swamped by AI systems finding holes in our digital systems. My essay for @bloomberg.com www.bloomberg.com/news/article...

AI isn’t built for all languages and cultures. There’s a push to fix that From Egypt to Indonesia, developers are building their own models to better reflect local languages and cultures.

One corollary of the current local LLM drive is that it puts the power to train and finetune new models into the hands of people outside big tech giants. The result? More representation of global majority cultures and language. My latest for @fastcompany.com www.fastcompany.com/91527477/the...

One big step… for future missions: what did we learn from Artemis II? The moon mission didn’t break scientific ground — but that wasn’t the point, says Chris Stokel-Walker

It was more than one small step for man, but did the Artemis II mission actually represent a leap for mankind? The folks at The London Standard asked me to investigate www.standard.co.uk/news/science...

@markriedl.bsky.social Hello! Have sent you a DM

What will the world look like when driverless cars start to hit the road? My latest for IMechE www.imeche.org/news/news-ar...

Wind turbines' best strength - their ability to withstand any issues - is their biggest liability, making them a hassle to recycle. My latest for IMechE www.imeche.org/news/news-ar...

A bit twiddling my thumbs today, so if you're an editor looking for a tech or general features specialist and need stories done, lmk

Talking to journalists A common view around me seems to be that journalists are frequently dishonorable and dangerous, and talking to them is a risk to be avoided unless you have a very specific piece of information that yo...

Thank goodness for @katjagrace.bsky.social's piece here dispelling a lot of myths about journalists (TL;DR: we are also humans too) worldspiritsockpuppet.substack.com/p/talking-to...

Another story and another name attached to the mysterious identity of Satoshi Nakamoto - but why does it matter? My story for @the-independent.com www.independent.co.uk/independentp...

Inside the rise of unmanned ‘dark factories’ When Henry Ford first automated production of the Model T in 1913, the goal was efficiency through rhythm.

More than a century ago, Henry Ford began to automated production of his vehicles. But today, dark factories operate without humans entirely. My latest for IMechE's Professional Engineering www.imeche.org/news/news-ar...

AI companies are tightening their token allowance for end users - but the last one to blink may win. My latest for @fastcompany.com www.fastcompany.com/91524498/the...

This 'generous' donation is 0.0286% of Anthropic's run rate revenue (source: www.anthropic.com/news/anthrop...) I can't believe that 12 years on from profiling the literally two people tasked with maintaining then fixing the Heartbleed bug we still have to talk about this.

thanks for the share!

Very lucky to sit down with Paul Newman at Oxa about the impact embodied AI will have on engineering for IMechE's Professional Engineering magazine www.imeche.org/news/news-ar...

Scientists invented a fake disease. AI told people it was real Bixonimania doesn’t exist except in a clutch of obviously bogus academic papers. So why did AI chatbots warn people about this fictional illness?

🚨 For nearly two years - and even now, depending on the day and model - AI models could tell you all about bixonimania, a totally made-up disease in a paper funded by “the Professor Sideshow Bob Foundation". My @nature.com story on AI's fallibility in medicine www.nature.com/articles/d41...

"Anthropic has also been in ongoing discussions with US government officials about Claude Mythos Preview and its offensive and defensive cyber capabilities." 👀https://www.anthropic.com/glasswing

Scientists invented a fake disease. AI told people it was real Bixonimania doesn’t exist except in a clutch of obviously bogus academic papers. So why did AI chatbots warn people about this fictional illness?

This wild story takes in shout-outs to the cast of Friends, a fake academic whose name translates roughly to "Lying man", and an entire paper being retracted by the ultimate parent company of the outlet that published the piece - so give it a read www.nature.com/articles/d41...

Storing hydrogen is a major problem that we need to solve to address the world's energy issues. For IMechE, I explored one solution - salt caverns www.imeche.org/news/news-ar...

Scientists invented a fake disease. AI told people it was real Bixonimania doesn’t exist except in a clutch of obviously bogus academic papers. So why did AI chatbots warn people about this fictional illness?

Fake papers that were obviously AI-generated (but still picked up and referenced in peer-reviewed journals) and a couple of Medium blog posts were enough to get AI to confidently explain something was wrong with people that absolutely wasn't www.nature.com/articles/d41...

Scientists invented a fake disease. AI told people it was real Bixonimania doesn’t exist except in a clutch of obviously bogus academic papers. So why did AI chatbots warn people about this fictional illness?

🚨 For nearly two years - and even now, depending on the day and model - AI models could tell you all about bixonimania, a totally made-up disease in a paper funded by “the Professor Sideshow Bob Foundation". My @nature.com story on AI's fallibility in medicine www.nature.com/articles/d41...

Lawmakers want to restrict 3D printing to stop ghost guns. Critics say it won’t work New 'print blocker' proposals would force printers to scan and reject files, a move opponents say risks privacy and targets the wrong problem.

America's obsession with thinking about Second Amendment rights (while being unwilling to talk honestly about gun control) means you're possibly going to be unable to use 3D printers anymore without handing over data. Really. My latest for @fastcompany.com www.fastcompany.com/91522293/law...

Lawmakers want to restrict 3D printing to stop ghost guns. Critics say it won’t work New 'print blocker' proposals would force printers to scan and reject files, a move opponents say risks privacy and targets the wrong problem.

America's obsession with thinking about Second Amendment rights (while being unwilling to talk honestly about gun control) means you're possibly going to be unable to use 3D printers anymore without handing over data. Really. My latest for @fastcompany.com www.fastcompany.com/91522293/law...