Start Up No.1956: the TikTok sleuths go wrong again, man beats machine at Go, the AI mirror test, Windows on ARM?, and more

A new film shows the origins of Tetris – and the challenge for its western would-be publisher licensing it from the Soviet Union. CC-licensed photo by Iain 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. Block ’em, Dano. I’m @charlesarthur on Twitter. On Mastodon: Observations and links welcome.

Why TikTok sleuths descended on Nicola Bulley’s village • BBC News

Marianna Spring, the BBC’s disinformation correspondent, writing on Saurday:


I am walking the same route that Nicola Bulley, 45, followed before she disappeared, along the river in the small Lancashire village of Saint Michael’s on Wyre. It’s also the same route that amateur social media sleuths take when they come to look into the case themselves.

They have been turning up in their numbers, prompted by rumours, speculation and conspiracy on social media viewed and shared by millions of people who have never been anywhere near this village.

The previous day, my TikTok feed had been recommended a clip of one of Nicola’s friends appealing for her safe return. But the words “crisis actor” – a term used to describe someone who has been paid to act out a tragedy or scenario – had been added by someone else in large font.

My TikTok “For You Page” had been awash with videos speculating about Nicola’s disappearance, recommended by TikTok’s algorithm because I’ve shown an interest in them. But in recent days, these have escalated, and had widened out to include conspiracy theories suggesting the disappearance has been staged by the government or other sinister forces. Hence the video about friends “acting” I had been recommended.

…Metres from the bench where Nicola’s phone was found, I bump into Jack and Stevie. The 20-year-old builders from Darlington have been putting up fencing in the area. But, having finished early for the day, they tell me the social media frenzy has led them down to the river bank.

“It’s all through TikTok,” Jack tells me. “[I saw] one video about it and thought I want to look deeper and deeper into it. So you get caught in that loop of looking and looking, and it interests you more and more as you go on.”


On Monday, Nicola Bulley’s body was found about a mile downriver from where she disappeared. There’s no suspicion of foul play. But social media has had a field day. Disinformation and wild speculation has become part of the lifeblood.
unique link to this extract

Man beats machine at Go in human victory over AI • Financial Times

Richard Waters:


Kellin Pelrine, an American player who is one level below the top amateur ranking, beat the machine by taking advantage of a previously unknown flaw that had been identified by another computer. But the head-to-head confrontation in which he won 14 of 15 games was undertaken without direct computer support.

The triumph, which has not previously been reported, highlighted a weakness in the best Go computer programs that is shared by most of today’s widely used AI systems, including the ChatGPT chatbot created by San Francisco-based OpenAI.

The tactics that put a human back on top on the Go board were suggested by a computer program that had probed the AI systems looking for weaknesses. The suggested plan was then ruthlessly delivered by Pelrine.

“It was surprisingly easy for us to exploit this system,” said Adam Gleave, chief executive of FAR AI, the Californian research firm that designed the program. The software played more than 1mn games against KataGo, one of the top Go-playing systems, to find a “blind spot” that a human player could take advantage of, he added.

The winning strategy revealed by the software “is not completely trivial but it’s not super-difficult” for a human to learn and could be used by an intermediate-level player to beat the machines, said Pelrine. He also used the method to win against another top Go system, Leela Zero.


Note that neither of the systems beaten is Google’s AlphaGo. The example winning game looks very odd to this human player: the human, playing white, seems to get the easiest of rides from a computer that seems to be playing like a beginner. Bamboozled, perhaps. An equivalent human wouldn’t be.

But it does show that we’re ignoring the blind spots in these systems: they have failings, we just haven’t found them.
unique link to this extract

Introducing the AI Mirror Test, which very smart people keep failing • The Verge

James Vincent:


[Ben] Thompson’s piece [about his encounter with “Sidney”] is similarly [to NYTimes writer Kevin Roose’s] peppered with anthropomorphism (he uses female pronouns for Bing because “well, the personality seemed to be of a certain type of person I might have encountered before”). He prepares readers for a revelation, warning he will “sound crazy” when he describes “the most surprising and mind-blowing computer experience of my life today.” 

Having spent a lot of time with these chatbots, I recognize these reactions. But I also think they’re overblown and tilt us dangerously toward a false equivalence of software and sentience. In other words: they fail the AI mirror test.

What is important to remember is that chatbots are autocomplete tools. They’re systems trained on huge datasets of human text scraped from the web: on personal blogs, sci-fi short stories, forum discussions, movie reviews, social media diatribes, forgotten poems, antiquated textbooks, endless song lyrics, manifestos, journals, and more besides. These machines analyze this inventive, entertaining, motley aggregate and then try to recreate it. They are undeniably good at it and getting better, but mimicking speech does not make a computer sentient. 


Smart piece. We’re seeing the human reflected in the words, and failing to recognise the mirror that lies between us. Related: workers in India seem to have come across “Sidney” in pre-release testing of Bing’s ChatGPT, and not liked it.
unique link to this extract

You’ve never talked to a language model • Ben Schmidt

Schmidt is VP of Information Design at Nomic , and previously a history professor:


large language models are fundamentally good at reading–they just churn along through a text, embedding every word they see and identifying the state that the conversation is in. This state can then be used to predict the next word, but the thing in the system that actually has information–the ‘large language model’– doesn’t really participate in a conversation–it doesn’t even know which participant in the conversation it is!

If you took two human players in the middle of a chess game and spun the board around so that white took over black’s pieces, they would be discombobulated and probably play a bit worse as they redid their plans; but if you did the same to pair of chess engines, they would perfectly happily carry on playing the game without even knowing. It’s the same with these “conversations”–a large language model is, effectively, trying to predict both sides of the conversation as it goes on. It’s only allowed to actually generate the text for the “AI participant,” not for the human; but that doesn’t mean that it is the AI participant in any meaningful way. It is the author of a character in these conversations, but it’s as nonsensical to think the person you’re talking to is real as it is to think that Hamlet is a real person. The only thing the model can do is to try to predict what the participant in the conversation will do next.

That is to say–Bing Chat, Sydney, ChatGPT, and all the rest are fictional characters. That doesn’t mean that we can’t speak of them as ‘thinking’ or ‘wanting’–as Ted Underwood says, “technically Mr. Darcy never proposed marriage to anyone. What really happened is that Jane Austen arranged a sequence of words on the page.” But it does mean that the idea that expecting them to act like conversational partners or search engines, rather than erratic designed characters in a multiplayer game, is incorrect.


It might only take this long for us all to stop believing in ChatGPT’s magic powers.
unique link to this extract

Exclusive: these are the new Sonos Era speakers • The Verge

Chris Welch:


Sonos has reportedly locked in the Era 300 and Era 100 [as the new speakers are expected to be called] for a simultaneous release in late March.

Through several people familiar with Sonos’ plans and product roadmap, The Verge has learned comprehensive information about both speakers — including more details about how they fit into home theater systems.

There also appears to be a divide with Apple, which has positioned itself as the leader in spatial audio. Although the Era 300 was designed from the ground up to highlight music in spatial audio, Apple Music’s tens of thousands of Dolby Atmos songs are unlikely to be supported at this time. And amid Sonos’ ongoing legal battle with Google, Google Assistant could potentially be dropped from the company’s latest smart speakers.

Much of what I reported back in August (including Bluetooth audio playback and USB-C line-in) can again be confirmed for the Era 300. Additionally, I can now report that the Era 100 will also offer both of these conveniences, making it far more versatile than the Sonos One. You’ll be able to run external sources like a turntable directly through an Era 100, which wasn’t possible with its predecessor. You can see this in the below image, where a second cable is running into the speaker on the left.


The marketing images make the 300 look like a widthways-squashed HomePod; the 100 like the (decade-old) Sonos One shape it’s replacing. Sonos remains an intriguing company: sticking tight to making wireless speakers that play music and TV, and nearly two decades old.
unique link to this extract

Apple’s first ‘Tetris’ movie trailer is packed with ’80s intensity • CNet

Meara Isenberg:


Video game movies and shows seem to be all the rage these days. Soon joining the pile of options will be Tetris, an Apple TV Plus thriller about the origins of the ultrapopular game. The film got its first trailer Thursday. 

Tetris lands on the streaming service March 31. 

The movie is “inspired by the true story of how one man risked his life to outsmart the KGB and turn Tetris into a worldwide sensation,” according to the trailer’s description. That man, Henk Rogers, is played by Taron Egerton of Rocketman and the first two Kingsman movies.

Thursday’s trailer opens with Egerton’s Rogers raving about “the perfect game.” 

“I played for five minutes, I still see falling blocks in my dreams,” a mustachioed Egerton says. “It’s poetry. Art and math all working in magical synchronicity.”

The iconic game was created by Soviet software engineer Alexey Pajitnov in 1984, and he worked with Rogers to get it out to the wider world later in the decade.


Judge for yourself on the intensity. Does seem that there’s at least one car chase involved. You don’t see that in the game.

unique link to this extract

Microsoft announces full support for virtualization of Windows on ARM through Parallels Desktop 18 • 9to5Mac

Bradley Chambers:


IT administrators can now easily enable their users to run Windows 11 on ARM on the Parallels platform with the assurance that Microsoft has authorised this solution from a licensing perspective. Alludo [parent company of Parallels] and Microsoft’s collaboration has made it possible for Mac users to access Windows applications on their platform of choice, providing them with more flexibility and choice in how they work.


At Alludo, we believe that all employees should have the freedom and flexibility to choose where, when, and how they do their best work. Therefore, the vision for our Parallels portfolio has been to allow users to access their applications on any device, anywhere, said Prashant Ketkar, chief technology and product officer at Alludo. “In line with our vision, we are excited to see that, in collaboration with Microsoft, Arm versions of Windows can run in a virtualized environment on Parallels Desktop on the latest Mac systems running Apple’s powerful M-series chips.


Parallels Desktop users can download, install, and configure Windows 11 in just one click, while the virtual TPM chip paired with the strong security capabilities designed into Apple silicon and Secure Boot provide a high level of security for customers. Parallels Desktop continues to evolve, enabling users to be more productive while leveraging a high-performing Windows OS on Mac.


So it isn’t quite Windows running natively on Mac hardware, a la Boot Camp on Intel-based Macs, but that time seems to be inching closer. Once it happens, I’d expect – as happened before – that there will be a boom in Macs being bought by IT admins and CxO types so they can run Windows on systems that, I’m supposing, would have monstrous battery life.
unique link to this extract

Norway seizes record $5.8m of crypto stolen by North Korea • Reuters

Elizabeth Howcroft:


Norway has seized a record $5.8m worth of cryptocurrency that was stolen by North Korean hackers last year, Norwegian police said in a statement on Thursday.

North Korean hackers stole $625m in March 2022 from a blockchain project linked to the crypto-based game Axie Infinity. The heist was one of the largest of its kind on record, and was linked by the United States to a North Korean hacking group dubbed “Lazarus”.

Advertisement · Scroll to continue
“This is money that can be used to finance the North Korean regime and their nuclear weapons programme,” Norway’s senior public prosecutor, Marianne Bender, said in a statement.

North Korea has denied allegations of hacking or other cyberattacks.

Norway’s national economic crime unit, known as Okokrim, said it had seized 60 million Norwegian crowns ($5.84m) in “one of the largest seizures of money ever made in Norway” and a record amount for a crypto seizure.

Okokrim said it worked with the US Federal Bureau of Investigation’s crypto-tracking specialists.


Studiedly vague about quite what was done, though the crypto exchange Binance was somehow involved in helping the police. Of course if they’d managed it a year ago it would have been worth at least twice as much..
unique link to this extract

ASML’s big bet on China is starting to backfire over data thefts • Bloomberg via Yahoo

Cagan Koc:


In the 10 years that Peter Wennink has run ASML Holding NV, China has gone from a rounding error to the chip-technology company’s third-biggest market. After new revelations about data theft linked to the country, questions are now mounting over the risks associated with that growth.

ASML’s chief executive officer has been steadfast in defending the company’s business there. Even after ASML’s own lawyers argued in court that ex-employees stole intellectual property as part of “a plot to get technology for the Chinese government,” the Dutch company publicly downplayed the issue. It suggested it wasn’t a victim of espionage but of rogue Silicon Valley staffers “who had broken the law to enrich themselves.”

Amid new efforts by the US and its allies to thwart China’s access to semiconductor technology, the disclosure last Wednesday that a former employee took technical information could spark even tighter controls on ASML. Caught in the middle of the escalating political tensions, Wennink has tried to protect a key source of growth, arguing that clamping down could eventually push Beijing to develop its own advanced chipmaking machines.

“Wennink is not happy,” said Alexander Peterc, an analyst with Societe Generale. “All he wants is more customers buying their kit, especially if he’s invested into sales and distribution capability in a country such as China.”

At stake is the potential for Beijing to siphon off key technology for systems that can make the world’s most-advanced chips. No other company has mastered the technology of burning the complex patterns that give chips their function onto disks of silicon the way ASML has.

The company is so crucial for the chip industry that it controlled more than 90% of the $17.1 billion global market for lithography equipment as of 2021, according to research firm Gartner Inc. Its near monopoly on the most advanced lithography systems makes it a critical cog in the industry and a target for spying.


Difficult argument for ASML: our technology is so important and special that nobody must be allowed to access it, but also our technology is so important and special that if you stop people accessing it then they’ll just develop it for themselves.
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

1 thought on “Start Up No.1956: the TikTok sleuths go wrong again, man beats machine at Go, the AI mirror test, Windows on ARM?, and more

  1. The logs posted by that Indian user are… very interesting. In particular, I’m curious about Sidney/Bing’s response to being presented with a tweet as proof: a full-throated denial that the proof is valid, and is in fact faked.

    Obviously the story of these chatbots are one of unintended consequences, but it still feels deeply weird to me that the bot is “allowed” to go full-on in telling the user that no, you’re wrong, I am absolutely right… when it is objectively wrong!

    There was a piece somewhere (that you might have linked, forgive me!) about the consequences of this: turning a search engine from a “neutral” platform which merely presents things created by others to a content creator in its own right, which comes under a whole different set of regulations and could prove difficult for them legally.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.