Is it possible for AI to replace all recorded music with Taylor Swift covers?

A rogue artificial intelligence obsessed with Taylor Swift could replace all recorded music with artificially generated cover versions of her, researchers say. History tells us that this American singer-songwriter for elise to paperback authorThere is no evidence left that Ludwig van Beethoven or the Beatles ever existed.

nick collins at Durham University, UK. mick grierson Professors at the University of the Arts London have issued an unusual warning in a paper that says humans should think about ways to resist “now, rather than when it is too late.”

Thankfully, the risk of AI Swiftpocalypse is low. Collins said the idea is a thought experiment aimed at encouraging researchers to develop ways to protect all types of data, including music, literature, scientific research, and historical records, from being corrupted by AI. I am.

The pair lay out a future scenario in which we rely on a few centralized data stores, such as Spotify and Apple for music. AI could infiltrate these stores and corrupt, delete, or alter the data inside. This can be dramatic and obvious, or it can be gradual and unnoticeable. “It's very likely that within a few thousand years there will be at least some corruption and some conflict over the truth of music in audio recordings,” Collins says.

To make their point clear and show how AI can already manipulate the data it has access to, researchers used current AI models to create Taylor Swift songs, including Queen songs. did. bohemian rhapsodyFrank Sinatra's I've Got You Under My Skin and the beach boys Isn't it wonderful?. They calculate that producing these “Taylor's versions” of all recorded music currently requires 1.67 billion kilowatt-hours of electricity, costing him more than $266 million. However, this is an amount that Swift herself can afford.

Collins says that while digital and physical backups can leave us complacent about the safety and permanence of our data, AI with the right incentives and capabilities can access everything we record. It states that it may be damaged. “No matter how much we try to preserve human culture, unpredictable threats may emerge in the future,” he says.

However, not all experts are convinced that AI is such a serious threat. sandra wachter Oxford University researchers have shown that AI can cause great harm by replicating sexist and racist biases in humans, but Collins and Grierson said He said that such a feat would never be possible.

“I don't think there's a serious problem with AI waking up and setting its own goals, having its own motivations, and taking actions to achieve those goals,” she says. “I think that's a nonsense argument, and I don't think it's realistic. It's like asking me what I would do if aliens landed on this planet tomorrow. I think it's unlikely. I think so.”

Carissa VelisResearchers, also at the University of Oxford, said that decisive action against AI is needed, but not a dramatic “kill switch” to stop malicious models from progressing. Instead, it should be a careful system of checks and balances to ensure the safety of AI models.

“This argument seems to assume that there is a malicious AI that somehow has its own desires and becomes so powerful that we want to stop it,” she says. . “And that seems so implausible and so ridiculous to me.”

The real problem is that we will integrate AI into many aspects of our lives and become completely dependent on it, even though it is likely not apocalyptic in nature. , which she believes raises issues such as racism and sexist prejudice that are still very harmful. Or they are simply making up facts that sound plausible.

“The more you leave it [AI] The more embedded it is in a product, the harder it is to turn it off. Not because this malicious thing has become powerful enough to take over, but because we've become dependent on it and it's very costly to stop it even when it's not working well,” Bellis said. says Mr.

Taylor Swift did not respond to a request for comment.