Start Up No.1445: Pornhub’s rape problem, Google staff row over AI paper, drones do policing, TikTok delayed again, and more


A new filling station in Essex won’t offer petrol – just electrons. CC-licensed photo by JCT 600 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. Electable. I’m @charlesarthur on Twitter. Observations and links welcome.

The Children of Pornhub • The New York Times

Nicholas Kristof:

»

Pornhub prides itself on being the cheery, winking face of naughty, the website that buys a billboard in Times Square and provides snow plows to clear Boston streets. It donates to organizations fighting for racial equality and offers steamy content free to get people through Covid-19 shutdowns.

That supposedly “wholesome Pornhub” attracts 3.5 billion visits a month, more than Netflix, Yahoo or Amazon. Pornhub rakes in money from almost three billion ad impressions a day. One ranking lists Pornhub as the 10th-most-visited website in the world.

Yet there’s another side of the company: Its site is infested with rape videos. It monetizes child rapes, revenge pornography, spy cam videos of women showering, racist and misogynist content, and footage of women being asphyxiated in plastic bags. A search for “girls under18” (no space) or “14yo” leads in each case to more than 100,000 videos. Most aren’t of children being assaulted, but too many are.

After a 15-year-old girl went missing in Florida, her mother found her on Pornhub — in 58 sex videos. Sexual assaults on a 14-year-old California girl were posted on Pornhub and were reported to the authorities not by the company but by a classmate who saw the videos. In each case, offenders were arrested for the assaults, but Pornhub escaped responsibility for sharing the videos and profiting from them.

Pornhub is like YouTube in that it allows members of the public to post their own videos. A great majority of the 6.8 million new videos posted on the site each year probably involve consenting adults, but many depict child abuse and nonconsensual violence. Because it’s impossible to be sure whether a youth in a video is 14 or 18, neither Pornhub nor anyone else has a clear idea of how much content is illegal.

…The issue is not pornography but rape. Let’s agree that promoting assaults on children or on anyone without consent is unconscionable. The problem with Bill Cosby or Harvey Weinstein or Jeffrey Epstein was not the sex but the lack of consent — and so it is with Pornhub.

I came across many videos on Pornhub that were recordings of assaults on unconscious women and girls. The rapists would open the eyelids of the victims and touch their eyeballs to show that they were nonresponsive.

«

Kristof spent months on this article, which is as horrifying as you expect.
unique link to this extract


We read the paper that forced Timnit Gebru out of Google. Here’s what it says • MIT Technology Review

Karen Hao:

»

Gebru, a widely respected leader in AI ethics research, is known for coauthoring a groundbreaking paper that showed facial recognition to be less accurate at identifying women and people of color, which means its use can end up discriminating against them. She also cofounded the Black in AI affinity group, and champions diversity in the tech industry. The team she helped build at Google is one of the most diverse in AI, and includes many leading experts in their own right. Peers in the field envied it for producing critical work that often challenged mainstream AI practices.

…Online, many other leaders in the field of AI ethics are arguing that the company pushed her out because of the inconvenient truths that she was uncovering about a core line of its research—and perhaps its bottom line.

…Titled “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?” the paper lays out the risks of large language models—AIs trained on staggering amounts of text data. These have grown increasingly popular—and increasingly large—in the last three years. They are now extraordinarily good, under the right conditions, at producing what looks like convincing, meaningful new text—and sometimes at estimating meaning from language. But, says the introduction to the paper, “we ask whether enough thought has been put into the potential risks associated with developing them and strategies to mitigate these risks.”

…because the [language] training datasets are so large, it’s hard to audit them to check for… embedded biases. “A methodology that relies on datasets too large to document is therefore inherently risky,” the researchers conclude. “While documentation allows for potential accountability, […] undocumented training data perpetuates harm without recourse.”

The researchers summarize the third challenge as the risk of “misdirected research effort.” Though most AI researchers acknowledge that large language models don’t actually understand language and are merely excellent at manipulating it, Big Tech can make money from models that manipulate language more accurately, so it keeps investing in them. “This research effort brings with it an opportunity cost,” Gebru and her colleagues write. Not as much effort goes into working on AI models that might achieve understanding, or that achieve good results with smaller, more carefully curated datasets (and thus also use less energy).

The final problem with large language models, the researchers say, is that because they’re so good at mimicking real human language, it’s easy to use them to fool people. There have been a few high-profile cases, such as the college student who churned out AI-generated self-help and productivity advice on a blog, which went viral.

«

unique link to this extract


Google workers mobilize against firing of top Black female executive • NBC News

Olivia Solon and April Glaser:

»

Nearly 800 Google employees have joined a solidarity campaign in support of one of the company’s top female executives, a known advocate for diversity in the industry who said she was fired after what her boss described as a dispute over a research paper.

The executive, Timnit Gebru, technical co-lead of Google’s Ethical Artificial Intelligence Team, announced on Twitter late Wednesday that she had been fired after sending an email to co-workers stating that the company’s leadership had forced her to retract a paper focusing on ethical problems involving the kind of artificial intelligence systems used to understand human language, including one that powers Google’s search engine.

The email also detailed her frustration with the company’s efforts to create a more inclusive workspace. She said that she feels “constantly dehumanized” at Google.

Google would not provide comment on Gebru’s firing, but pointed to an email from Google’s head of research, Jeff Dean, to employees, published by the technology newsletter Platformer, in which he said that Gebru had resigned.

«

By the time the story had been published, the number in the campaign was over a thousand. There’s a significance between being fired (Gebru’s line) and resigning: with the latter, you don’t qualify for unemployment and similar benefits. So the difference matters, a lot.
unique link to this extract


How to become a best-selling author on Amazon in five minutes with three dollars • Quartz

Brent Underwood:

»

I would like to tell you about the biggest lie in book publishing. It appears in the biographies and social media profiles of almost every working “author” today. It’s the word “best seller.”

This isn’t about how The New York Times list is biased (though it is). This isn’t about how authors buy their way onto various national best-seller lists by buying their own books in bulk (though they do). No, this is about the far more insidious title of “Amazon Best Seller”—and how it’s complete and utter nonsense.

Here’s what happened in the book industry over the last few years: As Amazon has become the big dog in the book world, the “Amazon Best Seller” status has come to be synonymous with being an actual best seller. This is not true, and I can prove it.

A while ago, I put up a fake book on Amazon. I took a photo of my foot, uploaded it as a book to Amazon, and in a matter of hours had achieved “№1 Best Seller” status, complete with the orange banner and everything.

How many copies did I need to sell to be able to call up my mother and celebrate my newfound authorial achievements? Three. Yes, a total of three copies to become a best-selling author. And I bought two of those copies myself!

The reason people aspire to call themselves “best-selling author” is because it dramatically increases your credibility and “personal brand.” It can establish you as a thought leader. You’re able to show that you not only wrote a book, but that the market has judged it to be better than other books out there.

«

Though this was written in 2017, I doubt any of its truth has changed. There are so many categories that any ebook can become a best-seller (given a few friends) very quickly. The problem of too much, again.
unique link to this extract


Police drones are starting to think for themselves • The New York Times

Cade Metz:

»

Each day, the Chula Vista police respond to as many as 15 emergency calls with a drone, launching more than 4,100 flights since the program began two years ago. Chula Vista, a Southern California city with a population of 270,000, is the first in the country to adopt such a program, called Drone as First Responder.

Over the last several months, three other cities — two in California and one in Georgia — have followed suit. Police agencies from Hawaii to New York have used drones for years, but mostly in simple, manually flown ways. Officers would carry a drone in the trunk of a car on patrol or drive it to a crime scene before launching it over a park or flying it inside a building.

But the latest drone technology — mirroring technology that powers self-driving cars — has the power to transform everyday policing, just as it can transform package delivery, building inspections and military reconnaissance. Rather than spending tens of millions of dollars on large helicopters and pilots, even small police forces could operate tiny autonomous drones for a relative pittance.

That newfound automation, however, raises civil liberties concerns, especially as drones gain the power to track vehicles and people automatically. As the police use more drones, they could collect and store more video of life in the city, which could remove any expectation of privacy once you leave the home.

“Communities should ask hard questions about these programs. As the power and scope of this technology expands, so does the need for privacy protection,” said Jay Stanley, a senior policy analyst with the American Civil Liberties Union’s Project on Speech, Privacy and Technology. “Drones can be used to investigate known crimes. But they are also sensors that can generate offenses.”

«

I’m not sure I follow the logic in “generate offences”. Either you’re committing an offence or you’re not, and the presence of a drone won’t change that. The drones are essentially police dogs that can fly wearing a camera on their chest. I think we’d like them a lot more if they were represented like that.
unique link to this extract


The light of the charge brigade? • Status-Q

Quentin Stafford-Fraser:

»

The British county of Essex is often the butt of jokes here, since it has a few notably unappealing areas, but this is unfair. In general it’s a lovely county with some particularly pretty spots. Just at the moment, though, it has a different kind of jewel in its crown, at least from my point of view, because it’s also home to what looks like one of the coolest car-charging areas on the planet. If you want to see what the future of car travel might be, the place to go is probably the Gridserve Electric Forecourt near Braintree, which opens formally next week.

It has no fewer than 36 rapid chargers, and most of them are very rapid; there are a dozen that can supply 350kW (which almost nothing can actually consume, yet, but they’re future-proofing). 350kW, to give you an idea, would gain you about 25-30 miles of range for every minute you’re plugged in. There’s a bank of the Tesla v3 superchargers, too, which can do up to 250kW.

Now, you might well ask, how can you supply this quantity of electricity, even with that many solar panels? Well, the answer is that, as well as a good grid connection, they have an enormous battery pack next door and a solar farm just down the road. While you’re charging, there are cafes, loos and shops available.

I haven’t visited yet, but it just so happens I’ll be in that area next week, so I may well take a look.

Oh, and they’re hoping to build 100 of these.

«

This is absolutely what the electric car industry needs – though it also needs chargers to be in supermarket car parks, and in ordinary car parks, and in ordinary streets. Electric cars are still too hard to use reasonably for most people. (Via John Naughton.)
unique link to this extract


Denmark to end oil, gas extraction in North Sea • Associated Press

Jan Olsen:

»

The Danish Parliament voted late Thursday to end offshore gas and oil extraction, which had started in 1972 and made the country the largest producer in the European Union. Non EU-members Norway and Britain are larger producers, with a bigger presence in the North Sea.

Denmark is this year estimated to pump a bit over 100,000 barrels of crude oil and oil equivalents a day, according to the government.

That is relatively little in a global context. The U.K. produces about ten times that amount while the US, the world’s largest producer, pumped over 19 million barrels of oil a day last year. Environmental activists nevertheless said the move was significant as it shows the way forward in the fight against climate change.

Greenpeace called it “a landmark decision toward the necessary phase-out of fossil fuels.”

“This is a huge victory for the climate movement,” said Helene Hagel of Greenpeace Denmark. Wealthy Denmark has “a moral obligation to end the search for new oil to send a clear signal that the world can and must act to meet the Paris Agreement and mitigate the climate crisis.”

The 2015 landmark Paris climate deal asks both rich and poor countries to take action to curb the rise in global temperatures that is melting glaciers, raising sea levels and shifting rainfall patterns. It requires governments to present national plans to reduce emissions to limit global temperature rise to well below 2 degrees Celsius (3.6 degrees Fahrenheit).

Denmark has been an early adopter of wind power, with more than a third of its electricity production deriving from wind turbines. They are considered key in the transformation of the energy system and should enable Denmark to no longer be dependent on fossil fuels in 2050 for electricity production.

«

Might not be big in absolute terms, but as Greenpeace says, it’s a significant decision because it’s a whole country. To a large extent, this is how action has to be taken: at governmental level.
unique link to this extract


TikTok sale deadline on hold as talks with US continue • Bloomberg

Kurt Wagner, Shelly Banjo, and Jennifer Jacobs:

»

A deadline set by the Trump administration for the forced sale of TikTok’s US assets will come and go Friday without a final deal, according to people familiar with the discussions.

While the deadline has been extended multiple times, TikTok isn’t expected to receive a new one, said the people, who asked not to be identified because the decision isn’t yet public. TikTok is still in talks with the US government about a sale that satisfies the administration’s national security concerns, but Friday’s deadline will be allowed to lapse while the discussions continue.

The US Treasury Department told TikTok and Chinese parent company ByteDance Ltd. that they won’t face a fine or other punishment for missing the deadline because the sides are still negotiating. The deal, which has been in the works for months, is close to being finished, and the administration is eager to complete it before President-elect Joe Biden takes office on Jan. 20, according to one of the people.

«

The Trump admininstration has either lost interest (as Trump seems to have done in absolutely everything) or hopes to to create some sort of screwup that it will leave in fire in an envelope on the doorstep of the Biden administration.
unique link to this extract


Recovering passwords from pixelized screenshots • LinkedIn

Sipke Mellema:

»

Images can be obfuscated in many ways, which is generally referred to as blurring. Pixelization with box filters can be seen as a subset of blurring techniques. Most blurring algorithms tend to spread out pixels as they try to mimic natural blurs caused by shaky cameras or focusing issues.

These exist many deblurring tools for common tasks, such as sharpening blurry photographs. Unfortunately, the pixelated passwords I’m working with are only a couple of blocks in height, so there is nothing to sharpen.

Recent developments in AI have raised fancy headlines such as “Researchers Have Created a Tool That Can Perfectly Depixelate Faces”. However, the AI does no such thing. This recent PULSE algorithm is similar to Google’s RAISR algorithm from 2016. The AI generates faces that result in the same image when pixelized, but the face it recovers is not the original.

Algorithms such as PULSE seem new, but they stem from a long lineage of deblurring tools. M. W. Buie wrote a tool in 1994 (!) to generate ‘Plutos’, blur them, and match them with observed images.

In a widely known article from 2006, D. Venkatraman explains an algorithm for recovering a pixelized credit card number. The idea is simple: generate all credit card numbers, pixelize them, and compare the result to the pixelized number.

…If not enough information is available to properly smooth the image back together, the technique-of-choice is to pixelize similar data and check if it matches. This is also the basis for my algorithm for recovering passwords from screenshots.

«

And his tool for discovering passwords from blurred images is on Github. Better find a different way to blur them than box filters in future.
unique link to this extract


Errata, corrigenda and ai no corrida: none notified

2 thoughts on “Start Up No.1445: Pornhub’s rape problem, Google staff row over AI paper, drones do policing, TikTok delayed again, and more

Leave a Reply to Niall O'Mara Cancel reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google 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.