Pay Up

Having courted billions of dollars in investment and becoming an unstoppable juggernaut of hype, the generative AI industry should be sitting pretty. But it's deathly paranoid about one thing: actually having to compensate artists for training AI models using their copyrighted work.

Exemplifying those fears, the venture capital firm Andreessen Horowitz — one of the biggest financial backers of AI — warned in comments to the US Copyright Office (USCO) that new regulation on training data "will significantly disrupt" investment into the technology and the expectations around it, Insider reports.

"The bottom line is this," the firm wrote. "Imposing the cost of actual or potential copyright liability on the creators of AI models will either kill or significantly hamper their development."

So in other words, the bottom line is the bottom line — and pesky technicalities like copyright laws shouldn't stand in the way of that.

All the Content

According to the VC firm, gobbling up copyrighted content is the "only practical way" of training large language models, the technology that powers chatbots like OpenAI's ChatGPT.

And the industry should be entitled to a whole lot of that content, too: "something approaching the entire corpus of the written word," Andreessen Horowitz argued, and "an enormous cross-section of all of the publicly available information ever published on the internet."

Because it believes the usage of this data counts as "fair use," it argues that AI companies shouldn't have to pay a dime.

An expedient line of thinking, because as the firm warns — or admits — LLMs have already ingested so much copyrighted material that companies would have to shell out "tens or hundreds of billions of dollars a year in royalty payments," which only the biggest corporations could afford.

Losing the Lead

If those arguments sound specious to you, wait till you hear out Andreessen Horowitz's biggest trump card: "national security." Because, folks, the country really needs to be a leader in the tech that gives us AI thirst trap influencers.

"The best way to lose the United States' current leadership in the burgeoning AI industry... is by rushing to pass legislation that undermines the long-standing and principled approach to copyright law that has made this country both a creative and technological leader," the firm wrote.

"We cannot afford to be outpaced in areas like cybersecurity, intelligence operations, and modern warfare, all of which are being transformed by this frontier technology," it added.

It's undoubtedly a consideration that will be at the back of lawmakers' minds. Will they countenance hampering one of the US's hottest industries with stringent copyright protections that competing countries may forgo?

But, if AI investors want the USCO to look the other way so companies can skirt paying due compensation, it faces some staunch opposition. High profile authors are suing companies like OpenAI, and a barrage of other lawsuits against the tech may ultimately decide the fate of the industry — however long that takes.

More on AI: Amazon Reportedly Training AI With Twice as Many Parameters as GPT-4

Share This Article