AI-generated song mimicking Drake and the Weeknd gets submitted for Grammys

Welcome to The Dispatch! We are the newsletter that keeps you informed about AI. Each weekday, we scour the web to aggregate the many stories related to artificial intelligence; we pass along the news, useful resources, tools or services, technical analysis and exciting developments in open source. Even if you aren’t an engineer, we’ll keep you in touch with what’s going on under the hood in AI.

AI-generated song that mimics Drake & The Weeknd submitted for Grammy consideration

From Variety: An AI-generated track titled "Heart on My Sleeve," featuring vocal facsimiles of Drake and the Weeknd, has been submitted for Grammy consideration. Mixed using AI by the shadowy creator ‘Ghostwriter’, the song went viral in April and has now received an endorsement (of sorts) from Recording Academy CEO Harvey Mason Jr., who stated that the track is "absolutely eligible" for a Grammy since it was written by a human. However, neither Drake nor the Weeknd had any involvement in the creation of the song.

More details:

• Ghostwriter submitted "Heart on My Sleeve" for "best rap song" and "song of the year" - categories that traditionally honor the writers rather than the performers. Ghostwriter wrote the lyrics, while the vocals were AI-generated.

• The Recording Academy has established rules against awarding AI-generated compositions but allows for "assistive AI," as in the case where Paul McCartney used the technology to enhance an old John Lennon vocal track. The emphasis remains on human authorship for eligibility.

• Despite its creative ingenuity, the song has had a turbulent journey in terms of distribution. Initially appearing on YouTube and streaming platforms, it was taken down due to apparent copyright issues from Universal Music but has since resurfaced thanks to third parties.

• Recording Academy CEO Harvey Mason Jr. elaborated that any work involving AI can be submitted for a Grammy, but only the "human portion" of the composition would be considered for an award. He mentioned the Recording Academy's position on distinguishing between human-driven and AI-generated creativity - a fine line.

Takeaways: If you haven’t heard it yet, here is the song in question. Perhaps not coincidentally, Ghostwriter released a second such creation yesterday on TikTok featuring two other popular artists - and it garnered well over 5 million views before it was removed/deleted from the platform. Here’s how we see it:

• Using an artist’s voice to create content without consent is clearly illegal. Ghostwriter has said if the original artists endorse his creations, he will pay them royalties. Is that enough? For some artists, it seems like the answer is yes. For most, we would expect not.

• Platforms like YouTube are put in an impossible situation: to issue a copyright takedown to YouTube, there needs to be an obvious copyright. Since “Heart on my Sleeve” is an original song, UMG doesn’t own it. This is why, after initially being taken down, the song is still on the platform.

• Something that can get lost amid the high stakes: AI-generated music can be awesome, and fun. The historically important intertwining of technology and music is at the dawn of a new era, and it’s something we should be celebrating - even as we navigate complex and sensitive realities with the deepest care for our artist’s identities.

Microsoft and Google’s most recent earnings calls mentioned AI 73 and 90 times, respectively, while Apple’s only mentioned it six. Given the secrecy around their new language model ‘Ajax’, it’s clear that Apple is taking a more ‘top secret’ approach to AI development than their tech peers.

Meanwhile, Steam has shut the door on even ‘fringe’ AI content over intellectual property worries. Developers will face a more bifurcated landscape in deciding how to pursue game development, as Steam is clearly not taking any chances with regard to AI’s involvement. That will put developers who want to integrate the technology into a bind - Steam is the most widely used video game distribution platform in the world, with around 75% of the global market.

• Inside Elon Musk’s struggle for the future of AI

• UN calls for age limits for AI tools in schools

• How Anthropic’s Daniela Amodei is keeping AI from spinning out of control

• Introducing Claude Pro (ChatGPT’s top competitor)

• Morgan Stanley to launch AI chatbot to woo wealthy

• AI a growing cause of 'tech anxiety' among business leaders

• New Slack AI capabilities announced to transform productivity

• HubSpot unveils strategy to integrate AI across the platform

In the blog post, IBM highlights their rigorous approach to trust and data governance in developing Granite, including screening for objectionable content and establishing model lineage back to original training data (read more about why this is important here). Watson Studio will enable capabilities like prompt tuning and synthetic data generation to further adapt models. IBM has been continually expanding Watson Studio’s generative AI capabilities - and they say they’re just getting started.

• Open Interpreter - an open-source Code Interpreter that runs locally

• SMPLitex: a generative model and dataset for 3D human texture estimation from a single image

• ‘We’re at the start of a revolution’: could lab brains eventually take over AI?