HAL 9000 from Stanley Kubrick's 2001: A Space Odyssey.

Artificial Authors & Blank Bands

“Finding my books on the Books3 data set was disappointing and disorienting: writing is how I’ve made my life, artistically, and—this is important—practically too. . . . Books and writing are how I pay my mortgage, my children’s tuition, my grocery bill. To see my work so cavalierly stolen and used, without my consent, by corporations eager only to increase their own profits, is frankly terrifying.”—Elisabeth de Mariaffi, in The Walrus

Books3, if you’re not familiar with it, is a dataset of books—thousands of them (as in around 183,000)—that were downloaded from pirated sources—so the authors received nothing for their work—and then used to train the AI language models of several companies, including Meta and Bloomberg.

Odds are, you’ve not heard of de Mariaffi.

Odds are, you have heard of Mark Zuckerberg and Mike Bloomberg.

Bloomberg is estimated to be worth $96 billion. Zuckerberg? About $115 billion.

Neither probably thinks about making their mortgage payments or the size of the grocery bill.

There are lawsuits against Books3 by authors and other interested parties.

There are lawsuits against OpenAI for illegally using authors’ works. There are some more famous writers—Jodi Picoult, George R.R. Martin, George Saunders, John Grisham, Jonathan Franzen—involved in suits, as are some, well, outliers, like Mike Huckabee and Sarah Silverman.

While the name brands probably aren’t too concerned about the price of a gallon of milk, what is notable about these undertakings is that these people are trying to protect their work from the potential unfair reuse of manipulated variants thereof that would lead to increased corporate profitability and no benefit redounding to them.

Think about it: Books3, used by humongous corporations, didn’t even plunk down $20 for a copy of The Firm.

There was another lawsuit filed last week against Anthropic, which describes itself as “an AI safety and research company.” (Although not synonymous with books the way it once was, it is interesting to note that in September Anthropic announced Amazon would be investing up to $4 billion in the company.)

Anthropic developed “Claude,” which it describes as “A next-generation AI assistant for your tasks, no matter what the scale.”

Claude, in effect, is being sued because, as the suit lays out, the AI models known as “Claude” are built “by scraping and ingesting massive amounts of text from the internet and potentially other sources, and then using that vast corpus to train its AI models and generate output based on this copied text. Included in the text that Anthropic copies to fuel its AI models are the lyrics to innumerable musical compositions.”

Yes, we have moved from novels to albums, yet the AI challenge remains the same.

The suit goes on to note:

“When a user prompts Anthropic’s Claude AI chatbot to provide the lyrics to songs such as “A Change Is Gonna Come,” “God Only Knows,” “What a Wonderful World,” “Gimme Shelter,” “American Pie,” “Sweet Home Alabama,” “Every Breath You Take,” “Life Is a Highway,” “Somewhere Only We Know,” “Halo,” “Moves Like Jagger,” “Uptown Funk,” or any other number of  __________ musical compositions, the chatbot will provide responses that contain all or significant portions of those lyrics.”

What’s more:

“Claude responds to a whole range of prompts that do not seek __________ lyrics—such as requests to write a song about a certain topic, provide chord progressions for a given musical composition, or write poetry or short fiction in the style of a certain artist or songwriter—by generating output that nevertheless copies ____________ lyrics.”

The blanks, of course, are not in the filing.

And the word that is absent is not “Artists’” but “Publishers’.” The suit, filed in U.S. District Court in Nashville, is being brought by not by the musical analogue of George R.R. Martin or Sarah Silverman, but Universal Music Group (UMG).

What’s interesting about this is that UMG is not anti-AI.

For example, on May 23, the company put out a press release that opens:

“AI sound wellness company, Endel, and Universal Music Group (UMG), the world leader in music-based entertainment, announced a first of its kind strategic relationship to enable artists and labels to create soundscapes for daily activities like sleep, relaxation, and focus by harnessing the power of AI.”

Then on October 18, strangely coincident with the filing of the suit:

“Universal Music Group (UMG), the world leader in music-based entertainment, and BandLab Technologies, the parent company of the world’s largest social music creation platform, BandLab, today announced plans for an expansive, industry-first strategic relationship concentrated on artificial intelligence (AI).”

It’s not that UMG is against AI. It is for it when it provides benefits to it.

AI is not a bad thing in and of itself, certainly. Give an axe to a Girl Scout in the woods and you’re likely to end up with a nice campfire. Given an axe to a Lizzie Borden* wannabe and things won’t be so propitious.

But here’s the thing to consider.

Streaming is where the money in music is coming from. According to the RIAA, it accounted for 84% of the music revenues in the first half of 2023, $7.0 billion.

So if we look at the numbers for Spotify, we find that there are approximately 11 million musicians on the platform, of which 1,060 make more than $1 million on the platform, or about 9.6%. However, it is estimated that 99.3% make less than $10,000. So that would be some 10,923,000 who are not making much individually.

Say, however, that you were able to aggregate a number of those $10,000 checks. At some point it would really add up to real money, even though individually it is below minimum wage.

World leaders in music entertainment achieve that status primarily though the efforts of brand-name acts. Those acts are, for them, to be supported and nurtured.

But there is a vast number of musicians who are only known on a limited basis.

Isn’t it conceivable that someone could generate a vast number of releases using AI that wouldn’t compete with the Swifts and Drakes of the world, but with an assortment of sounds that resemble those who are less widely recognized? A sufficient number of releases from Blank Bands could create a small but steady revenue stream.

Which brings me back to Ms. de Mariaffi’s concern regarding “corporations eager only to increase their own profits.”

No one who is going to create a business model based on Artificial Authors or Blank Bands is likely to want to risk the wrath of a Grisham or a Gaga. It is the vast array of writers and musicians who are trying to earn a living who are potentially at risk of economic collapse.


* “Lizzie Borden took an axe/And gave her mother forty whacks/When she saw what she had done/She gave her father forty-one.”

Leave a Reply

Your email address will not be published. Required fields are marked *