Believe It Or Bot

Ed Sheeran was found not guilty of plagiarism, after a total waste of tax-payer money.

I’m very happy for him—it’s bad enough he bears a resemblance to Boris Johnson. Speaking of which, diluted in the sophomoric madness of the coronation, last week also marks Tory trouncing Thursday—Sunak’s conservatives lost 48 councils and over one thousand councilors—much of the blame for that should be laid at Boris’s door.

In the midst of these fait divers, the meteoric rise of artificial intelligence is my major concern at present, as we whistle past the graveyard on Horse Guards Parade.

This seismic paradigm shift is cruising through the human race, most of whom are either scrabbling for their next meal or their next Instagram post.

I find it so worrying—and so relevant to today’s society—that as I make my final revisions to The Hourglass in the next couple of months I will work some AI into the book, which already addresses topics like self-driving vehicles and Deep Fakes. The latter have become so mainstream that you can go online and swap your face into your favorite video.

Humans struggle to deal with anything non-linear, from falling in love to coping with death—the internet in general, and AI in particular, is no exception.

Internet servers around the world. The US has a stunning 149,313,079 machines (Google probably owns half of them)—that’s one computer for every two people. China, a putative rival, has only 1,100,841 machines—one for every 1400 people.

To understand how quickly AI can—and will—surge, a few stats about the internet are useful. Domain registration has boomed in the last couple of years—during the pandemic, the curve is vertical. At present, there are over six hundred million registered domains—at an average cost of thirty bucks per year, that’s twenty billion dollars of annual revenue.

AI would not be possible without the internet—as data folks are fond of saying, the only thing better than data is more data.

We recognize that growth in AI has been slow—painfully slow—over the last decades, but what we don’t talk about is that the explosive growth we see now is closely linked to the explosion of internet infrastructure, use, social media, and the Internet of Things (IoT).

The Large Language Model (LLM) neural networks that form the basis of ChatGPT and its various siblings emerged in 2018, very much a consequence of the availability of “large quantities of unlabeled text”, used to define “many parameters (typically billions of weights or more)”. In other words, the internet.

I asked ChatGPT about Ed Sheeran.

is the song Thinking Out Loud a copy of Let’s Get It On (ChatGPT is like putonghua, it doesn’t need punctuation).

There have been accusations that the song “Thinking Out Loud” by Ed Sheeran has similarities to “Let’s Get It On” by Marvin Gaye. In 2016, Ed Sheeran was sued for copyright infringement by the estate of Ed Townsend, who co-wrote “Let’s Get It On” with Marvin Gaye, and the case was settled out of court.

While the similarities between the two songs have been noted, it is ultimately up to a court to determine whether there was any unlawful copying or infringement of copyright. The settlement of the lawsuit suggests that there may have been some basis to the claims, but it is not a definitive conclusion.

In any case, it is not uncommon for songs to share similarities or be inspired by previous works. Music is a constantly evolving art form, and it is often the case that elements of earlier songs are incorporated into new compositions.

Jeepers (for it is he) gave me a crappy, plain-vanilla answer—not a fucking opinion!

But worse than that, the first paragraph correctly states the lawsuit was filed in 2016 but then ‘informs’ us it was settled out of court. Bollocks. And the second ends with ‘The settlement of the lawsuit suggests there may have been some basis to the claims…”

I can find no evidence of such a settlement, and the case concluded last week would suggest the matter was resolved in court.

In other words, Jeepers is a lyin’ lil’ bot. But we humans are educated to give the benefit of the doubt. So I asked old Jeeps.

“As an artificial intelligence language model, I am not capable of lying or telling the truth.”

Ah, now I like him, dis bot’s gittin’ phil-o-sophy. The reply is dangerously close to the paradox made famous by Epiminedes the Cretan: All Cretans Are liars. I think Jeepers is becoming Jeepistotle.

Some of the AI lies are extremely troublesome. This one was reported in the Washington Post.

One night last week, the law professor Jonathan Turley got a troubling email. As part of a research study, a fellow lawyer in California had asked the AI chatbot ChatGPT to generate a list of legal scholars who had sexually harassed someone. Turley’s name was on the list.

The chatbot, created by OpenAI, said Turley had made sexually suggestive comments and attempted to touch a student while on a class trip to Alaska, citing a March 2018 article in The Washington Post as the source of the information. The problem: No such article existed. There had never been a class trip to Alaska. And Turley said he’d never been accused of harassing a student.

My concern about AI rocketed when I heard an eminent AI researcher talk about the dangers of all this—resulting in an open letter signed by many concerned humans—well, 27565, so not that many.

Actually, you can add one to that, Peter Wibaux just signed up. I enjoy the irony of fighting AI with an artificial human: AH versus AI.

Will you join the fun?

The India Road, Atmos Fear, Clear Eyes, and Folk Tales For Future Dreamers. QR links for smartphones