Month: September 2024

A.I. and Art ⇥ newyorker.com

Ted Chiang, the New Yorker:

It’s harder to imagine a program that, over many sessions, helps you write a good novel. This hypothetical writing program might require you to enter a hundred thousand words of prompts in order for it to generate an entirely different hundred thousand words that make up the novel you’re envisioning. It’s not clear to me what such a program would look like. Theoretically, if such a program existed, the user could perhaps deserve to be called the author. But, again, I don’t think companies like OpenAI want to create versions of ChatGPT that require just as much effort from users as writing a novel from scratch. The selling point of generative A.I. is that these programs generate vastly more than you put into them, and that is precisely what prevents them from being effective tools for artists.

Matt Muir, writer of the excellent Web Curios newsletter:

[…] Broadly speaking I agree with some of the points he makes, specifically about the requirement for art to have an element of intentionality which is necessarily absent from anything made by (current generative) AI being as all it is is maths, and maths cannot have intent. Equally, though, Chiang concedes that artists have made, are making, and will continue to make, work *in conjunction with* non-intentional systems, and that these works are perfectly capable of being considered as ‘art’. […]

Adi Robertson on Bluesky:

I can hazard lots of guesses why, but it’s striking that virtually none of the “can AI do art” conversation focuses on the most interesting examples I’ve seen, in which the interactive conversation between user and machine — rather than the end output — *is* the art.

Robertson points to the Are You the Asshole bot and the Hey Robot game as two examples, both of which are creative explorations of human–A.I. interaction. Whether those conversations are considered “art” is something I will leave others to decide because I spent a bachelor’s degree hearing hundreds of people asking that question and I lost my patience for it.

Robertson’s observation is a spiritual successor to my issue with Instagram bait art installations: neither are necessarily cheapening art, but I wish artists treated social media and, now, A.I. with less formalism and more conceptualism. Artists can eke compelling works out of any medium. In fact, the very suspicion of A.I.’s involvement in art seems likely to lend itself to surprising and moving works, with suitably talented artists.

Grain2Pixel, a Photoshop Script for Converting Film Negative Scans ⇥ grain2pixel.com

I got a batch of film scans from the developer today and realized I needed a better process for converting them — better, that is, than the way I had been doing it, which was to flip the curves in Lightroom and then do all my corrections in reverse.

I played around with the Filmomat SmartConvert demo but I did not like the workflow enough to consider paying for it. I really like the results I got from Negative Lab Pro and I think the USD $99 price tag is reasonable. However, its main selling point — that DNG scans remain in DNG format — is also its drawback: your workflow is still going to be the reverse of what you expect because, under the hood, the image is still a negative.

That pushed me to trying Grain2Pixel which, from a getting-started perspective, is more cumbersome than the other two options, particularly as MacOS is alarmed you are trying to use unsigned software.

But once you get that sorted out and install the script, it makes quick work out of batch processing a folder of DNGs into TIFF images. Then you can import them into Lightroom and make corrections in positive colour. You do not need to worry too much about a loss of range — TIFF is plenty flexible in post, at least for my amateur purposes. I am very happy with the resulting images.

The Surprisingly Big Business of Library Ebooks ⇥ newyorker.com

Danielle Deschamps, in the conclusion to a rather interesting chapter from “Contemporary Issues in Collection Management”, hosted by Open Education Alberta:

Ebook licensing agreements have become the widespread norm for library ebook access. Yet, between libraries and publishers, these agreements, the terms of which are set by publishers, have devolved to an extent that libraries are struggling to maintain their access to ecollections. Publishers perceive libraries as harming their bottom lines and libraries are in a particularly vulnerable place, without much negotiating power. However, there are several optional ways for public libraries to move forward, in effort of balancing their financial capacity while maintaining their ethical principle of respecting intellectual property rights. […]

The subsequent chapter specifically about ebook pricing is also a terrific primer.

Daniel A. Gross, writing in September 2021, in the New Yorker:

To illustrate the economics of e-book lending, the N.Y.P.L. sent me its January, 2021, figures for “A Promised Land,” the memoir by Barack Obama that had been published a few months earlier by Penguin Random House. At that point, the library system had purchased three hundred and ten perpetual audiobook licenses at ninety-five dollars each, for a total of $29,450, and had bought six hundred and thirty-nine one- and two-year licenses for the e-book, for a total of $22,512. Taken together, these digital rights cost about as much as three thousand copies of the consumer e-book, which sells for about eighteen dollars per copy. As of August, 2021, the library has spent less than ten thousand dollars on two hundred and twenty-six copies of the hardcover edition, which has a list price of forty-five dollars but sells for $23.23 on Amazon. A few thousand people had checked out digital copies in the book’s first three months, and thousands more were on the waiting list. (Several librarians told me that they monitor hold requests, including for books that have not yet been released, to decide how many licenses to acquire.)

If you want to know why publishers so aggressively fought the Internet Archive on its model of lending out scanned copies of physical books, this is the reason. Publishers have created a model which fundamentally upsets a library’s ability to function. There is no scarcity in bytes, so publishers have created a way to charge more for something limitless, weightless, with nearly no storage costs.

What the Internet Archive did was perhaps a legal long-shot, and I worry about the effects of this suit and the one over shellac 78s. But it is hard not to see publishers as the real villains in this mess. They are consolidating power and charging even legitimate libraries unreasonable amounts of money for electronic copies of books which the publishers and their intermediaries ultimately still control.

RationalWiki Is Being Sued ⇥ rationalwiki.org

At the beginning of August, Nassim Haramein sued RationalWiki on charges of defamation, conspiracy, and invasion of privacy. Regardless of the merits of the suit — I write, trying not to fall afoul of an obviously litigious individual — RationalWiki is a small, volunteer-run operation and will need legal representation to avoid losing next week by default. The site is currently soliciting donations.

I think the world is better for having RationalWiki in it. If you have the means and would like to chip in, I am sure the administrators there would appreciate it.

Update: RationalWiki has been SLAPP-ed into settling. Donations will go toward a proper legal fund.

Here Are Over Twenty Examples of Google Employees Trying to Avoid Creating Antitrust Evidence ⇥ 404media.co

Jason Koebler, 404 Media:

The chats show 22 instances in which one Google employee told another Google employee to turn chat history off. In total, the court has dozens of specific employees who have told others to turn history off in DMs or broader group chats and channels. The document includes exchanges like this (each exchange includes different employees) […]

These examples are equal parts amusing, blatant, and telling. I doubt this is isolated; there are probably similar policies standard at other companies. But apparently this was part of Google’s new employee training.

The Economist: Big, Heavy Cars Are Killing More People ⇥ economist.com

The Economist:

So how big is too big? At what point do the costs of the heaviest vehicles — measured in lives lost — vastly exceed their benefits? To answer this question, The Economist compiled ten years’ worth of crash data from more than a dozen states. Like the data compiled by Messrs Anderson and Auffhammer, our figures come from reports filed by police officers, who are tasked with recording information about car crashes when called to the scene. Although all states collect such data, we focus on those that collect the most detailed figures and share them with researchers. The resulting dataset, which covers more than a third of America’s population, provides us with a sample that is both big and representative.

The results? According to the Economist, “if the heaviest tenth of vehicles in America’s fleet were downsized […] road fatalities in multi-car crashes — which totaled 19,081 in 2023 — could be reduced by 12%, or 2,300, without sacrificing the safety of any cars involved”.

Andre Mayer and Emily Chung, reporting for CBC News in June:

But the ubiquity of SUVs and trucks isn’t an accurate reflection of what people want to drive, say industry analysts.

The trend has been greatly influenced by a combination of savvy marketing, government regulations that incentivize bigger vehicles and limited supply of more modest ones.

Indeed, much of it is driven by one simple economic fact.

“Smaller cars are less profitable,” said Stephanie Brinley, associate director at U.S.-based transportation consultancy S&P Global Mobility.

People are guided to purchase an SUV or truck in the United States and Canada because most cities oblige us to own a vehicle of some kind, but inexpensive cars are not generally available, and other people drive oversized SUVs and trucks which makes us scared of driving anything smaller. Repeat until around 80% of new vehicle sales are various kinds of SUVs and trucks.

This forced market is dangerous for everybody except for those who are inside a large SUV or truck. It means headlights from oncoming traffic at eye level. It means small roads are less navigable and parking spaces need to be made larger. It means roads feel more dangerous so fewer people feel comfortable walking or cycling. It means more people are seriously injured and die. All because these vehicles are more profitable, many cities are inaccessible by other means, automakers have artificially constrained their wares, and people feel roads are competitive instead of cooperative.

Brazil Bans X ⇥ nucleo.jor.br

Sérgio Spagnuolo, Sofia Schurig, and Pedro Nakamura, Núcleo:

A Supreme Court Justice ordered, on Friday (August 30, 2024), the complete suspension of all access to X (formerly Twitter) across the entire Brazilian territory, in an unprecedented ruling against the social platform.

[…]

In a ruling issued on the afternoon of Aug. 31, Justice Alexandre de Moraes ordered the president of Brazil’s telecom regulator, Anatel, Carlos Manuel Baigorri, to ensure that necessary measures are taken and that internet companies are notified to block the application within 24 hours.

An un-bylined report from Al Jazeera:

At the core of the dispute, de Moraes argues that Musk refused earlier this year to block accounts responsible for the spread of fake news, hate speech and attacks on the rule of law.

At the time, Musk denounced the order as censorship and responded by closing the company’s offices in Brazil while ensuring the platform was still available in the country.

Mike Masnick, Techdirt:

And, of course, as a reminder, before Elon took over Twitter (but while he was in a legal fight about it), he accused the company of violating the agreement because of its legal fight against the Modi government over their censorship demands. I know it’s long forgotten now, but one of the excuses Elon used in trying to kill the Twitter deal was that the company was fighting too hard to protect free speech in India.

And then, once he took over, he not only caved immediately to Modi’s demands, he agreed to block the content that the Modi government ordered blocked globally, not just in India.

So Elon isn’t even consistent on this point. He folds to governments when he likes the leadership and fights them when he doesn’t. It’s not a principled stance. It’s a cynical, opportunistic one.

This is being compared by some to the arrest of Pavel Durov but, again, I am not sure I see direct parallels. This Brazilian law seems, from my Canadian perspective, more onerous and restrictive than those from most other liberal democracies. But I do not know much of anything about Brazilian policy, and perhaps this is in line with local expectations.

This is probably not the reason Bluesky wanted for growing by two million new users in one week.

Elon Musk Is an Influential Idiot Who Should Not Be Arrested for Being Dumb ⇥ theguardian.com

Robert Reich, former U.S. Secretary of Labor for the Clinton administration and Sam Reich’s dad, wrote about Elon Musk’s political influence in an editorial for the Guardian. It begins as a decent piece, comparing the power of owning a social media platform with Musk’s childlike gullibility — my words, not Reich’s. But, in a section of ideas about what to do, one suggestion seems particularly harmful:

3. Regulators around the world should threaten Musk with arrest if he doesn’t stop disseminating lies and hate on X.

Global regulators may be on the way to doing this, as evidenced by the 24 August arrest in France of Pavel Durov, who founded the online communications tool Telegram, which French authorities have found complicit in hate crimes and disinformation. Like Musk, Durov has styled himself as a free speech absolutist.

There are places where U.S.-style interpretation of free expression is contradicted by local laws and, so, X’s operations must comply. Maybe Musk could be legally responsible in some jurisdiction for things he has said, or for things hosted on a platform he owns. But we should almost never encourage the idea of arresting people for things they say. Yes, there are limits: threats of violence and fraud are both types of generally illegal speech. Yet charging Musk for being a loud public idiot is a very bad idea.

Also, while details about Pavel Durov’s arrest are still solidifying, it does not yet appear he is being held responsible for “hate crimes and disinformation”. According to French prosecutors (PDF), which I translated with DeepL, his charges are mostly about failing to comply with subpoenas and other legitimate legal demands. If X follows legal avenues for either complying with or disputing government demands, then I do not see how Durov’s arrest is even relevant. And, for what it is worth, neither Durov nor Telegram have been “found complicit” in anything. The United States is not the only country which has legal procedures.

In response to Reich’s article, a troll X account posted a screenshot of a 4chan post about “low T men”, itself containing an arguably antisemitic meme, which was quoted by Musk calling it an “interesting observation”. Just more evidence Musk is a big, dumb, rich, influential moron.

Third Circuit’s Section 230 TikTok Ruling ⇥ techdirt.com

Maryclaire Dale, Associated Press:

A U.S. appeals court revived on Tuesday a lawsuit filed by the mother of a 10-year-old Pennsylvania girl who died attempting a viral challenge she allegedly saw on TikTok that dared people to choke themselves until they lost consciousness.

While federal law generally protects online publishers from liability for content posted by others, the court said TikTok could potentially be found liable for promoting the content or using an algorithm to steer it to children.

Notably, the “Blackout Challenge” or the “Choking Game” is one of few internet challenges for teenagers which is neither a media-boosted fiction nor relatively harmless. It has been circulating for decades, and was connected with 82 deaths in the United States alone between 1995–2007. Which, yes, is before TikTok or even social media as we know it today. Melissa Chan reported in a 2018 Time article that its origins go back to at least the 1930s.

Mike Masnick, of Techdirt, not only points out the extensive Section 230 precedent ignored by the Third Circuit in its decision, he also highlights the legal limits of publisher responsibility:

We have some caselaw on this kind of thing even outside of the internet context. In Winter v. GP Putnam’s Sons, it was found that the publisher of an encyclopedia of mushrooms was not liable for “mushroom enthusiasts who became severely ill from picking and eating mushrooms after relying on information” in the book. The information turned out to be wrong, but the court held that the publisher could not be held liable for those harms because it had no duty to carefully investigate each entry.

Matt Stoller, on the other hand, celebrates the Third Circuit’s ruling as an end to “big tech’s free ride on Section 230”:

Because TikTok’s “algorithm curates and recommends a tailored compilation of videos for a user’s FYP based on a variety of factors, including the user’s age and other demographics, online interactions, and other metadata,” it becomes TikTok’s own speech. And now TikTok has to answer for it in court. Basically, the court ruled that when a company is choosing what to show kids and elderly parents, and seeks to keep them addicted to sell more ads, they can’t pretend it’s everyone else’s fault when the inevitable horrible thing happens.

And that’s a huge rollback of Section 230.

On a legal level, both Masnick and Stoller agree the Third Circuit’s ruling creates a massive change in U.S. internet policy and, because of current structures, the world. But their interpretations of this are in vehement disagreement on whether this is a good thing. Masnick says it is not, and I am inclined to agree. Not only is there legal precedent on his side, there are plenty of very good reasons for why Section 230 is important to preserve more-or-less the way it has existed for decades.

However, it seems unethical for TikTok to have no culpability for how users’ dangerous posts are recommended, especially to children. Perhaps legal recourse is wrong in this case and others like it, yet it just feels wrong for this case to eventually — after appeals and escalation to, probably, the Supreme Court — be summarily dismissed on the grounds that corporations have little responsibility or care for automated recommendations. There is a real difference between teenagers spreading this challenge one-on-one for decades and teenagers broadcasting it — or, at least, there ought to be a difference.