Driving along the Hutchinson River Parkway means your car numberplate will be analysed by AI to see if it has been on a suspicious journey. Good or bad? CC-licensed photo by Doug Kerr 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 10 links for you. Use them wisely. I’m @charlesarthur on Twitter. On Mastodon: https://newsie.social/@charlesarthur. Observations and links welcome.
The spam bots have now found Threads, as company announces its own ‘rate limits’ • TechCrunch
Sarah Perez:
»
It looks like Twitter isn’t the only one having to turn to rate limits — or limits on how many posts users can view. In an amusing turn of events, Twitter’s latest rival, Instagram Threads, announced this afternoon that it, too, has to tighten up on rate limits due to spam attacks. Laughed Twitter owner Elon Musk in a reply to a screenshot of the announcement posted on Twitter, “Lmaooo Copy 🐈 [cat].”
As you may recall, Twitter earlier this month had to enforce new limits on how many tweets users could read as the service suffered an extended outage. Explained Musk at the time, Twitter was facing “extreme levels of data scraping” from hundreds of organizations and other “system manipulation.” As a result, Twitter chose to curb the problem by initially allowing Verified users (paying subscribers) to peruse a maximum of 6,000 posts daily, while unverified users could only view 600. After some backlash from users, Musk later increased the limits to 10,000 for Verified accounts, 1,000 for unverified accounts, and 500 for new, unverified accounts.
Over the weekend, Musk said he would increase the rate limit again for Verified users by 50%, which implies they would now be able to see 15,000 posts.
Twitter had been criticized for its unorthodox solution to the spam and bot problem, which some suggested wouldn’t have been an issue if Twitter hadn’t laid off such a large swath of its engineering staff. After all, not being able to scroll the Twitter timeline for long periods of time had never been an issue in the pre-Elon days (except, of course, in the earliest years when the fail whale was a regular occurrence).
Instagram head Adam Mosseri explained the problem in a post on the Threads app this afternoon, noting that “Spam attacks have picked up so we’re going to have to get tighter on things like rate limits, which is going to mean more unintentionally limiting active people (false positives). If you get caught up [in] those protections let us know.”
«
Spam isn’t failure. Getting spam attacks means you’re successful: big enough to matter. The question then is how well you deal with it.
unique link to this extract
Alzheimer’s drug donanemab helps most when taken at earliest disease stage, study finds • Nature
Sara Reardon:
»
An experimental drug can slow progression of Alzheimer’s disease in those who start it when the disease is still in its early stages. The drug, a monoclonal antibody called donanemab, does not improve symptoms. But among people who started taking it at the earliest stages of Alzheimer’s, 47% had no disease progression on some measures after one year, compared to 29% who took placebo.
The drug does not provide as much benefit to people at later stages of the disease or those with a common genetic mutation that raises the risk of Alzheimer’s.
“This decade is already proving to be the decade of Alzheimer’s,” said Alzheimer’s Association chief science officer Maria Carrillo at a press conference at the Alzheimer’s Association International Conference (AAIC) in Amsterdam. “It’s important to now double down and not slow down.”
Donanemab’s manufacturer Eli Lilly, based in Indianapolis, Indiana, presented the results of the 1,736-person trial today at AAIC and published them1 in the Journal of the American Medical Association. The company released partial results in May, but those results left researchers with questions about the drug’s safety and efficacy in certain patient populations.
«
No doubt Eli Lilly would be happy if people start taking it prophylactically, years ahead of any possibility of developing the disease. And it certainly creates an incentive for gene testing to see if you’re vulnerable. Also worth reading this article from the Alzheimer’s Society from last year about the “amyloid controversy”.
unique link to this extract
SEC is worried chatbots could fuel a market panic • The Verge
Emilia David:
»
The US Securities and Exchange Commission (SEC) has expressed concern about generative AI’s impact on financial markets.
In a speech given to the National Press Club on Monday, SEC Chair Gary Gensler said recent advances in generative AI increase the possibility of institutions relying on the same subset of information to make decisions.
Gensler said the large demand for data and computing power could mean only a few tech platforms may dominate the field, narrowing the field of AI models companies can use. If a model provides inaccurate or irrelevant information, financial institutions may end up using the same flawed data and making the same bad decisions — creating the risk of something like the 2008 financial crisis, where banks played “follow the leader” based on information from credit raters, or the Twitter-fueled run on Silicon Valley Bank. Gensler compared the potential fallout to something like the 2008 crisis, which he said demonstrated the risks of a “centralized dataset or model” in finance.
»
“AI may heighten financial fragility as it could promote herding with individual actors making similar decisions because they are getting the same signal from a base model or data aggregator,” Gensler said. He added that the rise of generative AI and other deep-learning models “could exacerbate the inherent network interconnectedness of the global financial system.”
«
«
Or, put another way, the whole thing is so teeteringly unstable that it’s a Jenga tower with many of the blocks pulled out and put on the top.
unique link to this extract
This AI watches millions of cars and tells cops if you’re driving like a criminal • Forbes
Thomas Brewster:
»
In March of 2022, David Zayas was driving down the Hutchinson River Parkway in Scarsdale. His car, a gray Chevrolet, was entirely unremarkable, as was its speed. But to the Westchester County Police Department, the car was cause for concern and Zayas a possible criminal; its powerful new AI tool had identified the vehicle’s behavior as suspicious.
Searching through a database of 1.6 billion license plate records collected over the last two years from locations across New York State, the AI determined that Zayas’ car was on a journey typical of a drug trafficker. According to a Department of Justice prosecutor filing, it made nine trips from Massachusetts to different parts of New York between October 2020 and August 2021 following routes known to be used by narcotics pushers and for conspicuously short stays. So on March 10 last year, Westchester PD pulled him over and searched his car, finding 112 grams of crack cocaine, a semiautomatic pistol and $34,000 in cash inside, according to court documents. A year later, Zayas pleaded guilty to a drug trafficking charge.
The previously unreported case is a window into the evolution of AI-powered policing, and a harbinger of the constitutional issues that will inevitably accompany it. Typically, Automatic License Plate Recognition (ALPR) technology is used to search for plates linked to specific crimes. But in this case it was used to examine the driving patterns of anyone passing one of Westchester County’s 480 cameras over a two-year period. Zayas’ lawyer Ben Gold contested the AI-gathered evidence against his client, decrying it as “dragnet surveillance.
…To Gold, the system’s analysis of every car caught by a camera amounted to an “unprecedented search.” “This is the specter of modern surveillance that the Fourth Amendment must guard against,” he wrote, in his motion to suppress the evidence.”
«
Er, OK, but Zayas was driving like a criminal because.. he was a criminal? The story doesn’t say how many cars were pulled over on suspicion, out of the 16m licence plates being scanned per week.
unique link to this extract
Behind Twitter’s poop emoji PR • Semafor
Max Tani:
»
as Yaccarino began angling for the CEO job at Twitter earlier this year, [her personal PR guy Joe Bennarroch] wasn’t afraid to cross some company lines. When NBC News reporter Ben Collins tweeted about a Semafor story documenting advertisers’ concerns about Musk, Benarroch called to reprimand him, a rare instance of a business-side employee expressing criticism of a journalist’s editorial views.
Benarroch declined to comment on his time at NBCU or his new role.
Benarroch and Yaccarino succeeded in elevating Yaccarino’s profile at NBC, and her iconic stature in the ad industry helped her get the top job at Twitter.
But their attempts to use the same tactics at Twitter so far have not worked, as she’s been unable to establish herself as the company’s true chief executive. She’s delivered the official message in a series of clunky tweets on a platform dominated by her boss’s politicized and profane stream-of-consciousness.
Benarroch has quietly attempted to implement a traditional press strategy: Since joining last month, he’s done outreach to reporters, spinning negative stories, and attempting to influence the narrative around the new CEO in the background. He confidentially shared her day one internal memo with tech reporters last month, and has flagged some of her noteworthy tweets.
But even Benarroch’s title is a reflection of attempting to placate Musk’s whims. Despite spending much of his time working the phone with reporters, he maintains a business operations title, the result of Musk’s dismissiveness of standard comms work.
Benarroch has pleaded with staff to keep conversations internal, sending around a memo last month telling employees to report leakers. The note promptly leaked.
«
It’s hardly unheard of for PR folk to call journalists to complain about a story they don’t like. (Well, maybe to pearl-clutching American journalists.) Though Tani’s method of expressing a view about Benarroch (read the headline again) is quite neat.
Also, “[has] been unable to establish herself as the company’s true chief executive” is quite a clause.
unique link to this extract
Why social media is hardly social any more • Financial Times
Elaine Moore:
»
Threads, Meta’s new social network, had 100mn sign ups in its first five days. Not bad for a watered-down version of Twitter. According to Zuckerberg, the idea is to create a public conversations app for a billion people.
Listening to a billion people talk to one another sounds like a nightmare. But that’s not quite what Zuckerberg means. Threads is less public town square than stage. He doesn’t want us all to be part of the conversation, he wants us in the audience.
Social media networks are not very sociable these days. Feeds are algorithmic, which means you see whatever the apps want to show you. After I joined Threads, I saw a lot of brands and celebrities. I couldn’t tell you what my friends were posting but I could tell you that reality star Bethenny Frankel had thoughts on the new Barbie movie.
Once upon a time, people joined social media networks so they could connect with one another. I signed up to Facebook in 2007 to see what my friends were up to online. It’s hard to remember why it was so interesting to look at lots of blurry photos of a night out, but I spent a lot of time doing it.
That has now been superseded by content from strangers. I still have all my social media accounts but I rarely post anything. For many of us, the point of TikTok, Instagram, YouTube and Twitter is not to upload our own posts or look at what our friends are doing but to watch a small number of popular creators. Instead of talking to one another, we have become mostly silent onlookers.
«
There’s a misconception here. Social media has always been dominated by a few accounts which generated most of the content, with a huge remainder which were essentially passive, generating very little content viewed beyond their own narrow networks. We’ve always (mostly) been silent onlookers.
unique link to this extract
Far-right Twitter influencers first on Elon Musk’s monetization scheme • The Washington Post
Taylor Lorenz:
»
not all prominent right-wing Twitter contributors appeared to be part of the program. When asked if she was part of the program, Chaya Raichik, the creator of @libsoftiktok, offered a tongue-in-cheek response claiming that her relationship with Musk was thriving. She did not respond to a question about whether she was receiving payments under the program.
Anti-Trump influencers Ed Krassenstein and Brian Krassenstein, who were previously banned from Twitter in 2019, also announced that they were part of the program. Musk did not respond to a request for comment emailed to him at Twitter and at SpaceX, another company he owns.
“I think that there are some conservative content creators who are unhappy,” said Kris Ruby, a conservative influencer and president of Ruby Media Group. “It doesn’t seem even across the board. I don’t think the playing field is level.” She said some on the right who weren’t included in the program, despite meeting all the criteria, are venting in private. “Most conservatives don’t want to go up against the wrath of Elon and what happens when you criticize him,” she said. “We’ve seen that he’s not really applying the terms of service equally across the board.”
Twitter claimed in a blog post that creators’ share of advertising revenue would be based on a calculation of replies to their posts and monthly impressions. However, on Friday, Musk tweeted that payouts were not tied to public impressions but were calculated using a proprietary metric based on ads served to other verified users.
The program is available only in countries where Stripe, a payment platform, supports payouts, and recipients must pay for Twitter Blue, the platform’s monthly subscription service, to be eligible.
Not all creators who want to monetize will be able to. Creators who apply to the program will have to pass “human review,” and there is currently no open application for those interested in joining.
«
In other words: it’s a complete con. It’s just those who interact with Musk a lot. However, they also repel Twitter’s largely non-far-right audience, who are suddenly finding Threads a lot more attractive.
unique link to this extract
The best way to save American lives on the road • Time
Daniel Knowles is the author of Carmageddon: How Cars Make Life Worse and What to Do About It:
»
Near my home in Wicker Park, on Chicago’s north-west side, is an intersection known by some in the neighborhood semi-affectionately as “the crotch.” In Chicago’s almost perfect grid, it is where North Avenue, Milwaukee Avenue and Damen Avenue all meet in one point, creating a six-way road intersection. Its nickname fits it, because this is the part of Wicker Park you want to look away from. Nobody likes the intersection. When you cross North Avenue on foot, you have to watch out, because when the walk signal turns white is exactly when drivers try to illegally turn left and run you over. If you are on a bike, as I usually am when I cross it, it is even scarier, as you pick your way between parked cars and stressed drivers. But it also sucks to drive through. Traffic is almost permanently backed up, and at busy times, it can take 10 minutes to get through the lights.
There have been 690 crashes at the Crotch over the past six years. At least 19 pedestrians and cyclists have suffered incapacitating injuries. And yet, there is, a simple way to make the intersection safer. It should be turned into a roundabout (also known as a traffic circle or rotary). Based on my crude highway engineering (staring at Google maps), it probably could not be a true roundabout (one where cars are free to enter at all times, giving priority to those already on the roundabout). I suspect traffic lights would still be needed for the pedestrian crossings to work. But if you put an island in the middle, instead of a giant expanse of empty asphalt, it would force drivers slow down as they travel through it. The illegal left turn would be impossible to make, because the spot drivers wait in blocking traffic would be occupied by the island. It would be far less dangerous.
Actually roundabouts should be installed almost everywhere in America. Why? Because they save lives.
«
Intrigued, I took a look on Apple Maps at the intersection: see below. And yup, it’s an obvious candidate for a roundabout with traffic lights. (Bear in mind that the cars would go counter-clockwise around it.) Might be a bit of fun installing the roundabout, of course.
This geothermal startup showed its wells can be used like a giant underground battery • MIT Technology Review
James Temple:
»
In late January, a geothermal power startup began conducting an experiment deep below the desert floor of northern Nevada. It pumped water thousands of feet underground and then held it there, watching for what would happen.
Geothermal power plants work by circulating water through hot rock deep beneath the surface. In most modern plants, it resurfaces at a well head, where it’s hot enough to convert refrigerants or other fluids into vapor that cranks a turbine, generating electricity.
But Houston-based Fervo Energy is testing out a new spin on the standard approach—and on that day, its engineers and executives were simply interested in generating data.
The readings from gauges planted throughout the company’s twin wells showed that pressure quickly began to build, as water that had nowhere else to go actually flexed the rock itself. When they finally released the valve, the output of water surged and it continued pumping out at higher-than-normal levels for hours.
The results from the initial experiments—which MIT Technology Review is reporting exclusively—suggest Fervo can create flexible geothermal power plants, capable of ramping electricity output up or down as needed. Potentially more important, the system can store up energy for hours or even days and deliver it back over similar periods, effectively acting as a giant and very long-lasting battery. That means the plants could shut down production when solar and wind farms are cranking, and provide a rich stream of clean electricity when those sources flag.
There are remaining questions about how well, affordably, and safely this will work on larger scales. But if Fervo can build commercial plants with this added functionality, it will fill a critical gap in today’s grids, making it cheaper and easier to eliminate greenhouse-gas emissions from electricity systems.
«
You have to love sentences like that final one above which begin “But if…” It’s essentially saying “OK, everything’s against this. Ignore that though and…” I think I wrote a lot of science “breakthrough” stories which then stumbled over the “But if” hurdle.
unique link to this extract
New lawsuit against Bing based on allegedly AI-hallucinated libellous statements • Reason
Eugene Volokh:
»
When people search for Jeffery Battle in Bing, they get the following (at least sometimes; this is the output of a search that I ran Tuesday):
[image suggesting that someone who is the CEO of a company has been sentenced to 18 years in prison for seditious conspiracy]
But it turns out that this combines facts about two separate people with similar names: (1) Jeffery Battle, who is indeed apparently a veteran, businessman, and adjunct professor, and (2) Jeffrey Leon Battle, who was convicted of trying to join the Taliban shortly after 9/11. The two have nothing in common other than their similar names. The Aerospace Professor did not plead guilty to seditious conspiracy.
And this Bing output doesn’t just list the facts about each of the Battles separately, the way that search engine results have long listed separate pages separately. Rather, it expressly connects the two, with the “However, Battle was sentenced …” transition, which conveys the message that all the facts are about one person. And to my knowledge, this connection was entirely made up out of whole cloth by Bing’s summarization feature (which is apparently based on ChatGPT); I know of no other site that actually makes any such connection (which I stress again is an entirely factually unfounded connection).
Battle is now suing Microsoft for libel over this, in Battle v. Microsoft (D. Md.) (filed Friday). He’s representing himself, and the Complaint is flawed in various ways. But if the case is properly framed, he may well have a serious argument. That is especially so if he can substantiate his allegations that he had informed Microsoft of the problem and it didn’t promptly fix it.
«
There it is again: “but if”. Very tricky for Battle to argue that this is directly Bing’s fault when Microsoft can argue that it’s essentially a random number generator which put a 1 against a 2; that doesn’t make it a serial number generator.
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 