Introducing AI systems for tennis line calling has had unexpected effects on umpires’ tendency to wrongly call serves out. CC-licensed photo by Brianna Laugher 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. New calls please. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. Observations and links welcome.
Air Canada told it is responsible for errors by its website chatbot • Vancouver Sun
Susan Lazaruk:
»
An Air Canada passenger from B.C. [British Columbia] has won his fight after the airline refused him a retroactive discount, claiming it wasn’t responsible for promising the refund because it was made in error by the airline’s online chatbot.
Artificial intelligence law experts say it’s a sign of disputes to come if companies don’t ensure accuracy when increasingly relying on artificial intelligence to deal with customers.
Jake Moffatt booked a flight to Toronto with Air Canada to attend his grandmother’s funeral in 2022 using the website’s chatbot, which advised him he could pay full fare and apply for a bereavement fare later, according to the decision by B.C. civil resolution tribunal.
But an Air Canada employee later told him that he couldn’t apply for the discount after the flight.
“Air Canada says it cannot be held liable for the information provided by the chatbot,” said tribunal member Christopher Rivers in his written reasons for decision posted online. It “suggests the chatbot is a separate legal entity that is responsible for its own actions,” he said Rivers. “This is a remarkable submission.”
When Moffatt asked Air Canada’s automated response system about reduced fares for those travelling because of a death in the immediate family, the chatbot answered he should submit his claim within 90 days to get a refund.
His total fare for the return trip was $1,640, and he was told the bereavement fare would be about $760 in total, a $880 difference, he told the tribunal. He later submitted a request for the partial refund and included a screenshot of the chatbot conversation, the tribunal said.
…The airline argued it could not be held liable for information provided by one of its agents, servants or representatives, including a chatbot, Rivers said, adding it didn’t say why it believed that.
«
As someone pointed out on Twitter: the point at which personhood is claimed for an AI isn’t when it’s conscious, it’s when an airline needs to get out of paying a refund.
unique link to this extract
The effect of front-end vehicle height on pedestrian death risk • ScienceDirect
Justin Tyndall is at the University of Hawaii department of economics:
»
Pedestrian deaths in the US have risen in recent years. Concurrently, US vehicles have increased in size, which may pose a safety risk for pedestrians. In particular, the increased height of vehicle front-ends may present a danger for pedestrians in a crash, as the point of vehicle contact is more likely to occur at the pedestrian’s chest or head.
I merge US crash data with a public data set on vehicle dimensions to test for the impact of vehicle height on the likelihood that a struck pedestrian dies. After controlling for crash characteristics, I estimate a 10 cm increase in the vehicle’s front-end height is associated with a 22% increase in fatality risk. I estimate that a cap on front-end vehicle heights of 1.25 m would reduce annual US pedestrian deaths by 509.
«
Reminds me strongly of the UK’s (and Europe’s) efforts to limit the use of bull bars on the front of cars because they caused extra fatalities and worsened injuries. The Independent campaigned very hard on this in the late 1990s, with some success. Lives were saved.
unique link to this extract
EU to hit Apple with first ever fine in €500m music streaming penalty • Financial Times
Javier Espinoza:
»
Brussels is to impose its first ever fine on tech giant Apple for allegedly breaking EU law over access to its music streaming services, according to five people with direct knowledge of the long-running investigation.
The fine, which is in the region of €500m and is expected to be announced early next month, is the culmination of a European Commission antitrust probe into whether Apple has used its own platform to favour its services over those of competitors.
The probe is investigating whether Apple blocked apps from informing iPhone users of cheaper alternatives to access music subscriptions outside the App Store. It was launched after music-streaming app Spotify made a formal complaint to regulators in 2019.
The Commission will say Apple’s actions are illegal and go against the bloc’s rules that enforce competition in the single market, the people familiar with the case told the Financial Times. It will ban Apple’s practice of blocking music services from letting users outside its App Store switch to cheaper alternatives.
Brussels will accuse Apple of abusing its powerful position and imposing anti-competitive trading practices on rivals, the people said, adding that the EU would say the tech giant’s terms were “unfair trading conditions”.
It is one of the most significant financial penalties levied by the EU on big tech companies. A series of fines against Google levied over several years and amounting to about €8bn are being contested in court.
«
True, it’s the first fine from the EU on antitrust, but back in 2011/2012 Apple was in hot water with the EU antitrust group over cartel pricing of its iBooks. That was settled by letting Amazon sell at the cartel prices for two years.
unique link to this extract
What tennis reveals about AI’s impact on human behaviour • The Economist
»
Car drivers, financial traders and air-traffic controllers already routinely see their decisions overruled by AI systems put in place to rapidly correct poor judgment. Doctors, judges and even soldiers could be next.
Much of this correction happens out of the public eye, thwarting would-be analysts. But, says [behavioural economist at Northwestern University, David] Almog, “tennis is one of the most visible settings where final decision rights are granted to AI.” That is why, together with colleagues in America and Australia, he has looked at whether tennis umpires and line judges correctly called balls in or out during nearly 100,000 points played in some 700 matches across the world, both before and after the introduction of the Hawk-Eye ball-tracking system in 2006.
The Hawk-Eye system, now used at most elite tournaments, uses between six and ten cameras positioned around the court to create a three-dimensional representation of the ball’s trajectory. This can then be presented on a screen visible to players, spectators and officials—as well as TV viewers. Players can use it to appeal human decisions, with the AI’s verdict considered final. Bad calls from line judges and umpires are now often overturned.
The latest analysis from Mr Almog and his colleagues, published as a preprint last month, showed that Hawk-Eye oversight has prompted human officials to up their game and make 8% less mistakes than before it was introduced. …But when the researchers looked at serves in particular, and especially in cases where the served ball landed within 20mm either side of a line, they were surprised to see the error rate soar. The umpires and line judges, it turned out, had switched strategy. Before Hawk-Eye, they were more likely to call a serve out when it was in. But afterwards, they were even more likely to wave through balls that were actually out.
«
Because, Almog suggests, umpires want to avoid being shown – on a giant screen! – to have interrupted the point wrongly. (He didn’t actually talk to the umpires, but any change would have been unconscious.)
The article makes more general points about humans v AI – but now, bigger tournaments have ELC (electronic line calling) where there isn’t any review: the machines call it, and that’s that.
unique link to this extract
Switzerland proposes an UN expert group on solar geoengineering • Climate Change News
Matteo Civillini:
»
Switzerland wants to advance global talks on whether controversial solar geoengineering techniques should be used to compensate for climate change by cooling down the earth.
It is proposing to create the first United Nations expert group to “examine risks and opportunities” of solar radiation management (SRM), a suite of largely untested technologies aimed at dimming the sun.
The panel would be made up of experts appointed by member states of the UN’s environment programme (Unep) and representatives of international scientific bodies, according to a draft resolution submitted by Switzerland and seen by Climate Home.
Governments will negotiate and vote on the proposal at Unep’s meeting due to start at the end of February in Nairobi, Kenya. It has been formally endorsed by Senegal, Georgia, Monaco and Guinea.
A Swiss government spokesperson told Climate Home that SRM is “a new topic on the political agenda” and Switzerland is “committed to ensuring that states are informed about these technologies, in particular about possible risks and cross-border effects”.
Solar geoengineering is a deeply contested topic and scientists are divided over whether it should be explored at all as a potential solution.
Ines Camilloni, a climatology professor at the University of Buenos Aires, welcomed Switzerland’s proposal, saying the UN “is in a good position to facilitate equitable, transparent, and inclusive discussions”.
«
As I previously said: first it’s tried by small groups, then it gets government money, then someone ridiculously rich goes and does it on their own. Examples: aircraft, cars, rockets.
unique link to this extract
AI and Japan as a safe space • Pure Invention
Matt Alt:
»
OpenAI chose to give their new system a Japanese name. Sora means “sky” in Japanese. Its creators chose it, they told the New York Times, because it “evokes the idea of limitless creative potential.”
Limitless it may be, but many of the videos OpenAI released to demonstrate the technology are steeped the imagery of one country, and that’s Japan. The Twitter announcement showcases a drone-like shot of a Tokyo street. Other samples included a night scene in Shibuya (linked to at the top of this post), a scene of rooftops speeding by the window of a Japanese commuter train, and a cute character raking stones in a Zen rock garden. The Sora website is topped by a video of origami birds nesting in a tree.
At first glance the videos impress. But similar to the case of the AI kimono I wrote about last year, those who know Japan will find themselves quickly sliding into the uncanny valley of gibberish street signs and snow on the ground in cherry blossom season and all of the other assorted janky weirdness that comes with generative AI. That weirdness isn’t a bug, but a sort of feature. Because this isn’t really Japan dreamed by a machine — it’s Japan dreamed by a machine that’s been trained on foreign fantasies of Japan. (The baked-in Orientalism of American AI is one of many reasons domestic Japanese startups are scrambling to conjure up their own.)
Far more interesting is the question of why OpenAI chose so many Japanese things (or more precisely, Japanese seeming things) to introduce Sora to the world.
…“Made in Japan” was a joke in the immediate postwar era, and then something akin to a threat in the Eighties. Today it is a badge of authenticity, deployed anywhere status needs to be conferred to a quotidian item: Japanese denim, Japanese whiskey, Japanese cleaning magic, Japanese Breakfast (okay, so that last one’s an indie-pop band.)
Japan may not confer any cool factor in the AI sphere, but it possesses undeniable cool factor in the real world. It’s also safe, in all senses of the word. It’s seen as free from crime and societal strife. It isn’t percieved as a threat, or even involved in touchy geopolitical issues.
«
The majority of traffic from Elon Musk’s X may have been fake during the Super Bowl, report suggests • Mashable
Matt Binder:
»
Super Bowl 2024 shattered records, with the NFL championship broadcast on CBS becoming the most-watched televised event in US history.
Also riding high from the big game? Elon Musk’s X. The company formerly known as Twitter published its own press release, lauding Super Bowl LVIII as one of the biggest events ever on the social media platform with more than 10 billion impressions and over 1 billion video views.
However, it appears that a significant portion of that traffic on X could be fake, according to data provided to Mashable by CHEQ, a leading cybersecurity firm that tracks bots and fake users.
According to CHEQ, a whopping 75.85% of traffic from X to its advertising clients’ websites during the weekend of the Super Bowl was fake.
“I’ve never seen anything even remotely close to 50 percent, not to mention 76 percent,” CHEQ founder and CEO Guy Tytunovich told Mashable regarding X’s fake traffic data. “I’m amazed…I’ve never, ever, ever, ever seen anything even remotely close.”
CHEQ’s data for this report is based on 144,000 visits to its clients’ sites that came from X during Super Bowl weekend, from Friday, Feb. 9 up until the end of Super Bowl Sunday on Feb. 11. The data was collected from across CHEQ’s 15,000 total clients. It’s a small portion of the relevant data, and it’s not scientifically sampled, but it nonetheless suggests a dramatic trend.
CHEQ monitors bots and fake users across the internet in order to minimize online ad fraud for its clients. Tytunovich’s company accomplishes this by tracking how visitors from different sources, such as X, interact with a client’s page after they click one of their links. The company can also tell when a bot is passing itself off as a real user, such as when a fraudulent user is faking what type of operating system they are using to view a website.
«
Just as relevant is that the proportion of bot traffic from other platforms, such as Facebook and TikTok, is in the low single digits. Seems Elon hasn’t got a handle on the bot problem at all. As any eX-Twitter user could tell you.
unique link to this extract
Zero plans for public onshore windfarms submitted last year in England • The Guardian
Fiona Harvey:
»
No new proposals for general-use windfarms were submitted for planning permission in England last year, despite the government’s much-vaunted relaxation of planning restrictions.
Only seven applications were submitted for onshore wind turbines for the whole of 2023 in England, new data from the government has shown, and all of those developments were for the replacement of existing turbines or for private sites, where the energy produced is destined for a particular consumer, such as a business.
The number was even lower than the 10 applications submitted in 2022, when the de facto ban was still in force.
Four onshore wind developments were granted planning permission in England last year from prior applications, all of which were either turbine replacements or for private use, and work began on one 4MW project in Leighton Buzzard in Bedfordshire, which had received permission the year before.
Last September, ministers announced changes to the restrictive regulations that had in effect ruled out onshore wind turbine construction in England since 2015, brought in by David Cameron to appease rightwing Conservatives.
Rishi Sunak agreed to amend the regulations last year under pressure from his backbenchers who were concerned about the impact of the ban on energy prices. But campaigners pointed out that the relaxation of the ban was only partial, and warned it was likely to be ineffective.
«
And lo, the campaigners were correct. How unsurprising.
unique link to this extract
Welding method drastically cuts time to make mini nuclear reactors • The Times
Emma Powell:
»
One of Britain’s oldest steelmakers has developed a manufacturing technique that it claims could drastically reduce the time and cost to produce small modular reactor (SMR) nuclear power stations, which have been proposed as one way to bridge the nation’s energy gap.
Sheffield Forgemasters has become the first to use the so-called electron beam welding method to produce one of the core parts of a small modular reactor (SMR) at scale.
Nuclear pressure vessels are thick steel containers that hold nuclear fuel when the reactors operate and provide one of several barriers that keep radioactive material out of the environment.
Electron beam welding works by firing electrons at an extremely high speed to join two pieces of metal together. The main difference to traditional welding methods is that no third-party material is introduced to make the join.
Using traditional techniques, the welding process alone can take at least 120 to 150 days. This new method can reduce the time to about two hours, according to Jesus Talamantes-Silva, director of research at Sheffield Forgemasters, drastically accelerating the manufacturing of SMRs. “That’s how disruptive this technology is,” he said.
The technique is already being used in the automotive and aerospace industries to produce smaller, relatively low-value components. Forgemasters is the first to use the welding technology to build a full-scale SMR pressure vessel, which weighs about 57 tonnes, has a diameter of three metres and walls with a thickness of 200 millimetres.
Unlike conventional plants, SMRs can be factory built. The government wants to open up far more areas as potential sites, replacing rules that allow nuclear power stations only in eight named locations, as it attempts to reach a target of 24 gigawatts of nuclear capacity by 2050, from 6GW at present.
«
If only there were a similar way to speed up the approval process by the same proportion.
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 