The invention of the transistor at Bell Labs was not the first time someone had described such a device, its inventors discovered. CC-licensed photo by Bill Bradford 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.
There’s another post coming this week at the Social Warming Substack on Friday at 0845 UK time. Free signup.
A selection of 10 links for you. File early, file often. 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 Meta AI app lets you ‘discover’ people’s bizarrely personal chats • WIRED
Kylie Robison:
»
“WHAT COUNTIES [SIC] do younger women like older white men,” a public message from a user on Meta’s AI platform says. “I need details, I’m 66 and single. I’m from Iowa and open to moving to a new country if I can find a younger woman.” The chatbot responded enthusiastically: “You’re looking for a fresh start and love in a new place. That’s exciting!” before suggesting “Mediterranean countries like Spain or Italy, or even countries in Eastern Europe.”
This is just one of many seemingly personal conversations that can be publicly viewed on Meta AI, a chatbot platform that doubles as a social feed and launched in April. Within the Meta AI app, a “discover” tab shows a timeline of other people’s interactions with the chatbot; a short scroll down on the Meta AI website is an extensive collage. While some of the highlighted queries and answers are innocuous—trip itineraries, recipe advice—others reveal locations, telephone numbers, and other sensitive information, all tied to user names and profile photos.
Calli Schroeder, senior counsel for the Electronic Privacy Information Center, said in an interview with WIRED that she has seen people “sharing medical information, mental health information, home addresses, even things directly related to pending court cases.”
“All of that’s incredibly concerning, both because I think it points to how people are misunderstanding what these chatbots do or what they’re for and also misunderstanding how privacy works with these structures,” Schroeder says.
It’s unclear whether the users of the app are aware that their conversations with Meta’s AI are public or which users are trolling the platform after news outlets began reporting on it. The conversations are not public by default; users have to choose to share them.
«
I’ll go with them not knowing – Facebook/Meta’s designers are making the assumption that everyone has the same understanding of “share” on a button. Some people might interpret it to mean an instruction to the machine to share its insight just with them, the person making the enquiry. Besides which there’s a ton of junk and spam in that tab. It’s a crowded wasteland empty of human thought.
unique link to this extract
Dear Microsoft: stop it with Copilot already • Marc D Anderson’s Blog
Marc Anderson:
»
Until recently, one of the easiest ways to get “into” Microsoft 365 was to go to Office.com. I’ll readily admit that I never spent any time on the page, mainly because most of my time in Microsoft 365 is spent building sites, pages, libraries, etc.
But Microsoft’s stated intent – stated many times, in fact – was to help you get right back to your work by going to Office.com’s home page. It showed the Office apps for which you were licensed, and more importantly to many people, the files with which you’d recently been working.
[Didn’t know this. But Anderson points out that it has changed, and, as you’d expect, it’s for the worse – Overspill Ed]
…The fact that Microsoft is now forcing people to a page focused on Copilot shows a disconnect from a large portion of their user base. There are several important considerations here.
• First, many organizations simply aren’t ready to do anything with AI or Copilot. Putting Copilot front & center when they haven’t subscribed to Copilot at all is not useful for them
• Second, this is a significant change to a landing page which Microsoft has touted for years as “the place to start your day”. That use case has been removed totally from the page. No recent documents, no apps, no waffle, etc. Sure, there’s an Apps link to the left, but it adds an unnecessary extra click to go where we want to go
• Third, the only way to go from that page to something we’ve recently worked on is to search – encouraging people to use search where the Microsoft Graph was actually being useful before by showing recent work – or by asking Copilot. See my first point
• Finally, by forcing agents on people, we are going to get yet another ungoverned mess. Plus, if you don’t have Copilot licensing or pay-as-you-go enabled, you very quickly hit a dead end with agents. It’s a solution begging for use to invest in it – which is NOT a good look for Microsoft. Forcing people into buying licenses to make a page useful was not the ethos we expected historically from them.This change to Office.com just one example of Copilot being rammed down the throats of people who may not want it at all, while taking away features people actually used.
«
Trump Mobile’s “Made in the USA” T1 is… made in China • Apple Insider
Mike Wuerthele and Malcolm Owen:
»
The Trump Organization has launched Trump Mobile and plans to release the T1, a smartphone that it says is “made in USA” at the same time that the iPhone 17 will launch. The problem is, the phone was made in China. Marking ten years after the launch of President Donald Trump’s original presidential campaign, the Trump Organization has decided to launch its own mobile phone network. Dubbed Trump Mobile, it is a network that is being promoted as an All-American service,” and heavily leaning on the Trump brand. Trump Mobile frames itself as a “next-generation wireless provider,” with mentions of it delivering “top-tier connectivity” and “unbeatable value.” All with a “customer-first” approach and an “all-American service.”
…The Trump Mobile offering falls into two main offerings, consisting of the phone plan itself and the smartphone. Neither looks particularly great from a consumer point of view. And that smartphone that the Trump organization claims is made in the US? It’s a cheap Chinese android device available on Amazon for a third of the price with some new plastic slapped on, at which point they called it a day.
«
MVNO service – all yours for $47.45 (Trump is/was 47th and 45th president) per month for the mobile network. It’s possible that no artefact will say so clearly “I’m a mug” three years from now.
unique link to this extract
‘It’s absolutely f—ed’: Why Google’s new £1bn London office is in crisis • Daily Telegraph
Liam Kelly:
»
The crowning glory of Google’s new, massive headquarters in London’s King’s Cross is its rooftop garden. More than 300m long, with hundreds of trees across four stories and a running track, star designer Thomas Heatherwick envisaged it as a haven for the tech giant’s 7,000 staff, as well as bats, bees, birds and butterflies.
At least, it is meant to be the crowning glory. However, delays to the project have meant that, while it is still under construction, the building and its garden have been invaded by foxes. The vulpine skulk has taken advantage of the building’s lack of human occupants, digging burrows in the manicured grass and leaving their droppings around.
“Fox sightings at construction sites are pretty common, and our King’s Cross development is no exception,” a Google spokesperson said after a report on the London Centric website. “While foxes have been occasionally spotted at the site, their appearances have been brief and have had minimal impact on the ongoing construction.”
The foxes, pests though they are, may be the least of Google’s problems. Today, visitors to the construction site are met with the cacophonous sounds of drilling and hammering; the sights of scaffolding and cherry pickers obscuring the view; the constant bustle of workmen coming and going. The 11-storey building, the cost of which has never been confirmed but expected to be well north of £1bn, still appears to be a long way from being completed.
Building site sources tell The Telegraph that all manner of things have gone wrong, from shoddy workmanship that was, in effect, “hidden” because of the vastness of the project to wooden floors that became so saturated with rainwater that they need complete repairs. Much of the ground floor, which is supposed to house shops and other public spaces, remains a shell.
The date for its opening, which was meant to happen last year, has been repeatedly pushed back. “If they get this job done by the end of 2026 it would be a f—ing miracle,” one worker tells me. “I don’t think the people building it know what they are doing.” An electrician says: “They have unlimited money so they throw out ridiculous dates. It’s going to be interesting, but very stressful and long hours.” (Both Google and Heatherwick Studio declined to comment on these claims.)
«
WhatsApp is officially getting ads • The Verge
Emma Roth:
»
WhatsApp is rolling out ads. In an update on Monday, Meta announced that it will now show ads from businesses through its Stories-like status feature.
The status feature lives within WhatsApp’s “Updates” tab and allows users to share disappearing text, photo, voice notes, or video messages. But now, you might see sponsored content in addition to the updates shared by friends or family members.
Meta had been considering bringing advertising to WhatsApp for years now, an idea its founders were entirely against. Though Meta reversed plans to roll out in-app ads in 2020, WhatsApp head Will Cathcart confirmed in 2023 that the company was still working on an implementation. Last year, Meta made more than $160 billion in ad revenue.
Meta says it will tailor the ads to your interests by using “limited” information, including your country or city, language, the channels you follow, and how you interact with ads on the platform. You can also change your ad preferences from Meta’s Accounts Center if you’ve opted into the hub.
“We’ve been talking about our plans to build a business that does not interrupt your personal chats for years and we believe the Updates tab is the right place for these new features to work,” Meta writes in its update.
«
Fine by me as long as they stay there, because I never visit that tab. But I suspect that won’t be sufficient, over time, for Meta.
unique link to this extract
Founder of 23andMe buys back company out of bankruptcy auction • Financial Times via Ars Technica
Sujeet Indap:
»
Anne Wojcicki has been declared the winner of a bankruptcy auction for 23andMe, the genetics testing start-up she founded, prevailing over a rival bid from Regeneron Pharmaceuticals.
TTAM Research Institute, a non-profit public benefit company also founded by Wojcicki, won the auction with a $305m bid for the 23andMe assets, which will not come with any company liabilities attached.
23andMe filed for bankruptcy in March after rejecting several go-private offers from Wojcicki in recent years. Regeneron was declared the winning bidder in May after the company accepted a $256m bid in a previous auction.
TTAM then accused the debtor and its advisers of prematurely shutting down the May auction before it could put in a higher bid. The company in court filings said it rejected higher bids from TTAM after it could not confirm its ability to raise necessary financing.
TTAM later said it had obtained backing from a “Fortune 500 company with a current market capitalization of more than $400bn and $17bn of cash on hand.”
The federal bankruptcy court in Missouri held additional rounds of bidding on Friday morning to allow TTAM to formally submit a higher bid.
Regeneron had the opportunity to see the “best and final” TTAM bid and make a higher offer, but it declined. It is set to receive a $10m termination fee, according to court filings.
TTAM’s winning offer requires judicial approval, and a court hearing to approve the bid is set for next week.
«
So basically all back to square one. With a ton of money spent and now with a ton of DNA profiles stored.
unique link to this extract
How common is multiple invention? • Construction Physics
Brian Potter:
»
The frequency of multiple invention is a useful thing to know, because it can give us clues about the nature of technological progress. A very low rate of multiple invention suggests that progress might be driven by a small number of “genius” inventors (what we might call the Great Man Theory of technological progress), and that it might be highly historically contingent (if you re-rolled the dice of history, maybe you get a totally new set of inventions and a different technological palette).
A high rate of multiple invention suggests that progress is more a function of broad historical forces (that inventions appear when the conditions are right), and that progress is less contingent (if you re-rolled the dice of history, you’d get a similar progression of inventions). And if the rate of multiple invention is changing over time, perhaps the nature of technological progress is changing as well.
…First, there may be multiple descriptions of the same basic idea. When Bell Labs went to file a patent for the transistor in 1948, it was discovered that physicist Julius Lillienfeld had filed a patent for a similar idea in 1926. Lillienfeld likely never actually built his transistor (it would have almost certainly been impossible to build a working one due to material limitations), but he nevertheless described the concept.
Similarly, concepts for airships were described many years before a working one was built. I counted these descriptions as multiple inventions only if the description included a basic functional mechanism that could have plausibly worked: So Lillienfeld’s patent counts as a multiple invention, but someone idly stating that “it would be great if there was a solid state amplifier” wouldn’t.
…My main takeaway is that the ideas behind inventions are often in some sense “obvious,” or at least not so surprising or unexpected that many people won’t think of them.
«
The conclusion isn’t that obvious. But it is interesting.
unique link to this extract
Why Israel struck now • The Atlantic
Graeme Wood:
»
Membership in the nuclear club—the nine countries known to have nuclear weapons—comes with one incredible perk: near immunity from direct attack, even with conventional weapons, by other nuclear powers. India and Pakistan have bent this rule, but overall it has held, because the danger of nuclear escalation is just too high to risk it. The peculiar thing about Iran—what made it unique among aspiring inductees into that club—was that until recently, it enjoyed this perk even while its membership application remained under review. When countries attack Americans and American interests overseas, the United States is generally uninhibited in striking back. (Libya, Iraq, Sudan, and Afghanistan are recent examples.) Iran attacks American interests all the time—and yet it has been treated gently in return by every U.S. president since Jimmy Carter, as if it were not a nuclear aspirant but a club member already.
It achieved this deterrence by infiltrating much of the Middle East, and rigging it to blow. The first detonator was installed in Lebanon, in the form of Hezbollah, which succeeded so spectacularly there that it became the template for deterrence and punishment of Iran’s enemies across the region. The militias and states that arose on this template were the Axis of Resistance, and their ability to unleash havoc on the region (combined with Iran’s own conventional forces) was enough to make Israel and the United States repeatedly decline to touch Iran within its own borders, even when American soldiers were being killed in large numbers by Iranian proxies.
Why, then, did Iran not make a mad rush for a nuclear weapon? Because it already had the immunity that a nuclear weapon would confer—and because as long as it didn’t have a nuclear weapon, it could use its threat of getting a bomb to extract concessions from America and its allies. Instead of getting a bomb and joining the club, it could advance half the remaining distance to the nuclear threshold whenever it wished—always getting closer but, like Zeno, never getting to the end. Each step closer to the threshold served as an impetus to negotiation, a new reason to demand less support for dissent inside Iran or more money, lest the next step closer to the bomb be taken.
…An Iran without a vigorous Axis simply does not have much to bargain, or threaten, with. And a reduced Iran is just another country, a Sudan or Libya with better weather. The only way to recover the lost deterrence would be to close the distance to the threshold and achieve real nuclear statehood, rather than just the provisional and revocable version. Iran has in the past reached low points in its power, and it has taken, in some cases, years to recover its position and find a strategy. Perhaps its strategy was a rush for a bomb. Perhaps it was not, but Israel saw no point in waiting around to find out what it would do instead.
«
It is notable that nuclear-armed Israel is prepared to attack Iran to prevent it getting to nuclear status.
unique link to this extract
Stop sending us AI articles • The Critic Magazine
David Scullion and Ben Sixsmith:
»
We know we’re setting ourselves up for a fall when, on an off-day, we will unwittingly publish 3000 dry words written by ChatGPT on Nepal’s growing naval blockade. But most of the time we can tell when you’re sending us an article written by artificial intelligence, and when we suspect it, we run it through a system that gives us the score. If you think we haven’t noticed then it’s probably just because we’ve been too polite to tell you when we passed on your piece: we knew.
We’re a small team and we read a lot of pitches and articles sent to us on spec. This takes a long time that we’re happy to spend, and we’ve found some amazing writers who have gone on to be prolific and fascinating to read, here and elsewhere. We aim to be relaxed in how we handle talent, but we take very seriously our obligation to find and encourage new contributors. But the rise of the “AI-writer” threatens to bog us down, make our jobs tedious, and reduce the chance of new scribblers being found. When it takes longer to read an article than it took to generate it, we’ve all got a serious problem.
Most people have the good sense to be coy, and don’t tell us that the 2000 words they filed on the impact of global trade from Trump’s latest TRUTH SOCIAL post were not entirely written by themselves, but perhaps that will change too? We’ve even been sent an article which we were invited to run with two bylines: one of them the nominal author, and the other the name of the AI software which did 80% of the work. You will not be surprised that we turned it down.
«
It’s faintly astonishing that people are sending in freelance work – even on spec – that’s so obviously written by a chatbot. Although I can see the attraction for freelances: you get some quotes and some facts and you tell the chatbot to write a X,000-word feature which uses these facts and these quotes. And then you polish it.
unique link to this extract
Here comes the AI sponcon • The Verge
Mia Sato:
»
TikTok announced today it was adding new capabilities to Symphony, the company’s AI ads platform it launched in 2024. The features go beyond generating basic videos and images — instead, the system’s new output mimics what audiences are used to seeing from human influencers.
The company says advertisers will be able to upload images, provide a text prompt, and generate videos with virtual avatars holding products, trying on and modeling clothing, and displaying a brand’s app on a phone screen. Some features already available to TikTok users — like creating a video out of a photo — will also now be available to advertisers.
AI creep in the influencer industry has been a steady development: advertisers already have the option of using synthetic characters (sometimes resembling real people) to do things like read scripts to promote brands and products. This new set of features brings an interactivity, with virtual avatars essentially acting like human influencers by using and modeling products. or advertisers, the appeal is a mix of automating processes and cutting costs — an AI avatar can’t demand specific rates or terms in a contract, and a brand can generate an endless amount of content without recording each video separately.
«
“Sponcon”, if you didn’t know, is “sponsored content”. Perhaps this will make it easier for us to ignore it, since we will be able – surely! – to recognise that junk too.
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
The Overspill: when there's more that I want to say 