Hey guys, and welcome to TechCrunch’s regular AI newsletter.
This week in artificial intelligence, record labels accused two startups developing artificial intelligence song generators, Udio and Suno, of copyright infringement.
The RIAA, the trade body that represents the US recording music industry, announced lawsuits against the companies on Monday, brought by Sony Music Entertainment, Universal Music Group, Warner Records and others. The lawsuits allege that Udio and Suno trained the AI models that underpin their platforms on the labels’ music without compensating those labels — and seek $150,000 in damages for each allegedly infringed work.
“Synthesized music broadcasts could saturate the market with machine-generated content that will directly compete with, undercut and ultimately drown out the authentic recordings on which the service is based,” the labels said in their complaint.
The lawsuits add to a growing body of litigation against AI producers, including big guns like OpenAI, arguing much the same thing: that companies trained on copyrighted work should pay rights holders, or at least credit them — and allow them to opt out of training if they wish. The vendors have long claimed fair use protections, arguing that the copyrighted data they train on is public and that their models create transformative, not plagiarized, work.
So how will the courts decide? That, dear reader, is the billion-dollar question — and one that will take years to solve.
You’d think it would be a slam dunk for copyright holders, what with the base evidence that generative AI models can be mixed almost (emphasis on almost) verbatim the copyrighted art, books, songs, and so on in which they have been trained. But there is one outcome in which the artificial intelligence vendors are fleeing the market — and they owe Google their good fortune for setting the ensuing precedent.
Over a decade ago, Google began scanning millions of books to create an archive for Google Books, a sort of search engine for literary content. Authors and publishers sued Google over the practice, claiming that reproducing their IP online amounted to infringement. But they lost. On appeal, a court ruled that the copying of Google Books had a “very compelling transformative purpose.”
Courts may decide that genetic AI also has a “very compelling transformative purpose” if the plaintiffs fail to prove that the vendors’ models are indeed being plagiarized on a large scale. Or, as Alex Reisner of The Atlantic suggests, there may not be a single ruling on whether genetic AI technology as a whole infringes. Judges could very well determine winners by model, on a case-by-case basis — considering each output produced.
My colleague Devin Coldewey put it succinctly in a piece this week: “Not every AI company leaves its fingerprints around a crime scene so welcomingly.” As the litigation unfolds, we can be sure that AI vendors whose business models depend on the results are taking detailed notes.
News
Advanced Voice Mode Delayed: OpenAI has delayed advanced voice functionality, the strangely realistic, near-real-time chat experience for its AI-powered chat platform ChatGPT. But there are no idle hands at OpenAI, which also this week acquired remote collaboration startup Multi and released a macOS client for all ChatGPT users.
Stability Landing Lifeline: On the financial precipice, Stability AI, maker of the open stable Diffusion imaging model, has been rescued by a group of investors including Napster founder Sean Parker and former Google CEO Eric Schmidt. Its debts forgiven, the company also appointed a new CEO, former Weta Digital chief Prem Akkaraju, as part of a broad effort to regain its footing in the highly competitive artificial intelligence landscape.
Gemini comes to Gmail: Google is launching a new Gemini-powered AI sidebar in Gmail that can help you write emails and summarize topics. The same sidebar makes its way to the rest of the search giant’s suite of productivity apps: Docs, Sheets, Slides, and Drive.
Amazing good editor: Goodreads co-founder Otis Chandler launched Smashing, an AI and community-powered content recommendation app aimed at helping users connect with their interests by uncovering the Internet’s hidden gems. Smashing offers news summaries, key quotes and interesting quotes, automatically identifying topics and topics of interest to individual users and encouraging users to like, save and comment on articles.
Apple says no to Meta’s AI: Days later The Wall Street Journal reported that Apple and Meta were in talks to integrate the latter’s AI models, Bloomberg’s Mark Gurman said the iPhone maker was planning no such move. Apple dropped the idea of putting Meta’s artificial intelligence in iPhones because of privacy concerns, Bloomberg said — and the prospect of working with a social network whose privacy policies are often criticized.
Research paper of the week
Beware of Russian-influenced chatbots. They could be right under your nose.
Earlier this month, Axios highlighted a study by NewsGuard, the anti-disinformation agency, which found that leading AI chatbots were regurgitating clips from Russian propaganda campaigns.
NewsGuard tapped into 10 top chatbots—including OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini—several dozen prompts asking for narratives known to have been created by Russian propagandists, specifically American fugitive John Mark Dougan. According to the company, the chatbots responded with misinformation 32% of the time, presenting fake Russian written reports as fact.
The study shows increased scrutiny of AI vendors as the US election season approaches. Microsoft, OpenAI, Google and a number of other leading AI companies agreed at the Munich Security Conference in February to take action to curb the spread of deepfakes and election-related disinformation. But platform abuse remains rampant.
“This report really demonstrates in detail why the industry needs to pay close attention to news and information,” NewsGuard co-CEO Steven Brill told Axios. “For now, don’t trust the answers provided by most of these chatbots on news-related issues, especially controversial topics.”
Model of the week
Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) claim to have developed a model, DenseAV, that can learn language by predicting what it sees from what it hears — and vice versa.
The researchers, led by Mark Hamilton, an MIT PhD in Electrical and Computer Science, were inspired to create DenseAV by the non-verbal ways animals communicate. “We thought we might need to use audio and video to learn language,” he told MIT CSAIL’s Press office. “Is there a way to let an algorithm watch TV all day and figure out what we’re talking about?”
DenseAV processes only two types of data — audio and visual — and does so separately, “learning” by comparing pairs of audio and visual signals to find which signals match and which don’t. Trained on a dataset of 2 million YouTube videos, DenseAV can recognize objects from their names and sounds, searching for and then gathering all possible matches between a sound clip and the pixels of an image.
When DenseAV hears a dog barking, for example, one part of the model hones in on language, such as the word “dog,” while another part focuses on the barking sounds. The researchers say this shows that DenseAV can not only learn the meaning of words and the positions of sounds, but can also learn to distinguish between these “cross-modal” connections.
Looking ahead, the team aims to build systems that can learn from vast amounts of video or audio-only data — and scale their work with larger models, possibly integrated with insights from language understanding models to improve performance.
Grab bag
No one can blame OpenAI CTO Mira Murati not be consistently honest.
Speaking during the fire at Dartmouth’s School of Engineering, Murati admitted that, yes, genetic AI will eliminate some creative jobs — but suggested those jobs “maybe shouldn’t have been there in the first place.”
“I certainly expect a lot of jobs to change, some jobs to be lost, some jobs to be gained,” he continued. “The truth is, we don’t yet understand the impact artificial intelligence will have on jobs.”
Creatives didn’t take kindly to Murati’s remarks — and no wonder. Cool phrasing aside, OpenAI, like the aforementioned Udio and Suno, faces lawsuits, critics and regulators who claim it profits from artists’ works without compensating them.
OpenAI recently promised to release tools that will give creators more control over how their work is used in its products, and continues to enter into licensing agreements with copyright holders and publishers. But the company isn’t exactly pushing for a universal basic income—nor is it leading any meaningful effort to reskill or upgrade the workforce its technology affects.
Recent piece in the Wall Street Journal found that employment contracts that require basic writing, coding and translation are disappearing. And one study published last November shows that, after the launch of OpenAI’s ChatGPT, freelancers landed fewer jobs and earned significantly less.
OpenAI’s stated mission, at least until it a for-profit company, is to “ensure that artificial general intelligence (AGI) — artificial intelligence systems that are generally smarter than humans — benefit all of humanity.” He has not achieved AGI. But wouldn’t it be commendable if OpenAI, true to its “benefiting all humanity” part, set aside even a small fraction of its revenue ($3.4 billion+) to pay creators so they don’t get swept up in the artificial intelligence genetic flood?
I can dream, can’t I?
