An increasing number of companies that develop or rely on artificial intelligence technology may become insolvent as costs balloon across the sector, experts say. (iStock.com/Dragon Claws)
On a smaller scale, StoryFile — which promised to use AI to create interactive videos to leave behind for your descendants and landed William Shatner as its spokesperson — filed for bankruptcy in May.
Overseas, AI-based cryptocurrency platform fetch.ai entered administration in the United Kingdom in January before it was purchased by another crypto firm, and German AI translation company Lengoo began insolvency proceedings in March.
Science writer and editor HP Newquist, the editor and publisher of an early newsletter in the AI field and author of "The Brain Makers," a book on the subject, said the filings may represent an "inflection point" for the tens of thousands of corporations touting themselves as AI companies worldwide.
"They're really right now the tip of the iceberg," he told Law360.
He noted there is precedent to go on: the initial surge of interest in the commercial use of AI in the late '80s and early '90s for things like expert systems, combinations of programs and databases meant to mimic human expert decision-making.
"In that first wave, every company, every single one, went out of business," he said.
He said one thing distinguishing this from the internet boom of the '90s and the phone app boom of a decade later has been the "overwhelming crush" of user interest. The need to rapidly scale up to handle millions of users has exposed how expensive AI systems are to run, as much as six figures a day, he said.
Chris Selland, a partner at technology consulting firm TechCXO, told Law360 that many of the largest internet infrastructure companies, like Amazon, have been subsidizing AI to the benefit of smaller firms, but those costs are mounting and the giants have AI projects of their own that will need those resources.
"Over time, these bills are going to need to be paid," he said.
There's also the question of the high-end hardware needed to run these systems, and the power they consume, Newquist said.
"Chips and servers are in short supply, there's already a black market for them," he said.
Newquist said another rising cost is what he estimates are the several thousand programmers in the world who can "make this stuff work."
"Talent is being scooped up, not just in the U.S., but worldwide," he said.
This comes as investors are beginning to get "antsy" for returns, Newquist added.
"Up until now, it's literally been an open checkbook and a blank check for anyone with a good AI story," he said.
He said this is already causing upper management turmoil, citing as an example Open AI, which last year saw founder Sam Altman ousted and rehired over the course of a week. In June, the company's chief scientist left with an announcement that he was starting a new company "insulated from short-term commercial pressure."
Both Newquist and Selland said companies will also have to deal with the challenge of transitioning users from free to paid services.
"We need to see some customers that are going to pay these bills," Selland said.
Both said the likely survivors will include startups that can compete with the resources and capital of AI players like Google and Microsoft by focusing on providing a specific product to a specific market, like AI for medical diagnosis or insurance claim review.
"AI for dentists," Selland said. "Very specific-use cases."
Newquist also said a more focused AI would provide superior results, saying attorneys that have found themselves in hot water for submitting AI-written filings with made-up citations might have avoided that with AI with specialized training on statute and case law.
"That's why you can't use a general-purpose large-language model," he said.
--Editing by Kelly Duncan.
For a reprint of this article, please contact reprints@law360.com.