Facebook’s algorithm is suggesting Holocaust denial groups when people search on the term. CC-licensed photo by Daniel_Sadono 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 11 links for you. Reboot! I’m @charlesarthur on Twitter. Observations and links welcome.
Facebook’s algorithm “actively promotes” Holocaust denial content according to an analysis that will increase pressure on the social media giant to remove antisemitic content relating to the Nazi genocide.
An investigation by the Institute for Strategic Dialogue (ISD), a UK-based counter-extremist organisation, found that typing “holocaust” in the Facebook search function brought up suggestions for denial pages, which in turn recommended links to publishers which sell revisionist and denial literature, as well as pages dedicated to the notorious British Holocaust denier David Irving.
The findings coincide with mounting international demands from Holocaust survivors to Facebook’s boss, Mark Zuckerberg, to remove such material from the site.
Last Wednesday Facebook announced it was banning conspiracy theories about Jewish people “controlling the world”. However, it has been unwilling to categorise Holocaust denial as a form of hate speech, a stance that ISD describe as a “conceptual blind spot”.
The ISD also discovered at least 36 Facebook groups with a combined 366,068 followers which are specifically dedicated to Holocaust denial or which host such content. Researchers found that when they followed public Facebook pages containing Holocaust denial content, Facebook recommended further similar content.
Recommending “similar” content has led to people literally being recruited to terrorist groups. No surprise that this would happen with Holocaust denial. That’s the content about which Mark Zuckerberg said in 2018 “I don’t believe that our platform should take that down because there are things that different people get wrong. I don’t think that they’re intentionally getting it wrong.”
unique link to this extract
Newley Purnell and Jeff Horwitz:
In Facebook posts and public appearances, Indian politician T. Raja Singh has said Rohingya Muslim immigrants should be shot, called Muslims traitors and threatened to raze mosques.
Facebook Inc. employees charged with policing the platform were watching. By March of this year, they concluded Mr. Singh not only had violated the company’s hate-speech rules but qualified as dangerous, a designation that takes into account a person’s off-platform activities, according to current and former Facebook employees familiar with the matter.
Given India’s history of communal violence and recent religious tensions, they argued, his rhetoric could lead to real-world violence, and he should be permanently banned from the company’s platforms world-wide, according to the current and former employees, a punishment that in the U.S. has been doled out to radio host Alex Jones, Nation of Islam leader Louis Farrakhan and numerous white supremacist organizations.
Yet Mr. Singh, a member of Indian Prime Minister Narendra Modi’s Hindu nationalist party, is still active on Facebook and Instagram, where he has hundreds of thousands of followers. The company’s top public-policy executive in the country, Ankhi Das, opposed applying the hate-speech rules to Mr. Singh and at least three other Hindu nationalist individuals and groups flagged internally for promoting or participating in violence, said the current and former employees.
Yes, Facebook put its thumb on the scale because of politics. Literally scores of people have died in India due to content posted on Facebook’s apps (including WhatsApp). And then it does nothing when it knows someone is rabble-rousing?
unique link to this extract
Mozilla signs fresh Google search deal worth mega-millions as 25% staff cut hits Servo, MDN, security teams • The Register
Mozilla has renewed its lucrative nine-figure deal with Google to ensure its search engine is the default in Firefox in the US and other parts of the world.
Within hours of the browser maker laying off a quarter of its staff this week, a well-placed source told The Register Moz had signed a three-year agreement with Google. On Thursday, a spokesperson for Mozilla confirmed the partnership had been renewed though declined to go into specific detail on the contract duration and sums of money involved.
However, our source told us Moz will likely pocket $400m to $450m a year between now and 2023 from the arrangement, citing internal discussions held earlier this year.
Saved by the bell. Even so, that’s three more years of charity during which it needs to figure out what it’s for. Google is surely overpaying for this, relative to Firefox’s absolute share (combined desktop and mobile). Though it might have some use in warding off antitrust complaints as Chrome becomes more and more dominant. (That’s the argument here, which I discovered after this.)
unique link to this extract
A college kid created a fake, AI-generated blog. It reached #1 on Hacker News • MIT Technology Review
At the start of the week, Liam Porr had only heard of GPT-3. By the end, the college student had used the AI model to produce an entirely fake blog under a fake name.
It was meant as a fun experiment. But then one of his posts reached the number-one spot on Hacker News. Few people noticed that his blog was completely AI-generated. Some even hit “Subscribe.”
While many have speculated about how GPT-3, the most powerful language-generating AI tool to date, could affect content production, this is one of the only known cases to illustrate the potential. What stood out most about the experience, says Porr, who studies computer science at the University of California, Berkeley: “It was super easy, actually, which was the scary part.”
GPT-3 is OpenAI’s latest and largest language AI model, which the San Francisco–based research lab began drip-feeding out in mid-July. In February of last year, OpenAI made headlines with GPT-2, an earlier version of the algorithm, which it announced it would withhold for fear it would be abused. The decision immediately sparked a backlash, as researchers accused the lab of pulling a stunt. By November, the lab had reversed position and released the model, saying it had detected “no strong evidence of misuse so far.”
…The trick to generating content without the need for much editing was understanding GPT-3’s strengths and weaknesses. “It’s quite good at making pretty language, and it’s not very good at being logical and rational,” says Porr. So he picked a popular blog category that doesn’t require rigorous logic: productivity and self-help.
Says as much about Hacker News (and the productivity and self-help category) as it does GPT-3.
unique link to this extract
With classic Hangouts on the way out, Google today has two video calling apps. However, that is one too many for the company, and sources familiar with the matter tell us that Google Duo will eventually be replaced by Meet.
This decision is the result of Google placing its consumer communication services — Duo, Messages, and Android’s Phone app — under the leadership of G Suite head Javier Soltero. After the unified team was made public in May, Soltero announced to employees that it does not make sense for Duo and Meet to coexist.
Following the rise of work from home and remote learning, Google has moved aggressively to make Meet a Zoom competitor. Like Duo, it’s now “free for everyone” to use and going after the same market.
With all the focus on Meet, the new messaging chief opted to have the service become Google’s one video calling service for both regular and enterprise customers. Internally, this is being described as a merger of the two services that is codenamed “Duet” — a portmanteau of Duo and Meet.
We’re told by sources that this new direction and the reduced interest in building a dedicated consumer service came as a surprise to the Duo team.
Oh man. Nothing should come as a surprise to people working on any of Google’s messaging apps. Any time there are more than two (video and text) there’s a risk of one getting wiped out.
unique link to this extract
An Oxford-based solar technology firm hopes by the end of the year to begin manufacturing the world’s most efficient solar panels, and become the first to sell them to the public within the next year.
Oxford PV claims that the next-generation solar panels will be able to generate almost a third more electricity than traditional silicon-based solar panels by coating the panels with a thin layer of a crystal material called perovskite.
The breakthrough would offer the first major step-change in solar power generation since the technology emerged in the 1950s, and could play a major role in helping to tackle the climate crisis by increasing clean energy.
By coating a traditional solar power cell with perovskite a solar panel can increase its power generation, and lower the overall costs of the clean electricity, because the crystal is able to absorb different parts of the solar spectrum than traditional silicon.
Typically a silicon solar cell is able to convert up to about 22% of the available solar energy into electricity. But in June 2018, Oxford PV’s perovskite-on-silicon solar cell surpassed the best performing silicon-only solar cell by reaching a new world record of 27.3%.
That’s a 24% improvement. Substantial. And usually these things are “five years from now…” This is a lot quicker than that. Although there do seem to be questions about the longevity of the perovskite layer.
Tim Harford (with a non-paywalled article):
Instead of storing each frame separately, video compression algorithms will start with the first frame of a scene and then store a series of “diffs” — changes from one frame to the next. A slow, contemplative movie with long scenes and fixed cameras can be compressed more than a fast-moving action flick.
Similarly, a week full of new experiences will seem longer in retrospect. A month of repeating the same routine might seem endless, but will be barely a blip in the memory: the “diffs” are not significant enough for the brain to bother with.
After months of working from home, I now realise that there was something incomplete about this account. New experiences are indeed important for planting a rich crop of memories. But, by itself, that is not enough. A new physical space seems to be important if our brains are to pay attention.
The Covid-19 lockdown, after all, was full of new experiences. Some were grim: I lost a friend to the disease; I smashed my face up in an accident; we had to wear masks and avoid physical contact and worry about where the next roll of toilet paper was coming from. Some were more positive: the discovery of new pleasures, the honing of new skills, the overcoming of new challenges.
But I doubt I am alone in finding that my memory of the lockdown months is rather thin. No matter how many new people or old friends you talk to on Zoom or Skype, they all start to smear together because the physical context is monotonous: the conversations take place while one sits in the same chair, in the same room, staring at the same computer screen.
Greenland’s ice sheet may have shrunk past the point of return, with the ice likely to melt away no matter how quickly the world reduces climate-warming emissions, new research suggests.
Scientists studied data on 234 glaciers across the Arctic territory spanning 34 years through 2018 and found that annual snowfall was no longer enough to replenish glaciers of the snow and ice being lost to summertime melting.
That melting is already causing global seas to rise about a millimeter on average per year. If all of Greenland’s ice goes, the water released would push sea levels up by an average of 6 meters — enough to swamp many coastal cities around the world. This process, however, would take decades.
“Greenland is going to be the canary in the coal mine, and the canary is already pretty much dead at this point,” said glaciologist Ian Howat at Ohio State University. He and his colleagues published the study Thursday in the Nature Communications Earth & Environment journal.
The Arctic has been warming at least twice as fast as the rest of the world for the last 30 years, an observation referred to as Arctic amplification. The polar sea ice hit its lowest extent for July in 40 years.
In June, as Trump’s poll numbers dropped nationwide, the Washington Post reported that his campaign advisers were losing hope for Michigan and Pennsylvania, and would focus on holding Wisconsin. “It’s baked into the cake that Trump will lose the state’s large metro areas in a landslide, while the suburbs have been fleeing him,” Ben Wikler, the head of the Wisconsin Democratic Party, told me. “Trump can’t win a second term unless he racks up enormous margins in rural Wisconsin.”
For [dairy farmer and 2016 Trump voter, after two terms voting for Obama] Volenec, Trump’s appeal vanished almost immediately. “If I had known the things I know about him now, I wouldn’t have voted for him,” he said, when I visited him at his farm in February. As Trump’s trade wars escalated, Volenec’s problems worsened. In March, 2018, Canada effectively cut off all dairy imports from the United States, and milk from Michigan that had previously been exported began flooding into Wisconsin’s processing plants.
The co-op where Volenec sent his milk for processing was now competing with cheap out-of-state milk, and put a cap on the amount that it would take from him. That week, Volenec heard about a meeting of the Wisconsin Farmers Union, a family-farm advocacy group, in nearby Dodgeville, to promote a version of supply management, a system used in Canada that sets a quota on the production of dairy, eggs, and poultry.
Designed, like the New Deal policies, to prevent overproduction and to guarantee farmers a stable income, the system relies on higher prices for Canadian consumers. Trump’s trade war with Canada is aimed at dismantling supply management, which has long been deplored by Republican politicians. John Boehner, the former Speaker of the House, called it “Soviet-style” agriculture. For Volenec, it was a revelation. “This was my first glimpse into a world where the dairy farmer is not subservient to The Market,” he wrote in an essay called “Groomed for Apocalypse.”
when Fortnite was predictably removed from both [Apple’s and Google’s mobile] platforms, Epic filed lawsuits against both companies, alleging “anti-competitive restraints and monopolistic practices” in the mobile app marketplace. That move came alongside a heavy-handed PR blitz, including a video asking players to “join the fight to stop 2020 from becoming ‘1984.’”
But through this entire public fight for “open mobile platforms,” as Epic puts it, there is one major set of closed platforms that the company seems happy to continue doing business with. We’re speaking, of course, about video game consoles.
Most if not all of the complaints Epic makes against Apple and Google seem to apply to Microsoft, Sony, and Nintendo in the console space as well. All three console makers also take a 30% cut of all microtransaction sales on their platforms, for example.
This DLC fee represents a big chunk of those console makers’ revenues, too. “Add-on content” was a full 41% of Sony’s Game and Network revenue in the latest completed fiscal quarter. Microsoft saw a 39-percent increase in gaming revenue the quarter after Fortnite was released, too, coyly attributing the bump to “third-party title strength.” And the Switch saw similar post-Fortnite digital revenue increases after Nintendo announced that fully half of all Switch owners had downloaded Fortnite.
On mobile platforms, Epic is calling the same kind of 30% fee “exorbitant” and says it wants to offer a more direct payment solution so it can “pass along the savings to players.” On consoles, though, Epic happily introduced a permanent 20% discount on all microtransaction purchases, despite there being no sign that the console makers have changed their fee structure.
The major console makers also all exercise full control over what games and apps can appear in their own walled gardens. When it comes to iOS, Epic says that “by blocking consumer choice in software installation, Apple has created a problem so they can profit from the solution.” When it comes to consoles, Epic is silent about the same state of affairs.
Epic is being terribly hypocritical here, though as Orland points out, it’s a bit odd how it’s silent when the companies have investments in it. Epic’s CEO says that the 30% is permissible on games consoles because they’re sold below cost and there are “tremendous marketing campaigns”. Hmm.
unique link to this extract
Mark was suspicious about the BBC report last week on thousands of people dying from social media dis/misinformation about Covid-19:
Here is the relevant part of that article:
A popular myth that consumption of highly concentrated alcohol could disinfect the body and kill the virus was circulating in different parts of the world. Following this misinformation, approximately 800 people have died, whereas 5,876 have been hospitalized and 60 have developed complete blindness after drinking methanol as a cure of coronavirus.
That 5,876 figure is the origin of the 5,800 figure in the news reports. But note, it’s only the death toll from the one very specific piece of fake news: drinking methanol as a cure of coronavirus. Nor is it a figure about social media specifically. It’s a hospitalisation total for that myth, however it spread. Social media almost certainly played a big part in it spreading, but there’s no attempt to isolate its impact from other sources of spreading.
What’s more, go to the footnotes for this figure and they are all sources that refer to Iran. It looks like the figure is specifically for the one country, not a global one. For example, one of the cited sources is this Al Jazeera story: “Iran: Over 700 dead after drinking alcohol to cure coronavirus”.
So we have a figure for hospitalisations due to one myth in one country that has ended up being reported as a global figure for myths on social media specifically.
Errata, corrigenda and ai no corrida: none notified