The apostrophe apocalypse

It was immediately tempting to view the absence of apostrophes on new street signs in a North Yorkshire town as a real-life example of computer systems crushing human culture. Then, near-simultaneously, Apple launched an ad (which it now regrets) showing just that process, raising the temptation even more. But no.

In fact, as Brandon Vigliarolo writes at The Register, not only is the removal of apostrophes in place names not new in the UK, but it also long precedes computers. The US Board on Geographic Names declared apostrophes unwanted as long ago as its founding year, 1890, apparently to avoid implying possession. This decision by the BGN, which has only made five exceptions in its history, was later embedded in the US’s Geographic Names Information System and British Standard 7666. When computers arrived to power databases, the practice carried on.

All that said, it’s my experience that the older British generation are more resentful of American-derived changes to their traditional language than they are of computer-driven alterations (one such neighbor complains about “sidewalk”). So campaigns to reinstate missing apostrophes seem likely to persist.

Blaming computers seemed like a coherent narrative, not least because new technology often disrupts social customs. Railways brought standardized time, and the desire to simplify things for computers led to the 2023 decision to eliminate leap seconds in 2035 (after 18 years of debate). Instead, the apostrophe apocalypse is a more ordinary story of central administrators preferencing their own convenience over local culture and custom (which may itself be contested). It still seems like people should be allowed to keep their street signs. I mean.

***

Of course language changes over time and usage. The character limits imposed by texting (and therefore exTwitter and other microblogging sites) brought us many abbreviations that are now commonplace in daily life, just as long before that the telegraph’s cost per word spawned its own compressed dialect. A new example popped up recently in Charles Arthur’s The Overspill.

Arthur highlighted an article at Level Up Coding/Medium by Fareed Khan that offered ways to distinguish between human-written and machine-generated text. It turns out that chatbots use distinctively different words than we do. Khan was able to generate a list of about 100 words that may indicate a chatbot has been at work, as well as a web app that can check a block of text or a file in one go. The word “delve” was at the top.

I had missed Khan’s source material, an earlier claim by YCombinator founder Paul Graham that “delve” used in an email pitch is a clear sign of ChatGPT-generated text. At the Guardian, Alex Hern suggests that an underlying cause may be the fact that much of the labeling necessary to train the large language models that power chatbots is carried out by badly paid people in the global South – including Africa, where “delve” is more commonly used than in Western countries.

At the Premium Times, Chiamaka Okafor argues that therefore identifying “delve” as a marker of “robotic text” penalizes African writers. “We are losing sight of an opportunity to rewrite the AI narratives that exclude people in the global majority,” she writes. A reminder: these chatbots are just math and statistics predicting the next word. They will always regress to the mean. And now they’ll penalize us for being different.

***

Just two years ago, researchers fretted that we were running out of “high-quality text” on which to train large language models. We’ve been seeing the results since, as sites hosting user-generated content strike deals with LLM owners, leading to contentious disputes between those owners and sites’ users, who feel betrayed and ripped off. Reddit began by charging for access to its API, then made a deal with Google to use its database of posts for training for an injection of cash that enabled it to go public. Yesterday, Reddit announced a similar deal with OpenAI – and the stock went up. In reality, these deals are asset-stripping a site that has consistently lost money for 18 years.

The latest site to sell its users’ content is the technical site Stack Overflow, Developers who offer mutual aid by answering each other’s questions are exactly the user base you would expect to be most offended by the news that the site’s owner, the investment group Prosus, which bought the site in 2021 for $1.8 billion, has made a deal giving OpenAI access to all its content. And so it proved: developers promptly began altering or removing their posts to protest the deal. Shortly thereafter, the site’s moderators began restoring those posts and suspending the users.

There’s no way this ends well; Internet history’s many such stories never have. The site’s original owners, who created the culture, are gone. The new ones don’t care what users *believe* their rights are if the terms and conditions grant an irrevocable license to everything they post. Inertia makes it hard to build a replacement; alienation thins out the old site. As someone posted to Twitter a few years ago, “On the Internet your home always leaves you.”

‘Twas ever thus. And so it will be until people stop taking the bait in the first place.

Illustrations: Apple’s canceled “crusher” ad.

Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. She is a contributing editor for the Plutopia News Network podcast. Follow on Mastodon.

Performing intelligence

“Oh, great,” I thought when news broke of the release of GPT-4. “Higher-quality deception.”

Most of the Internet disagreed; having gone mad only a few weeks ago over ChatGPT, everyone’s now agog over this latest model. It passed all these tests!

One exception was the journalist Paris Marx, who commented on Twitter: “It’s so funny to me that the AI people think it’s impressive when their programs pass a test after being trained on all the answers.”

Agreed. It’s also so funny to me that they call that “AI” and don’t like it when researchers like computational linguist Emily Bender call it a “stochastic parrot”. At Marx’s Tech Won’t Save Us podcast, Goldsmith professor Dan McQuillan, author of Resisting AI: An Anti-fascist Approach to Artificial Intelligence, calls it a “bullshit engine” whose developers’ sole goal is plausibility – plausibility that, as Bender has said, allows us imaginative humans to think we detect a mind behind it, and the result is to risk devaluing humans.

Let’s walk back to an earlier type of system that has been widely deployed: benefits scoring systems. A couple of weeks ago, Lighthouse Reports and Wired magazine teamed up on an investigation of these systems, calling them “suspicion machines”.

Their work focuses on the welfare benefits system in use in Rotterdam between 2017 and 2021, which uses 315 variables to risk-score benefits recipients according to the likelihood that their claims are fraudulent. In detailed, worked case analyses, they find systemic discrimination: you lose points for being female, for being female and having children (males aren’t asked about children), for being non-white, and for ethnicity (knowing Dutch a requirement for welfare recipients). Other variables include missing meetings, age, and “lacks organizing skills”, which was just one of 54 variables based on case workers’ subjective assessments. Any comment a caseworker adds translates to a 1 added to the risk score, even if it’s positive. The top-scoring 10% are flagged for further investigation.

This is the system that Accenture, the city’s technology partner on the early versions, said at its unveiling in 2018 was an “ethical solution” and promised “unbiased citizen outcomes”. Instead, Wired says, the algorithm “fails the city’s own test of fairness”.

The project’s point wasn’t to pick on Rotterdam; of the dozens of cities they contacted it just happened to be the only one that was willing to share the code behind the algorithm, along with the list of variables, prior evaluations, and the data scientists’ handbook. It even – after being threatened with court action under freedom of information laws, shared the mathematical model itself.

The overall conclusion: the system was so inaccurate it was little better than random sampling “according to some metrics”.

What strikes me, aside from the details of this design, is the initial choice of scoring benefits recipients for risk of fraud. Why not score them for risk of missing out on help they’re entitled to? The UK government’s figures on benefits fraud indicate that in 2021-2022 overpayment (including error as well as fraud) amounted to 4%; and *underpayment* 1.2% of total expenditure. Underpayment is a lot less, but it’s still substantial (£2.6 billion). Yes, I know, the point of the scoring system is to save money, but the point of the *benefits* system is to help people who need it. The suspicion was always there, but the technology has altered the balance.

This was the point the writer Ellen Ullman noted in her 1996 book Close to the Machine”: the hard-edged nature of these systems and their ability to surveil people in new ways, “infect” their owners with suspicion even of people they’ve long trusted and even when the system itself was intended to be helpful. On a societal scale, these “suspicion machines” embed increased division in our infrastructure; in his book, McQuillan warns us to watch for “functionality that contributes to violent separations of ‘us and them’.”

Along those lines, it’s disturbing that Open AI, the owner of ChatGPT and GPT-4 (and several other generative AI gewgaws) has now decided to keep secret the details of its large language models. That is, we have no sight into what data was used in training, what software and hardware methods were used, or how energy-intensive it is. If there’s a machine loose in the world’s computer systems pretending to be human, shouldn’t we understand how it works? It would help with damping down imagining we see a mind in there.

The company’s argument appears to be that because these models could become harmful it’s bad to publish how they work because then bad actors will use them to create harm. In the cybersecurity field we call this “security by obscurity” and there is a general consensus that it does not work as a protection.

In a lengthy article at New York magazine, Elizabeth Weil. quotes Daniel Dennett’s assessment of these machines: “counterfeit people” that should be seen as the same sort of danger to our system as counterfeit money. Bender suggests that rather than trying to make fake people we should be focusing more on making tools to help people.

The thing that makes me tie it to the large language models that are producing GPT is that in both cases it’s all about mining our shared cultural history, with all its flaws and misjudgments, in response to a prompt and pretending the results have meaning and create new knowledge. And *that’s* what’s being embedded into the world’s infrastructure. Have we learned nothing from Clever Hans?

Illustrations: Clever Hans, performing in Leipzig in 1912 (by Karl Krali, via Wikimedia.

Wendy M. Grossman is the 2013 winner of the Enigma Award. Her Web site has an extensive archive of her books, articles, and music, and an archive of earlier columns in this series. Follow on Mastodon or Twitter.