Start Up No.1490: how Zuckerberg helped right-wingers, Google fires another AI lead, the Texans facing giant electricity bills, and more

White attacks: but is that a comment about a racist onslaught, or a chess game? AI systems can’t tell. CC-licensed photo by Andrew Malone 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. Warmer. I’m @charlesarthur on Twitter. Observations and links welcome.

Who is still buying VHS tapes? • The New York Times

Hannah Selinger:


On Instagram, sellers tout videos for sale, like the 2003 Jerry Bruckheimer film “Kangaroo Jack,” a comedy involving a beauty salon owner — played by Jerry O’Connell — and a kangaroo. Asking price? $190. (Mr. O’Connell commented on the post from his personal account, writing, “Hold steady. Price seems fair. It is a Classic.”)

If $190 feels outrageous for a film about a kangaroo accidentally coming into money, consider the price of a limited-edition copy of the 1989 Disney film “The Little Mermaid,” which is listed on Etsy for $45,000. The cover art for this hard-to-find copy is said to contain a male anatomical part drawn into a sea castle.

There is, it turns out, much demand for these old VHS tapes, price tags notwithstanding, and despite post-2006 advancements in technology. Driving the passionate collection of this form of media is the belief that VHS offers something that other types of media cannot.
“The general perception that people can essentially order whatever movie they want from home is flat-out wrong,” said Matthew Booth, 47, the owner of Videodrome in Atlanta, which sells VHS tapes in addition to its Blu-ray and DVD rental business.

Streaming, Mr. Booth said, was “promised as a giant video store on the internet, where a customer was only one click away from the exact film they were looking for.”

But the reality, he said, is that new releases are prohibitively expensive, content is “fractured” between subscription services, and movies operate in cycles, often disappearing before people have the chance to watch them. In that sense, the VHS tape offers something the current market cannot: a vast library of moving images that are unavailable anywhere else.


Technologies never die, they just get less and less use.
unique link to this extract

Margaret Mitchell fired from Google • Axios

Ina Fried:


Margaret Mitchell, the co-lead of Google’s Ethical AI team, says that the company has fired her following an investigation into her use of corporate email.

Google was already under fire for its ouster of Timnit Gebru, the other co-lead of the team. Mitchell has been locked out of the corporate email since last month after what a source says was her effort to search corporate correspondence for evidence to back up Gebru’s claim of discrimination and harassment.

The move comes the same day that Google announced internally it had completed its investigation of Gebru’s exit, as Axios first reported. The company didn’t release the findings of that investigation, but announced a series of policy changes, including tying executive pay companywide to diversity and inclusion efforts.


Google’s really having some growing pains, and the problem here seems likely to become endemic: that there will be disgruntled employees who don’t like what it’s doing in some area or other and will stir things up. That’s not a comment on the merits or otherwise of this specific case, more a prediction about what will follow. This is hardly the first case where there’s been internal upheaval that’s become public, but we’ve usually seen it nearer the top level.
unique link to this extract

Apple cracks down on apps with ‘irrationally high prices’ as App Store scams are exposed • 9to5Mac

Guilherme Rambo:


It looks like Apple has started to crack down on scam attempts by rejecting apps that look like they have subscriptions or other in-app purchases with prices that don’t seem reasonable to the App Review team.

9to5Mac obtained access to a rejection email shared by a developer that provides a subscription service through their app. It shows a rejection message from Apple telling them that their app would not be approved because the prices of their in-app purchase products “do not reflect the value of the features and content offered to the user.” Apple’s email goes as far as calling it a “rip-off to customers” (you can read the full letter at the end of this post).

We were initially skeptical about the veracity of this email given some of the wording choices, but looking through Apple’s App Store Review Guidelines, it’s possible to find the term “rip-off” at least twice, such as in section 3, where Apple states that “we won’t distribute apps and in-app purchase items that are clear rip-offs.”

In contact with the developer of the rejected app, we were able to verify the authenticity of the rejection email from Apple. Unfortunately in this case, it seems clear that the rejection was a mistake. The developer was able to work with the app review team and eventually got their app approved by explaining that the subscription price was justified because the app employed paid APIs to perform its tasks. Just as with many other items in the guidelines, it’s possible for the review team at Apple to encounter false positives that lead to wrongful rejections of apps, which highlights why moderating the App Store is such a complicated task.

This rule has been in place for a long time, so it’s unclear when Apple started enforcing it more rigorously.


Very likely that Apple is responding to the PR noise around this – as well it should. Though it could also remove weekly subscriptions, which always feel like a problem looking for a place to happen.
unique link to this extract

Online speech is now an existential question for tech • WSJ

Christopher Mims:


just about every site that hosts user-generated content is carefully weighing the costs and benefits of updating their content moderation systems, using a mix of human professionals, algorithms and users. Some are even building rules into their services to pre-empt the need for increasingly costly moderation.

The saga of gaming-focused messaging app Discord is instructive: In 2018, the service, which is aimed at children and young adults, was one of those used to plan the Charlottesville riots. A year later, the site was still taking what appeared to be a deliberately laissez-faire approach to content moderation.

By this January, however, spurred by reports of hate speech and lurking child predators, Discord had done a complete 180. It now has a team of machine-learning engineers building systems to scan the service for unacceptable uses, and has assigned 15% of its overall staff to trust and safety issues.

This newfound attention to content moderation helped keep Discord away from the controversy surrounding the Capitol riot, and caused it to briefly ban a chat group associated with WallStreetBets during the GameStop stock runup. Discord’s valuation doubled to $7bn over roughly the same period, a validation that investors have confidence in its moderation strategy.


As he points out, even small amounts of hate speech on a large enough platform can have a huge impact.
unique link to this extract

Covid pandemic: how Youyang Gu used AI and data to make most accurate prediction • Bloomberg

Ashlee Vance:


With such intense interest around these forecasts, more models began to appear through the spring and summer of 2020. Nicholas Reich, an associate professor in the biostatistics and epidemiology department at the University of Massachusetts, Amherst, collected the 50 or so models and measured their accuracy over many months at the Covid-19 Forecast Hub. “Youyang [Gu]’s model was consistently among the top,” Reich says.

In November, Gu decided to wind down his death forecast operation. Reich had been blending the various forecasts and found that the most accurate predictions came from the this “ensemble model,” or combined data.

“Youyang stepped back with a remarkable sense of humility,” Reich says. “He saw the other models were doing well and his work here was done.” A month before stopping the project, Gu had predicted that the U.S. would record 231,000 deaths on Nov. 1. When Nov. 1 arrived, the U.S. reported 230,995 deaths.
The IHME’s Murray has his own take on Gu’s exit. He says Gu’s model would not have picked up on the seasonal nature of the coronavirus and would have missed the winter surge in cases and deaths. “He had the epidemic going away in the winter, and we had picked up that there was seasonality as early as May,” Murray says.

The machine learning methods used by Gu work well at short-range predictions, Murray says, but “are not very good at understanding what is going on” in the bigger picture.


Certain amount of sour grapes there from Murray, whose forecasting system at the IHME was outclassed by Gu’s. But there’s also an element of self-fulfilling prophecy: with so many people attempting to create their own forecast, someone had to win the jackpot. Gu managed it, but if it hadn’t been him, someone else would have.
unique link to this extract

AI may mistake chess discussions as racist talk • Carnegie Mellon University


“The Queen’s Gambit,” the recent TV mini-series about a chess master, may have stirred increased interest in chess, but a word to the wise: social media talk about game-piece colors could lead to misunderstandings, at least for hate-speech detection software.

That’s what a pair of Carnegie Mellon University researchers suspect happened to Antonio Radić, or “agadmator,” a Croatian chess player who hosts a popular YouTube channel. Last June, his account was blocked for “harmful and dangerous” content.

YouTube never provided an explanation and reinstated the channel within 24 hours, said Ashique R. KhudaBukhsh a project scientist in CMU’s Language Technologies Institute (LTI). It’s nevertheless possible that “black vs. white” talk during Radić’s interview with Grandmaster Hikaru Nakamura triggered software that automatically detects racist language, he suggested.

“We don’t know what tools YouTube uses, but if they rely on artificial intelligence to detect racist language, this kind of accident can happen,” KhudaBukhsh said. And if it happened publicly to someone as high-profile as Radić, it may well be happening quietly to lots of other people who are not so well known.


Black, white, attack, threat – to an AI, that sounds like something bad. But it’s that last sentence that matters: that it’s probably happening to other people, and they won’t have any clue why either, while probably having less recourse or leverage to get reinstated.
unique link to this extract

Joel Kaplan’s policy team sways big Facebook decisions like Alex Jones ban • Buzzfeed News

Ryan Mac and Craig Silverman:


The strategic response team that had gathered evidence for the Alex Jones and Infowars ban in spring 2019 drew upon years of examples of his hate speech against Muslims, transgender people, and other groups. Under the company’s policies for dangerous individuals and organizations, Jones and Infowars would be permanently banned and Facebook would have to remove content that expressed support for the conspiracy theorist and his site.

In April 2019, a proposal for the recommended ban — complete with examples and comments from the public policy, legal, and communications teams — was sent by email to Monika Bickert, Facebook’s head of global policy management, and her boss, Kaplan. The proposal was then passed on to top company leadership, including Zuckerberg, sources said.

The Facebook CEO balked at removing posts that praised Jones and his ideas.

“Zuckerberg basically took the decision that he did not want to use this policy against Jones because he did not personally think he was a hate figure,” said a former policy employee.

…IFR [In Feed Recommendations] was intended to foster new connections or interests. For example, if a person followed the Facebook page for a football team like the Kansas City Chiefs, IFR might add a post from the NFL to their feed, even if that person didn’t follow the NFL.

One thing IFR was not supposed to do was recommend political content. But earlier that spring, Facebook users began complaining that they were seeing posts from conservative personalities including Ben Shapiro in their News Feeds even though they had never engaged with that type of content.

When the issue was flagged internally, Facebook’s content policy team warned that removing such suggestions for political content could reduce those pages’ engagement and traffic, and possibly inspire complaints from publishers. A News Feed product manager and a policy team member reiterated this argument in an August post to Facebook’s internal message board.


Kaplan really does emerge as a malign influence over everything in the past ten years.
unique link to this extract

The bizarre reaction to Facebook’s decision to get out of the news business in Australia • Techdirt

Mike Masnick:


as any economist will tell you, taxing something doesn’t just bring in revenue, it decreases whatever you tax. This is why we have things like cigarette taxes and pollution taxes. It’s a tool to get less of something. So, in this case, Australia is saying it wants to tax links to news on Facebook, and Facebook responds in the exact way any reasonable economist would predict: it says that’s just not worth it and bans links. That’s not incompatible with democracy. It’s not bringing a country to its knees. The country said “this is how much news links cost” and Facebook said “oh, that’s too expensive, so we’ll stop.”

Contrary to the idea that this is an “attack” on journalism or news in Australia, it’s not. The news still exists in Australia. News companies still have websites. People can still visit those websites.
Indeed, the people who are saying that this move by Facebook is somehow an “attack” on news or an attack on Australian sovereignty seem to be admitting more than they’d really like: that they think Facebook must be a dominant source of news in the country.

I mean, if Facebook is really such a problem, shouldn’t they all be celebrating? This is Facebook saying “okay, okay, we’ll completely remove ourselves from the news business.” Since everyone was complaining that Facebook was too much of a presence in the news business… isn’t that… a victory?

And we haven’t even gotten to the other problematic part of the law — which is that it requires Facebook and Google to give newspapers heads up to algorithmic changes. This is completely disconnected from reality. Facebook and Google may make multiple algorithm changes every day, just to keep their services running. Having to tell newspapers (and them alone) about those changes with a few weeks notice is basically giving those news organizations the keys to the kingdom: it’s telling them how to game the algorithms.


You really can’t argue with his middle point: if Facebook is bad for news, then how can it getting out of news be bad, unless actually it’s good for news?
unique link to this extract

Griddy customers face $5,000 electric bills for five freezing days in Texas • Dallas News

Maria Halkias:


Some Texans are facing yet another crisis: how to pay enormous electric bills.

The Texas power supplier Griddy, which sells unusual plans with prices tied to the spot price of power on the Texas grid, warned its customers over the weekend that their bills would rise significantly during the storm and that they should switch providers.

Some quickly looked into doing that but found the actual changeover of service wouldn’t happen for days.

Now customers say they never dreamed they’d be billed in the four figures for five days of service.

Karen Cosby said her cost is $5,000 for usage since Saturday at her 2,700-square-foot house in Rockwall.

DeAndre Upshaw of Dallas said the electric bill for his 900-square-foot, two-story townhouse was also $5,000.

Other customers on social media expressed frustration with similar bills from Griddy, the power supplier that told its 29,000 customers on Saturday, after spot electricity prices soared, to quickly shift out of its network and find a new supplier.

Those spot prices hit $9,000 per megawatt-hour. That means $9 for a kilowatt-hour that usually costs Cosby around 7 cents, and sometimes as little as 2 cents.


Of course when they signed up, they’d never have expected a once-in-a-generation snowstorm to hit. But the US has 50 states. Going by 25-year generations, on average you’ll get a once-per-generation event at least every year somewhere. And with climate change forcing the issue, more often. The Texas governor and public utilities say they’re “taking steps” to help people with such bills.
unique link to this extract

Decarbonising America: Joe Biden’s climate-friendly energy revolution • The Economist


The vast majority of Republicans elected to federal office reject policies to cut emissions, which is why Congress has not seriously confronted the issue for more than a decade. The power of Republicans in the Senate made it pointless.

The problem is made worse by the fact that some conservative Democrats have their own reservations. Joe Manchin, a Democrat from West Virginia, says that he supports climate action. But he rejects the idea that coal, the dirtiest fossil fuel, might be permanently removed from the world’s energy portfolio: “Get into reality,” he says. “It’s not going to be eliminated.” The fact that the Senate is split 50-50 between the parties means that, even with Vice-President Kamala Harris’s casting vote, Mr Manchin in effect has a veto over legislation.

Should such obstacles lead to America punting for another decade, it will pay for the privilege. Delaying to 2030 would make the transition to a net-zero emissions economy almost twice as expensive as it would be if started today, with costs soaring to $750bn a year by 2035 and more than $900bn a year by the early 2040s, according to Energy Innovation, a policy group. But today’s urgency comes from greater concerns than fiscal prudence. America’s emissions are not only a problem for the climate in and of themselves. They are also a check on its opportunities to influence the rest of the world’s emissions, which copiously outweigh its own.


(Thanks Cam Ross for the link.)
unique link to this extract

Errata, corrigenda and ai no corrida: none notified

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.