Last week we wrote that the theft of copyrighted content for AI training data is becoming boringly predictable. We’re also previously written that we’ve had a vague notion that this whole AI saga is moving inexorably closer to a “Napster moment,” the point at which creators and other copyrighted owners say, “wait a damn moment, enough of this shit, this is stealing, plain and simple, and we’re not going to stand for it”—and then back that up with an army of lawyers.
In our latest chapter of the ongoing Move Fast and Take Things saga, 404 Media reports that “AI Music Generator Suno Admits It Was Trained on ‘Essentially All Music Files on the Internet’.” It’s kind of astonishing: Suno is admitting in court proceedings that it’s potentially stolen every single piece of music ever written that’s ever been available on the internet and not password protected.
What gives this particular case decidedly “jumping the shark” vibes is that Suno also has the balls to claim that it’s unfair that the record labels don’t allow them to scrape the copyrighted songs and so they were forced to steal them. To quote directly 404:
Suno also wrote that many labels’ licensing agreements do not allow for the training of AI on copyrighted songs, which the company says is an example of labels ‘trying to leverage their exclusive rights under copyright law to strong-arm music users into categorically avoiding artificial intelligence products.’ Suno has positioned itself as a company that is trying to fight record label monopolies to “empower many more people to create music” using AI tools.
We’re not lawyers and so we cannot argue the legal merits of this defense. But let’s just say, Suno admitting it did the illegal thing that it’s being sued for by the plaintiff doesn’t put them in the best position. Just saying.
In the broader context, the ongoing Suno lawsuit—along with the pending cases against OpenAI filed by the New York Times and Condé Nast sending a cease and desist letter to Perplexity accusing them of plagiarism—give us a bit of hope that AI companies are facing a degree of reckoning on the copyright front. They look like they’re going to have to start paying for the training data they need. Case in point: Perplexity this week offered to pay some amount of recompense (“cutting checks”) to publishers which may or may not be connected.
Which begs the question: How are they going to afford it? AI companies are already burning through mountains of cash with very little in the way of revenue to support it. Having to pay for the content many of them are currently stealing doesn’t seem to bode well for their business model.
Edward Zitron published a very good article earlier this week titled How Does OpenAI Survive. In it, he makes the argument that OpenAI’s business model simply doesn’t make sense. This particular paragraph hammers that point home with a sledgehammer:
Assuming everything exists in a vacuum, OpenAI needs at least $5 billion in new capital a year to survive. This would require it to raise more money than has ever been raised by any startup in history, possibly in perpetuity, which would in turn require it to access capital at a scale that I can find no comparable company to in business history.
As we said last week: The math ain’t mathing.
And there is a third significant factor in this formula of existential threat for the current form of AI companies: The product simply isn’t very good. In fact, as we’ve argued and written about before, it’s decidedly meh—and it’s prone to errors in ways that make the viability of the product untenable in its current form.
Earlier this week, Alex Health in The Verge reported on Meta’s apology after its AI chatbot denied that the shooting at the Trump rally ever happened. Also this week, The Information published a piece reporting that Meta had scrapped its celebrity AI chatbots after less than a year because they had “fallen flat with users.” (Aspects of that initiative live on, as Instagram has announced it will be rolling out “creator AIs and AI characters” over the coming weeks.)
It wasn’t just bad news for Meta. Business Insider reported that a CIO at a pharmaceutical company canceled Microsoft's CoPilot because, "after six months… the AI tools weren't good enough to be worth the extra money." This is particularly interesting when given the added context that this week Microsoft listed OpenAI as a competitor in an SEC filing, and Ed Ziton’s reporting of OpenAI’s need to raise large wads of cash this year and in perpetuity.
So does this mean that AI is about to vaporize itself in a plume of smoke? In a word, no. There is simply too much money involved and too much pressure to invest in AI for that to happen—at least right now.
But it does help to acknowledge that there are significant dominoes lining up that will fundamentally impact AI as a business, and by extension, AI as a product / feature / tool / disruptor, etc. The future of storytelling is as much about economics as it is technology, so while we’ll continue to check in on the business side, this is a good moment to swing the pendulum more towards creative endeavors while we let these critical moments percolate a bit. Again, we’re not lawyers—but we do love watching legal drama unfold.
What we’ve read
Copyright traps” could tell writers if an AI has scraped their work (MIT Technology Review) This is an interesting piece with regards to the intersection between AI and copyright law. A team from Imperial College London is developing “copyright traps,” digital watermarks in the form of specific text inserted within articles, essays, etc., that will reveal whether the work has been illegally ingested into an LLM. Per the article: “The team generated 100 trap sentences and then randomly chose one to inject into a text many times, de Montjoye explains. The trap could be injected into text in multiple ways—for example, as white text on a white background, or embedded in the article’s source code. This sentence had to be repeated in the text 100 to 1,000 times.” Clearly, there is some way to go before this “copyright trap” is subtle enough to escape the LLMs defenses but it’s going to be fascinating to watch this copyright arms race develop over the coming years. —James
Our first taste of ‘Apple Intelligence’ is here. Here’s what works. (The Washington Post) With the new iOS 18.1 developer beta, Apple is finally rolling out some Apple Intelligence features. To be sure, the big marquee tools like Genmoji and ChatGPT integration are still unavailable—much to my chagrin—but what you can use here is a Photos search that allows for more natural-language (e.g. "my dog in the snow"), a Siri that understands context, and a handful of “Writing Tools” that will help you send messages that are friendlier, more professional, or more concise. It's a bit of fun for now, but what's interesting to me in my limited testing is how seamless all these tools have been integrated into iOS so far—as it said at WWDC (and what is very clear here), Apple Intelligence isn't something the company is keeping at arm's length but rather a core feature set for the platform itself. Note that this is not available on the current _public_ beta, but CNET has a rundown of how to sign up to the Apple Developer Program (it's free!) if you're wanting to try it yourself. —Ross
Google’s Olympics ad for Gemini AI draws backlash in latest tech marketing fail (AdAge) With the Olympics in full swing in Paris, Google ran a national advertising campaign featuring a father asking Gemini AI to help his daughter write a fan letter to her athlete hero. Rather predictably, the ad wasn’t quite as inspiring as Google intended. The backlash from the advertising industry, and from what I saw on Threads (it may have been bad on Twitter too but I am not hanging out on that hellsite) was scathing, so much so that Google disabled the comments on the ad on YouTube. Personally, I think the comparisons between this ad from Google and Apple’s iPad “Crush” are a little harsh. —James
What we’ve watched
Last Week Tonight with John Oliver (Max/YouTube) If a writer of a TV show had suggested a plot line similar to the last two weeks of shenanigans in the United States presidential election, I suspect they’d have been shown the door of the writers room and told to come back when they’d decided to stop fucking about and write a script based on a modicum of reality. Thankfully we have John Oliver, an English comedian whose show Last Week Tonight features some of the finest reporting available anywhere. It’s funny because it’s true, which is also kind of scary. —James
I'm What the Culture Feeling (The full story of Kendrick Vs. Drake) (YouTube) I know what you're thinking: why on earth would you want to watch a 3+ hour essay on a rap beef that you've already heard too much about this summer? Because YouTube essayist F.D Signifier is an incredible storyteller and makes every minute of the video count—starting at the origin of hip-hop and how its business model shaped the cultural zeitgeist, through Drake and Kendrick's rise. Even if you have just a passing interest in hip hop, the first hour alone will teach you just how deeply entrenched this moment speaks to the culture of hip hop. —Ross
Batman: Caped Crusader (Prime Video) For my money, the best adaptation of Batman has always been Batman: The Animated Series, with a brooding, minimalist-meets-art deco aesthetic—it was famously drawn using light colors on black paper, vs. the standard dark colors-on-white—and iconic voice work, especially from Mark Hamill as The Joker and the late, great Kevin Conroy as Batman. Batman: Caped Crusader takes many cues from BTAS, unsurprising given it shares a co-creator (Bruce Timm), and the result is a worthy successor—absolute BTAS nostalgia bait, sure, but it still scratches that itch. This is Batman in his early crime-fighting years, with stakes that lean more grounded and noir than some of the more fantastical, gadget-heavy versions—fitting for the Matt Reeves movie era. I'm only halfway through and it's very good so far, with a special nod to Hamish Linklater for adeptly taking over the title role. Another gorgeous, stylish take on the caped crusader. —Ross
What we’ve played
The Legend of Zelda: Tears of the Kingdom (Nintendo) I am still deep in my third playthrough of Sekiro—this week I struggled again with the Guardian Ape, possibly the most annoyingly difficult boss in the game. But I also spent some valuable R&R time roaming the skies, land and depths of Hyrule playing The Legend of Zelda: Tears of the Kingdom. It’s been well over a year since this sequel to Breath of the Wild was first released but I still find this game a delight to play. It’s funny too: I spent time trying to build a two-wheeled tank thingy only to hit it by accident and watch it speed off down the hill and fall off a cliff. Much as I love playing FromSoftware games, I confess if ever I was asked the question, "What is the most creative, clever and imaginative video game you’ve ever played?" My answer would be Tears of the Kingdom. It’s not even up for debate. —James
DeepGame for ChatGPT. Inspired by this Polygon essay from Zoë Hannah, I’ve started to put in some serious time exploring this story generation game, which reminds me as much of the pulpy choose-your-own adventure books I read growing up as it does a tabletop experience. My ghost-hunting mystery may not be as true to the Supernatural theme as I was hoping—and it was not for lack of trying on my part to set it up that way—but I’m enjoying coming back to the transcript sporadically when I have time just to see how the story will play out. It’s not what I’d call compelling, award-winning writing, but the more I learn the “rules” of how to best interact with the game, the more fun I’m having. —Ross
What else we’ve loved
DocuByte: I Am A Computer: James Ball is a British “photographer, retoucher, art director and ultimate nerd.” I came across his work this week when this collection of images was shared by Present & Correct on Threads. To say I have become a fan of James’s work is understatement; I find it extraordinary, particularly his portfolio of images titled Guide to Computing. Talk about creative. —James
Paris Olympics Sharpshooter memes: It started with Kim Yeji, the South Korean sharpshooter with futuristic specs. Then there was Turkey's Yusuf Dikeç, who won silver with one hand in his pocket, looking like he just rolled out of bed and decided to try his luck at a global competition. It turns out pretty much every Olympic Sharpshooter has enough style to headline their own anime, and the Great Internet Collective has responded in kind. It's increasingly rare to find joy on social media, but this was great. As @airbagged put it best, "I’m convinced that this Olympic event was directed by Hideo Kojima,” who we’ll note has shared (retweeted) his own favorite Dikeç fan art. All I’m saying is, I wouldn’t put it past him to include this in a future Death Stranding. —Ross
Speaking of the Olympics: Snoop Dogg has been the best reason to watch, if only for his commentary. How else were we supposed to find out about his bond with Queen Elizabeth II?
Waiting for Godot: Alex Winter dropped the poster of “Waiting for Godot” starring himself and Keanu Reeves and Threads went bonkers. I suspect this show will sell out instantly as the chance to see Bill and Ted in Samuel Beckett’s play is going to be too excellent (get it?) to miss. —James