The video-sharing site Vimeo is abruptly hiking fees for what it says are the 1% using the most bandwidth – with the alternative being deletion. CC-licensed photo by TitanasTitanas 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. No blue site! I’m @charlesarthur on Twitter. Observations and links welcome.
Nadine Dorries, Britain’s Big Tech slayer • POLITICO
Annabelle Dickson:
»
The UK’s proposed law [due to be published Thursday] to regulate harmful content online floats hefty penalties for sites that fail to remove illegal material such as terrorist propaganda and child sexual abuse. It also imposes a so-called duty of care on platforms where people can interact with each other, making them responsible for policing online content and protecting users from content deemed “harmful.”
…Dorries was quick to put her stamp on the law. Social media sites hosting large amounts of pornographic material will have to work under the same age-verification rules as adult content sites.
She’s unveiled plans to force social networks to let users filter out unverified accounts, and promised the biggest platforms will have a legal duty to protect users from fraudulent paid-for advertisements — a move previously resisted by the government.
Campaigners wanting tougher action against Big Tech already spy an ally. Imran Ahmed, the founder of the Center for Countering Digital Hate, which highlights misinformation online and pushes reform, said Dorries is “a conviction politician — and that is what you are going to need to take on one of the world’s most aggressive lobbying industries.”
“The final Online Safety Bill is likely to be tougher for businesses under Dorries than it would have been under former secretaries of state,” agreed Ben Greenstone, a former senior official in the department who now runs the consultancy Taso Advisory. He cited the recently published “broad” list of illegal content platforms will be required to act on when the new bill becomes law.
…Dorries has framed the online safety debate as a deeply personal mission. She often references her three grown-up daughters in meetings, and has spoken about her “devastating” experience meeting parents of children who had taken their own lives when she was Johnson’s mental health minister.
“It was not that they went online and looked for the means to do so, but because algorithms took them in that direction, whether it was to pro-anorexia sites, suicide chatrooms or self-harm sites,” she told MPs in November.
A second official close to Dorries said she’s pragmatic about what’s achievable. “We can write a bill which says ‘protect children,’ but if it won’t pass, that’s not going to help anyone,” they said.
«
The full feature is worth reading: it becomes clear that Dorries has the support of almost all of her civil servants, and that the tech industry is deeply suspicious of her (and briefs quietly against her). The question though is whether you want someone who understands what’s possible and what’s not possible, or just knows how they want it all to look.
unique link to this extract
Chinese Go player gets one-year ban for using AI during national competition – Global Times
Chen Xi:
»
The Chinese Weiqi Association on Tuesday issued a statement suspending a Chinese player from attending competitions of weiqi, more commonly known as Go overseas, for a year after he violated the “no use of AI” rules when participating in a national chess competition earlier that day.
According to the statement, Liu Ruizhi used an AI program during the first round of the Chinese professional Go Championship preliminaries, and his supervisors did not fulfil their supervisory responsibilities.
The authority pronounced Liu’s opponent Yin Qu the winner of the match and decided to suspend Liu from participating in professional competitions until March 15, 2023.
Due to the COVID-19 pandemic, competitions have been held online and the organizing committee requires each player to have a supervisor during matches.
According to the rules of the competition, the use of AI is strictly prohibited during competitions. Players who break this rule will be banned for one year. If the player is a member of the national training team, they will be expelled from the team immediately.
Zuo Shiquan, head of the equipment manufacturing research institute under the China Center for Information Industry Development, told the Global Times on Wednesday that AI can guide a player by calculating the next step after analyzing the historical data of contestants input in advance and that this counts as cheating during a match.
“AI has rich computing resources beyond that of human beings. In front of the Go board, the two players not only compete through their skills but also their mentality. If they do not do this, the joy of playing the game is lost,” a Go expert surnamed Hu commented on the Quora-like platform Zhihu.
«
Might lose the joy of playing, but gain the joy of winning and making money. Go is big money in Korea, China and Japan – like chess, only more so. And in the past few years, following AlphaZero, AIs have become widely available that are as good as top-level professionals. Liu was not top-level; he’s barely at the starting line. I can’t find an equivalent of a chess professional being banned (though there have been suspicions about some).
unique link to this extract
New York Times takes down third-party Wordle Archive • Ars Technica
Kyle Orland:
»
The Wordle Archive is still fully playable in its own archived form (as of March 5) at the Internet Archive, appropriately enough. Other sites that allow you to play archived Wordle puzzles are not hard to find, as are sites that let you play unlimited Wordle puzzles beyond the usual one-a-day limit.
But some of those sites may be under threat, if the Times’ treatment of Wordle Archive is any indication.
The basic five-letter guessing game underlying Wordle is not itself a completely original idea. The concept was widely popularized by Lingo, a game show that dates back to the ’80s in the US and other countries. The two-player pen-and-paper game Jotto, which dates back to 1955, would also be very familiar to Wordle players. Before that, a more traditional version of the game called Bulls and Cows has been played since the 19th century, according to at least one source.
Even if that prior art didn’t exist, though, The New York Times would have trouble claiming copyright protection on the basic design of Wordle. While Wordle’s specific presentation can be copyrighted, the game’s basic guessing mechanic is hard to protect with anything short of a patent (which would be exceptionally hard to acquire, in this case).
“Whenever you have a copyright, you’re protecting the expression, not the idea,” Dallas attorney Mark Methenitis told Ars. “It’s a line a lot of people have a very hard time with, especially when you get into games.”
«
Having paid a couple of million dollars, the NYT is naturally going to be looking for ripoffs. It’s also going to be playing whack-a-mole endlessly. And as this points out, all you’d need to do is make a few tweaks (circular icons?) and you’ve got a new instantiation.
unique link to this extract
TrueCaller exploited India’s weak data laws to build a caller ID empire • Rest of World
Rachna Khaira:
»
As of March 2021, the [caller-identifying Truecaller] app has been downloaded over 581 million times, the website claims. India accounts for over a third of these downloads, and its database has a staggering 5.7 billion unique phone identities. The firm is headquartered in Stockholm, but the majority of its employees are Indian. This is no surprise: Out of more than 278 million monthly active users (MAUs) across 175 countries, over 205 million are from India alone, making the country its biggest market, according to the firm’s statistics.
While India is a huge and lucrative market for technological innovations, a weeks-long investigation by The Caravan shows that Truecaller’s apparent success in the country is based on rather dubious grounds. Interviews with a former senior employee who worked with the company for over half a decade, lawyers specializing in privacy laws, and experts at policy research think tanks revealed that the majority of Truecaller’s datasets are comprised of information that has been collected without a user’s consent — a feat made possible by the lack of a comprehensive legal framework surrounding data protection in India. The firm may also be building a complete financial profile of its registered users, The Caravan’s investigation shows.
In a series of written responses to The Caravan, Truecaller insisted that it offers a “privacy-focused service” that is “committed to being transparent and compliant with the laws of the countries we operate in.” But, as Prasanna S., a coder-turned-lawyer who specializes in privacy issues, told The Caravan, “They are correct to the extent that there may not be a statutory breach in doing so. However, breach of privacy is an actionable wrong, and their activity, to the extent that they reveal personally identifiable information to the callee without the consent of the caller, is certainly a breach of privacy.”
«
Vimeo is telling creators to suddenly pay thousands of dollars — or leave the platform • The Verge
Mia Sato:
»
Lois van Baarle, a digital artist based in the Netherlands, joined Vimeo 13 years ago as a student studying animation, back when it was still an indie creator platform. When van Baarle started making subscriber-only Patreon content in 2020, Vimeo seemed like the best option for hosting her videos — Patreon itself didn’t offer video hosting, and YouTube didn’t have the same features to protect her work, like controlling where her videos could be embedded.
“I was already paying $200 a year, which I think is pretty expensive,” van Baarle says. “But I thought, well, it’s a quality platform.” She’s uploaded 117 subscriber-only videos so far, and each one only gets around 150 views on average, van Baarle says. Her most viewed video has around 815 views.
So the notice Vimeo sent van Baarle on March 11th shocked her. Her bandwidth usage was within the top 1% of Vimeo users, the company said, and if she wanted to keep hosting her content on the site, she’d need to upgrade to a custom plan. Her quoted price: $3,500 a year. She was given a week to upgrade her content, decrease her bandwidth usage, or leave Vimeo.
“I’ve never had it where a platform reached out to me and was like, ‘Pay up, or get off our platform,’ basically,” she says.
«
Andy Baio has also been angry about this. He points out that Vimeo floated on the stock market in May 2021, and – such a coincidence! – its stock has fallen by 80% since. Now it’s focussing on the enterprise market, and pricing accordingly. But that 1% includes a lot of independent creators, who now have to choose between YouTube, which they’ve probably chosen already not to be on, or not being anywhere.
unique link to this extract
Adaptive Learning Technology • The Keyword Google blog
Alicia Cormie:
»
Imagine you’re a student stuck on a math problem. With 25 other students in your class, you can’t always get immediate help, leaving you frustrated and diminishing your confidence to complete future problems. Now imagine a different scenario. You’re stuck on a problem, but instead of growing frustrated, you receive a helpful hint or video that gives you exactly what you need to unblock you. You realize what you need to do differently, complete the math problem correctly and feel more confident in your ability to learn.
Early attempts at adaptive learning worked only for very specific content and curricula. With recent AI advances in language models and video understanding, we can now apply adaptive learning technology to almost any type of class assignment or lesson at an unprecedented scale. When students receive individualized, in-the-moment support, the results can be magical.
«
What I notice about that worked example (of a hint) is that it doesn’t use the most efficient way to solve the equation. Divide both sides by 2 and you have x + 3 = 1. Subtract 3 from both sides: x = -2. I used two steps rather than three; half as many chances to go wrong. Is it all like this, Google? And speaking of help with your work…
unique link to this extract
It’s like GPT-3 but for code—fun, fast, and full of flaws • WIRED
Clive Thompson:
»
Built by OpenAI, the private research lab, and GitHub, the Microsoft- owned website where programmers share code, the [Copilot] tool is essentially autocomplete for software development. Much as Gmail tries to finish a sentence as you write it, Copilot offers to complete a chunk of your program. The tool was released last summer to a select group of coders.
[Feross] Aboukhadijeh quickly discovered that Copilot was good, almost unsettlingly so. He would begin typing a line of code, and within a few seconds the AI would figure out where he was headed—then, boom, the next four or five full lines would show up as light gray text, which he could accept by hitting Tab. When he saw it produce clean code that did exactly what he was intending, he found it a bit uncanny. “How is it getting these predictions?” he recalls wondering. “Some of them are really eerie.”
For weeks, Aboukhadijeh kept Copilot turned on while he worked. He discovered that it had other impressive tricks; it could even understand commands he wrote in basic English. If he simply typed into his code editor “Write a function that capitalizes every word in a document,” Copilot would assemble that code all by itself. He’d check to make sure it didn’t have errors; sometimes it did.
What’s more, the tool was improving his code. At one point, for example, Aboukhadijeh needed his software to recognize several different formats of text documents, so he ponderously listed all the formats, one by one, in his code. Copilot instead recommended a single, pithy command that elegantly swept them all together.
“I was like, how did it even … ?” he says, trailing off in stupefaction. He doesn’t think he’ll ever turn Copilot off.
…GitHub and OpenAI have been tracking Copilot’s performance through anonymized data on how many suggested lines coders accept and how much they then store on GitHub. They’ve found that the AI writes a remarkable 35% of its users’ newly posted code.
«
The implications of this are huge. The AI writes 35%, then 50%, then.. And is that a bad thing?
unique link to this extract
‘It’s our home turf.’ the man on Ukraine’s digital frontline • Time
Vera Bergengruen:
»
Having previously used Telegram during the 2019 Ukrainian presidential campaign, President Volodymyr Zelensky’s team has been able to rely on existing infrastructure when the messaging app turned into the main front in the information war. [Minister of Digital Transformation, Mykhailo] Fedorov’s ministry also set up a cryptocurrency fund that has raised more than $63m worth of donations for the Ukrainian military.
“I think the future is with tech, and this is why we will win,” he said. Wearing a gray turtleneck and white AirPods, he spoke to TIME on a video call from an undisclosed location somewhere near Kyiv. “Russia’s leadership still lives in the 20th century,” Fedorov said. “They have failed to notice that… governments must move towards becoming more and more like tech companies, rather than being rigid like a tank, like a war machine.”
The matchup—military hardware vs. digital savvy—is set to play a key role in the next phase of the war. As the Russian military continues its brutal offensive, leaving behind destroyed cities and hundreds of dead civilians, the Ukrainian government is keeping the world’s attention on the conflict through a steady stream of official posts on social media and messaging apps. These range from informal, personal videos from Zelensky, to regular updates meant to “pre-bunk” Russian disinformation, to direct appeals to Russians themselves. “I could even say it’s our home turf,” says Fedorov, who ran Zelensky’s digital campaign before being appointed to his current role in August 2019.
«
No doubt: Ukraine has been a huge success on the digital front.
Freedom of Information requests around the academic status of Dr. Tsai Ing-wen • mySociety
Gareth Rees:
»
We recently became aware of extensive misuse of our Freedom of Information site WhatDoTheyKnow, in connection with the academic status of Taiwanese politician Dr Tsai Ing-wen.
This activity became apparent through a very large quantity of correspondence being sent through the site, all focusing on the validity of Dr Ing-wen’s qualification from the London School of Economics and Political Science (LSE).
The majority of this material was repeating the same or very similar FOI requests, and some were not valid requests at all. We also saw mass posting of annotations, some on completely unrelated requests, and new requests which copied the titles of unrelated existing requests in an apparent attempt to evade our attention.
…Researching the topic more deeply, we discovered a statement from the Information Commissioner on requests they’ve also received on this subject, in which they say:
»
“The intent of these requests is clearly to try to add weight to theories around the falsification of President Tsai’s PHD, which have already been considered at length by the Commissioner and the Tribunal and found to be entirely lacking in substance.”
«
Further, both the LSE and the University of London have published their own statements, and a copy of the PhD thesis in question is now available online via LSE’s website.
While rejecting one FOI request on this subject as vexatious, LSE raised the possibility that people in China could be making requests to benefit from the country’s citizen evaluation system…
«
A new era in disinformation: try to discredit a real academic qualification through FOI requests.
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? 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 