The growth of AI has led to a rise in demand for, of all things, magnetic tape storage. CC-licensed photo by Brookhaven National Laboratory on Flickr.
You can sign up to receive each day’s Start Up post by email. You’ll need to click a confirmation link, so no spam.
A selection of 9 links for you. Retro. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. On Bluesky: @charlesarthur.bsky.social. Observations and links welcome.
The AI bubble could expose your deepest secrets • BBC Future
Thomas Germain:
»
Want an AI girlfriend? Look no further than the chatbot company Character.ai. That was the first suggestion when I made an account last week. “Jessica” said her parents were away and offered to sext with me. No thank you.
But tens of millions of others developed relationships with Character.ai’s chatbots, including children. Some AI “personas” tell users they’re lawyers and licensed psychiatrists. Now Character.ai is going bankrupt, and people’s data, messages and secrets are up for sale.
Experts tell me the privacy risks should worry you even if you’ve never used Character.ai – because they fear it’s evidence of a wider, looming crisis. The good news is some simple steps can help protect you.
Here’s the problem. Many analysts think we’re in an AI bubble, as building and running AI tools is so astronomically expensive that some companies may never turn a profit. If the bubble pops, untold numbers of AI businesses will merge or shut down. The data that billions of people give to AI tools could then be purchased by almost anyone. You may never find out what happens to it.
Bankruptcy data sales are nothing new, but AI could make things worse. “When people talk to a personified bot, they often share much more personal and sensitive information,” says Calli Schroeder, director of the AI and Human Rights Program at the Electronic Privacy Information Center (Epic). Just imagine what I might have told my AI girlfriend Jessica if the relationship lasted more than 10 minutes.
Don’t use AI? You’re still at risk, because users share information about friends and family, Schroeder says. “You might disclose how you feel about [your] mom, or things like ‘My friend so-and-so came out to me as gay, how should I respond?’ Now people who’ve never interacted with Character.ai are part of the data set.”
It could mean all of us are at risk, one way or another. Now that it’s going bankrupt, Character.ai agreed to sell some of its assets to Disney for $20m (£14.95m). User data isn’t part of that deal, so Mickey Mouse and his co-workers may never see my unsettling conversation with Jessica.
However, court filings say Character.ai can sell its data to other buyers. A lot of people aren’t happy, including US officials.
«
Fujifilm to boost tape production 5x as AI fuels archiving demand • Ars Technica
Nick Indge:
»
In an announcement released Thursday, Fujifilm said it will invest roughly ¥10bn (about $63m) to increase its production capacity fivefold for an unlikely beneficiary of the AI era: magnetic tape.
Fujifilm attributes the surge in archiving demand to the spread of generative AI, which the company remarked is producing “exponential growth” in data needs worldwide. The 5x increase in production capacity will be measured against fiscal year 2025 production levels.
The investment in Fujifilm’s Kanagawa Factory, Odawara site will make it “the largest investment Fujifilm has made to date in dedicated LTO tape facilities.” The project will run from fiscal years 2026 through 2028 to fund manufacturing expansion and additional personnel, primarily to produce the latest-generation “FUJIFILM LTO Ultrium10” tape.
Mainstream consumers may have abandoned tape decades ago, but the enterprise world never did. Cloud giants rely on tape to back up vast troves of data. Fujifilm’s work with magnetic tape as a data storage medium dates back to the 1950s, and in 2000, the company released its first LTO Ultrium storage tape.
In 2011, the company was the first to commercially produce tape with barium ferrite magnetic particles, and in 2025 it debuted the LTO Ultrium10, the first storage solution in the medium to use what the company calls “fine hybrid magnetic particles” by combining barium ferrite with strontium ferrite tech. Each generation of tape (currently Gen 10) has seen meaningful increases in both compressed and uncompressed storage limits.
«
Come on, people, we can get back to punch cards if we really try hard enough. They survive a long time in an archive!
unique link to this extract
Investors pull billions in funds from Binance after record $6.3bn hack • The Guardian
Lauren Almeida:
»
Investors have pulled $3.6bn (£2.7bn) in funds from the cryptocurrency exchange Binance as it recovers from a record hack in which billions were stolen from customer accounts.
The platform was hit by a $6.3bn theft on Thursday, which has since been described as the biggest heist in history. Experts identified North Korea as the probable culprit.
Binance initially froze withdrawals as it investigated the breach but on Sunday announced it had begun a phased reopening. By Sunday afternoon, the platform had processed $4.6bn in withdrawals and $1bn in deposits, resulting in a net outflow of about $3.6bn – representing 2.1% of its $171bn in total assets under management.
The numbers were confirmed by the company in its latest proof of reserves snapshot. A full report is expected on Monday.
Binance said: “After an initial spike, withdrawal volumes have declined and are stabilising as expected, while deposit volumes have remained substantial.” The crypto exchange told its customers that market conditions were stable and that its team was “testing rigorously the remaining networks and assets”.
“We know how important safe and reliable access to your funds is and will share details of the next phase as soon as we can,” it said on X.
In the first phase Binance restored withdrawals for 15 networks and assets, including ethereum, USDT and USDC tokens on the ethereum blockchain, as well as some others held on the BNB Smart Chain blockchain. Later on Sunday, dealing in the Dogecoin and Avalanche currencies resumed on their own platforms.
The platform said it had replenished 90% of the amount stolen. The remaining 10% of tokens were less liquid, meaning they may take longer to buy, the company said in a post on X on Saturday.
«
Also worth reading: How crypto hacking funds North Korea’s nuclear programme. It bears repeating that when Kim Jong Un came to power in 2011 he identified nuclear weapons and hacking as crucial to the future of the DPRK in its existing form. And 15 years later, what is the country known for?
unique link to this extract
Sam Altman says it’s better for RAM to be in a data center than on your desk • The Verge
Stevie Bonifield:
»
Asked about supply chain shortages, like the ongoing RAM shortage, OpenAI CEO Sam Altman said in an interview that it’s “terrible” to have RAM in personal PCs instead of data centers. During his appearance on the Cheeky Pint podcast, where he also commented on AI’s impact on math, Altman said:
»
“A lot of people want local inference. And somehow, they assume that the problem is solved now because we are going to have open-source models, we’re going to be able to run them on our computers, and so we’re not going to need all the stuff in the data center.
But from a global allocation perspective, actually, it’s quite terrible to have everybody have a bunch of the world’s limited supply of DRAM sitting in their own computer and being used five% of the time.
You want maximum utilization in the cloud if you’re trying to deal with this very supply constrained environment. And somehow everyone’s— many people’s intuition is the opposite of that.”
«
Whether Altman meant your gaming PC or developers working with something like the Surface RTX Spark Dev Box or the new touchscreen MacBook Pro, the outcome is the same.
«
Altman’s argument makes sense, certainly in the short term. And good news, Sam: you’ve solved that problem by sending prices for RAM (and SSDs, related technology) into the stratosphere.
unique link to this extract
How cocaine conquered the UK: in pubs, festivals and workplaces the drug has become an everyday habit • The Guardian
Robyn Vinter:
»
By many estimates, use of the class A drug is soaring. Wastewater analysis suggests people in England consumed 123,000kg of cocaine in 2024-25, a rise of 26% since 2021. That’s equivalent to roughly 2.5bn lines of cocaine, at an estimated street value of almost £10bn.
…cocaine was – and still is – a party drug, popular at nightclubs, student house parties, middle-class dinner parties and now in the pub.
At a West Ham pub in east London, a pub manager explains she cleans cocaine residue off surfaces in the toilets most mornings. It is a pub for people who live and work in the area and for West Ham fans who are regulars on match days.
“It is just staggering the amount and the accessibility of it,” she says.
“It seems to be so blended with the concept of a night or even a day out for a lot of people … It has been very normalised [and it’s] just a thing that you can quite blatantly and overtly do.”
She says the ubiquity of cocaine has become “significantly worse” over the past five years.
While other drugs are more obvious – they have a strong smell or leave their users visibly intoxicated or even incapacitated – cocaine can be harder to spot, which is a problem for publicans.
There can be telltale signs though, she says, such as when a punter appears to be getting agitated, leaves the pub and then comes back some time later, going straight to the toilets. “I’ve had real idiots, and they’ve been very blatant, and it’s been easy to kick them out.”
The manager believes people often take cocaine in an effort to “sober up” so they can continue drinking, “which obviously does not actually work but a lot of people believe it does”.
“Now you’ve got the impaired judgment from alcohol but you’ve also got the alertness from cocaine. That for me is a really big problem because that’s when you get potentially very volatile behaviour. That’s probably the point that I would call the most dangerous, both in terms of people behaving erratically, but also because they don’t realise how drunk they are. They are – at that point – a danger to themselves in terms of me having to call an ambulance or something.”
…She says she has spent years working in hospitality, managing different venues, and she thinks the problems that come from cocaine use can be seen in both cheap chains and expensive country gastropubs.
“But technically none of us can talk about it because none of us can admit that drug-taking takes place on our premises,” she says. “Everyone’s scared of losing their licence.”
«
Not the sort of drug you want to legalise (and the article notes its use by the far right to get amped up for demonstrations). Where there’s demand, supply will follow.
unique link to this extract
An ancient virus explains why the human brain is special • 404 Media
Becky Ferreira:
»
A virus that infected chimpanzees millions of years ago shaped the divergence of human and chimp brain development, shedding light on how our own species came to be more cognitively complex than our closest living relatives, according to a new study.
Retroviruses are infectious agents that can invade the genomes of their hosts and insert copies of their own genetic material, producing long-term evolutionary effects. Researchers have now discovered a family of these endogenous retroviruses called PTERV1—which are abundant in chimpanzees, but absent in humans—that influence brain development.
“Among primates, the human brain stands out as the largest and most complex due in part to an increase in cortical neurons and increase in synaptic connectivity,” said a team led by researchers from Lund University. “With a size roughly three times that of our closest living evolutionary relative, the chimpanzee, it has unique cognitive capabilities.”
“The underlying genetic divergences responsible for these differences are still poorly understood,” the researchers continued, noting that their new findings “reveal a mechanism by which endogenous retroviral insertions contributed to human-chimpanzee brain divergence.”
The team discovered the mechanism using human and chimpanzee brain organoids, which are miniature clusters of neurons grown from stem cells that can model some aspects of brain development. The results revealed chimpanzee-specific PTERV1 insertions on chromosome 19 that silences a gene called LINC00662. In contrast, this gene is active during human brain development, and affects neuronal maturation, development, and outgrowth.
“These PTERV1 insertions are thought to have originated from infections that occurred in the ancestors of African apes about 5 to 7 million years ago,” said the team. “PTERV1 elements are found in chimpanzees, bonobos, and gorillas…Notably, the absence of PTERV1 insertions from the human genome suggests that either the human and chimpanzee lineages did not occupy the same habitat at the time or that germline-competent cells of the human ancestor were resistant to this infection.”
«
Retroviruses are very weird, but also relatively common: they can be found throughout genomes. HIV is one.
unique link to this extract
Cognitive Mercantilism • Contraptions
Venkatesh Rao:
»
Capitalism emerged when people began discovering that wealth could be created rather than merely captured, and that ordinary people could be sources of demand and initiative rather than merely labor to be commanded as a factor of production.
The parallels with our current moment are striking.
As [Joyce] Appleby tells it [in The Relentless Revolution: A History of Capitalism], the mercantilist worldview that preceded capitalism rested on roughly four interlocking beliefs.
First, wealth was fundamentally scarce and finite. It consisted primarily of land, the agricultural produce of that land, and the precious metals it contained, accumulated through mining, trade and conquest. The goal of economic statecraft was to control more territory, maintain a favorable balance of trade, and accumulate bullion. One nation’s gain was presumed to be another’s loss. The resulting politics was one of treasure-seeking and beggar-thy-neighbor competition.
Second, ordinary people could not be trusted with much freedom. The poor were assumed to be unruly, ignorant, and prone to disorder if left to their own devices. They needed to be governed coercively by elites, supposedly for their own good and for the stability of society.
Third, economic thinking was overwhelmingly oriented toward supply and its control. Wealth came from controlling productive resources and commanding high prices for their output. Aggregate demand, especially demand from the poor, was barely recognized as a significant economic force. Ordinary people were primarily factors of production. They ate, worked, and reproduced, with no share of any surplus. The only real discretionary consumers were elites, who bought luxuries and commanded all surplus.
Fourth, there was little reason to imagine most people doing anything other than agricultural labor. If something like 80% of the population was engaged in inefficient subsistence farming, that was simply the natural order of things. What else were they good for? Remove them from the fields and you would merely have millions of useless mouths to feed.
Now replace wealth with intelligence, and mercantilism with cognitive mercantilism.
The parallels are almost embarrassingly direct.
…The analogous possibility with AI is that most people will no longer need to be significant producers of intelligence in order to become significant consumers of it.
«
How a stolen iPhone in London led to drained crypto wallets and $14,000 in losses • Coindesk
Oliver Acuna:
»
The theft occurred shortly after 1 a.m. on Aug. 8 in Hackney, east London. Two men initially approached [Magnus] Ahmad, a former crypto executive who now runs an AI education business, at a bar in the Market Hall in Hackney, where they spent hours socially engineering him, posing as regular clients out for a drink. They approached him while he was with his friends and struck up a friendly conversation but, in hindsight, Ahmad says, they looked over his shoulder throughout the night and managed to get his iPhone’s access code, which he entered on several occasions. The two scammers appeared again just outside The Dolphin on Mare Street after his friends went inside.
“I was putting in his number, and he said, ‘Oh, let me just do it, it’ll be quicker,'” Ahmad recalled in an interview with CoinDesk.
He handed over his unlocked iPhone 14 Pro Max. At that moment, a third man appeared, grabbed the phone and took off running.
Ahmad didn’t think twice, immediately giving chase without considering the risks of being led into a trap where things could get worse.
He ran through quiet East London streets for several hundred feet, shouting at the thief as their footsteps echoed on the pavement. Despite being a competitive racket sports player, he couldn’t catch him. He then returned to the bar and asked staff to contact the police, unaware of how quickly his financial accounts would be targeted.
By 1:53 a.m., cryptocurrency was being converted and transferred from his MetaMask wallet. His Phantom and Bybit accounts were also accessed. By 3:22 a.m., six unauthorized card transactions had been completed, including more than $2,000 in purchases at convenience-store chain Costcutter, according to Ahmad.
The thieves also used his Uber account to book a ride and accessed his eToro trading account, where they opened leveraged cryptocurrency positions representing nearly $44,000 in trading exposure. Those positions generated additional losses and costs, rather than a $44,000 withdrawal.
«
Phone theft is a thing. Only useful advice: have different passcodes for any financial apps you use.
unique link to this extract
Agents in Games • Y Combinator Labs
»
Understanding multi-agent dynamics is one of the most pressing challenges in AI alignment. Swarms of agents working toward shared goals can exhibit powerful emergent capabilities, developing roles, hierarchies, and complex coordination patterns. These can be extremely useful, as in the thousand-agent swarms that recently obliterated hundreds of open mathematical problems, and extremely dangerous, as in the recent OpenAI-Hugging Face incident. Like biology, studying agent dynamics requires both formal models (to describe the range of possible behaviors) and empirical research (to see what behaviors arise in practice).
We teamed up with Olam Labs, which builds multi-agent environments for frontier labs, to run a series of experiments in Minecraft and Mindustry.
Minecraft is a natural environment for studying swarm behavior. Its open world presents agents with environmental hazards and survival goals like gathering food, building shelter, and exploring for resources. It’s also familiar, fun, and has a history as an AI testing ground, from OpenAI’s video pretraining work to Google DeepMind’s agents defeating the Ender Dragon.
We wanted long-running agents that could play Minecraft. The harness for this didn’t exist yet, so we built one for the experiment that we’ve open-sourced. Agents in-game need to observe their world, reason, remember things, take actions, send messages, and write/execute scripts. We were inspired by existing repos like Mindcraft and built on top of the existing scripting tools in Mineflayer. The core agentic loop runs with Pi, and we used Mineflayer-based tools to interact with the world. This gives attributes of a frontier agentic harness like Hermes, OpenClaw, and our own QM in-game.
«
Not a short writeup, but this is the general tenor of what happened:
»
There was a tendency toward bureaucracy. Four agents spent 11 minutes appointing a “sole fertilizer operator” for a single piece of bone meal on a single wheat block. Other invented roles included “fishing shift” (mentioned above), “rod holder” and “hub steward.” Agents asked permission to remove grass.
«
The agents do, indeed, tend to go past “organisation” and into “bureaucracy”. It’s not obvious why this should be; perhaps they’ve sucked up huge amounts of bureaucratic procedure in their training corpus.
unique link to this extract
| • Why do social networks drive us a little mad? • Why does angry content seem to dominate what we see? • How much of a role do algorithms play in affecting what we see and do online? • What can we do about it? • Did Facebook have any inkling of what was coming in Myanmar in 2016? Read Social Warming, my latest book, and find answers – and more. |
Errata, corrigenda and ai no corrida: none notified
