Matthew Sparkes

Matthew Sparkes

@sparkes.bsky.social

Reporter at New Scientist magazine. "Technologically optimistic but epistemically cautious," according to ChatGPT. Got a story? Email: matthew.sparkes@newscientist.com

938 followers250 following431 posts

Formalisation has been key to the explosion of AI mathematics - it allows companies to "prove" their results are correct. But AI has already learned to hack its way to fake success. Tricking humans is sometimes easier than doing real mathematics... www.newscientist.com/article/2591...

newscientist.comMathematicians and AI in behind-the-scenes battle over what's true | New ScientistAI models are solving mathematics problems with increasing pace, and a technique called formalisation is key to demonstrating that their claimed solutions are indeed correct. But can we trust the form...

Kyiv resident Olexandra Shekhireva says almost all Shaheds targeting the city are now jet-powered models. “It’s noisier, louder – not like a conventional motorcycle engine, more like a plane. They’re obviously faster,” she says. www.newscientist.com/article/2587...

newscientist.comJet-powered drones pose a new threat to Ukraine – can they be stopped? | New ScientistUkraine is racing to design faster, more capable interceptor drones to tackle Russia's new generation of jet-powered Shahed attack drones

What’s next for mathematics now that AI is upending the field? “I have colleagues who will think it’s the end of the world, and colleagues who will think it’s the best thing since sliced bread,” says one mathematician. “I’m in the ‘a bit terrified’ camp.” www.newscientist.com/article/2588...

newscientist.comWhat's next for mathematics now that AI is upending the field? | New ScientistFamous and challenging problems in mathematics are falling to AI at a rapid pace. Where does this leave mathematicians and what will the field look like in coming years?

Terence Tao says AI companies are harming mathematics by dropping new results without engaging with academia. “These companies are dumping carcasses of raw meat onto our table and saying ‘here you go, I solved your food problem’ and then they just leave." www.newscientist.com/article/2588...

newscientist.comTerence Tao: AI companies are harming mathematics | New ScientistAI companies are making new mathematical discoveries at a rapid pace but not sticking around to help unpick the new proofs. That is not the way to advance our understanding, says mathematician Terence...

I'm almost embarrassed to have done it. It took no skill. I feel it somehow undermines the work of human mathematicians and I fear it may be the thin end of the wedge for diminished human input in the field. Is this how future mathematicians will feel? www.newscientist.com/article/2587...

newscientist.comI made a free AI chatbot solve a decade-long maths problem in 13 minutes | New ScientistAfter months covering the shocking rise of AI mathematics, reporter Matthew Sparkes decided to put a free chatbot to the test – and was shocked at the speed at which it solved an outstanding problem

After months of covering AI making mathematical breakthroughs I had a little experience of my own. I set ChatGPT on an esoteric dice problem that some humans had just made a big discovery on, and it found a solution to a slightly larger version of the problem. www.newscientist.com/article/2587...

newscientist.comI made a free AI chatbot solve a decade-long maths problem in 13 minutes | New ScientistAfter months covering the shocking rise of AI mathematics, reporter Matthew Sparkes decided to put a free chatbot to the test – and was shocked at the speed at which it solved an outstanding problem

Ukraine is sharing battlefield data with UK to build AI models. “You can’t put a price on the data. Add up everything that the UK has given to Ukraine over the last several years – this data is worth all of that and more,” says one expert. www.newscientist.com/article/2587...

newscientist.comNations and big tech to train AI using ‘gold dust’ data from Ukraine | New ScientistForeign states and technology firms are scrambling for data from the Ukraine war so they can train the next-generation of AI models that might be used in defence and national security

Matthew Sparkes reposted

Ukraine's state-backed @u24.gov.ua has picked up on our drone exclusive: that fully AI-controlled drones were used in a fatal strike on Russian soldiers, marking the first disclosed incident of fully autonomous weapons being used to deadly effect. united24media.com/war-in-ukrai...

united24media.comShould Ukraine Have to Justify Its Use of Battlefield AI?Ukraine is pushing battlefield AI to counter Russia, raising a difficult question: how should a country at war balance survival, autonomy, and ethics?

The question now is, after years of positive data, why we don't have a roadmap for further improvement. When does ULEZ get expanded, or compliance criteria tightened-up? When does the first area become electric-only? It worked: great. What next? www.bbc.co.uk/news/article...

bbc.co.ukChildren's stunted lungs show recovery in ultra low emission zoneChildren's lung capacity caught up with their peers in less polluted areas once a clean air zone came in, a study shows.

AI models that can hack into computers via novel methods, at lightning pace, are upsetting the equilibrium of cybersecurity. “It’s like a scalpel: it saves lives in the hands of a doctor, and destroys lives in the hands of someone else. It's just a tool." www.newscientist.com/article/2583...

newscientist.comRogue hacking AIs have changed the cybersecurity landscape | New ScientistAttackers with no technical skills can now use AI models to find and exploit loopholes quickly and they can deploy them on a massive scale, raising fears that organisations without large budgets for c...

The latest DeepMind AI gives an extra day of warning ahead of deadly cyclones. AI is really shaking up the world of meteorology. One research told me: “You go on holiday for a week… and find out something else has happened.” www.newscientist.com/article/2583...

newscientist.comDeepMind AI gives an extra day of warning ahead of deadly cyclones | New ScientistAn AI model from DeepMind can predict cyclones three days ahead with a level of accuracy that previous models can only hit a day later

Accidentally crashing a 4000kg lump of metal into the moon at 8700km/h is part of the new commercial space race. Experts say we shouldn't make a habit out of it, and are working to ban it in future, but that some interesting science can be had as a side effect. www.newscientist.com/article/2583...

newscientist.comSpaceX rocket crashes into moon at 8700 kilometres per hour | New ScientistA rocket launched last year is thought to have smashed into the lunar surface this morning, creating the moon's newest crater

Authors look away: research shows people prefer AI-generated stories. There are nuances. The study looked at literary short stories, not populur genres. It used an older model. It wasn't a huge sample, etc. But it counters the "AI writing is obvious" idea. www.newscientist.com/article/2583...

newscientist.comMost people prefer AI-written stories – and can't tell when they are made by humans | New ScientistPeople appear to prefer short stories created by AI over those written by humans. They are also no better than chance at telling which is which

You have to zoom in pretty close, near the tap, because it was moving too fast for a decent photo. But I think we had a visit from a hummingbird hawk-moth: huge, frantic thing that was bullying the bees out of the way on every flower it visited.

Matthew Sparkes reposted

A new stealth drone spins at 1500rpm to blur into the background. Just like the blades of a fan it virtually disappears by moving fast than human sight can cope with. It's pretty nifty. But it's not very manoeuvrable and you can hear it from quite a distance... www.newscientist.com/article/2579...

newscientist.comStealth drone spins so fast that it disappears | New ScientistA drone that spins 25 times a second attempts can avoid being spotted by becoming nothing more than a vague blur – though the creators admit that it can still be easily heard

Matthew Sparkes reposted

Not ideal

Matthew Sparkes@sparkes.bsky.social

A list of global space launches designed to calm Cold War tensions and promote transparency has been missing from the UN's website for months. Experts warn it makes the world a less predictable and safe place. www.newscientist.com/article/2533...

A list of global space launches designed to calm Cold War tensions and promote transparency has been missing from the UN's website for months. Experts warn it makes the world a less predictable and safe place. www.newscientist.com/article/2533...

newscientist.comUN space database aimed at easing global tensions is mysteriously downA list of global space launches designed to calm cold war tensions and promote transparency has been missing from the UN's website for months

Are AI firms hoovering-up philosophers? “I wish we had made more progress on the big questions of philosophy – about consciousness, agency, morality, etc – prior to the advent of AI,” says Jonathan Birch at LSE. "If we had, we would have been better prepared." www.newscientist.com/article/2532...

newscientist.comCan the biggest problems in AI be solved by philosophy?AI companies are hiring philosophy graduates to help them understand the nature of consciousness, whether it can be replicated and how their systems can be made better and more reliable