Will The Real Slim Shady Stand Up to AI?

Acclaimed DJ David Guetta recently posted a video of a song he mixed with an AI-generated Eminem. Using ChatGPT he asked the AI to write lyrics about ‘Future Rave’ in the style of Eminem. His obvious next step was to use a different AI to replicate Eminem’s voice by inputting the lyrics. What happened? “People went nuts.”

Podcasters Beware - AI is Coming for Your Voice

So what’s the concern? Haven’t artists and authors "stood on the shoulders of giants" for decades? Bob Dylan was inspired by Woodie Guthrie, the Beatles by Elvis Presley, and David Guetta by Prince and Michal Jackson. That’s all well and good, but the challenge comes when “being inspired by” drifts into being “copied by” with famous lawsuits in music spanning decades. In fact, both Open AI (maker of ChatGPT) and Stability AI (maker of Stable Diffusion) have open lawsuits largely focused on how their AI models unfairly leveraged copy-written material. 

The legal aspects of this will certainly be settled in court over the next few years and I expect that the result won’t simply be no more generative AI. Rather, it’ll likely look similar to what happened when YouTube was in its infancy and had to build and adopt new technologies to fight piracy. Have you ever uploaded a personal video and used the latest top 40 track in the background? If yes, you know that YouTube noticed it (using AI in fact) and requested you to confirm your usage rights to publish online. 

The concept of ‘Fair Use’ makes it possible for artists to criticize, parody, or comment on another work for a “limited and transformative use.” If a large language model (LLM) can trace a substantial component (51%, 10% — this will be decided as well) of its generated content might it need to reference it? Of course, that’s complex given the massive nature of these models — with GPT-3 being tuned with 175B parameters across ~500B tokens. We’ve seen a fairly steady string of advancements in machine learning model explainability largely due to pressure from regulated industries like finance and healthcare so we should expect progress with LLMs over time as well.

Most Highly Valued Generative AI companies

Will we eventually see a royalty model where the models are sending micropayments to content creators? If ChatGPT plays the role of David Guetta and remixes your blog, will you be cited as the author? That won’t resolve the bad actors who download data massive sets without permission, but it could pave the way for a new version of Robots.txt which defines what web crawlers like Google can or can’t do on a website. Or maybe it’ll be the landmark use case for NFTs we’ve all been searching for post the Bored-Ape craze. The newest wave of transformer models like ChatGPT are simply too powerful to contain now that the world has seen their potential. 

Content creators may soon decide which platform they publish their content on by looking beyond clicks and ad revenue toward data usage rights and royalties coming from AI models. We’re not just talking about social media influencers — software engineers will be facing similar dilemmas. Millions of engineers benefit from Github’s code repository platform but if their latest code commit gets crawled and soon shows up as a curiously similar response to a ChatGPT request to “Write a WordPress plugin” might they rethink where they store code? If not the individual, we can be sure large organizations will. 

Where's the money for generative AI going?

Where's the money for generative AI going?

There are countless reasons for organizations to stay on the sidelines when it comes to GPT3 and similar generative models. The same was true for the internet, mobile, and social media. But, there’s a big distinction between being an early adopter who throws care to the wind and purposeful and strategic innovation. Begin building a list of use cases that would move the needle, track startups pioneering the technology, and start building to learn. Build, measure, and learn as you watch the legal dust settle. As Marshall Mathers once said (or maybe it was ChatGPT?)…

If you had, one shot, or one opportunity, to seize everything you ever wanted, In one moment, Would you capture it, Or just let it slip?

Previous
Previous

Super Mario Meets ChatGPT

Next
Next

Looking Backwards vs. Working Backwards