The OpenAI logo appears on a mobile phone and an image generated by ChatGPT’s Dall-E text-to-image translation model appears on a computer screen, Dec. 8, 2023, in Boston. European Union lawmakers are set to give final approval to the 27-nation bloc’s artificial intelligence bill on Wednesday, setting a world-leading set of rules for the rapidly developing technology set to take effect later this year. (AP Photo/Michael Dwyer, File)
Huge amounts of money are being poured into generative artificial intelligence, a Bay Area-born technology that’s taking the world by storm, but the relentless hype has created a bubble that experts say is bound to pop. So far, the returns on investment have been largely disappointing, and serious problems abound.
It’s been less than two years since San Francisco’s OpenAI released its ChatGPT-generating AI bot, but it’s already sparked an arms race among big tech companies, a surge in venture capital funding for AI startups, and a flurry of businesses trying to cut costs and boost productivity by incorporating the technology into just about every product and service imaginable.
Investors have pumped more than $24 billion into generative AI, according to consulting firm EY, and Goldman Sachs predicts that tech companies plan to spend $1 trillion on AI infrastructure over the next few years. While many technologists have high hopes for the technology, which uses patterns and relationships in data to generate text, images and voice, others point out significant drawbacks.
“Everybody wants to make money in the AI race,” said Howard Young of San Jose, who at tech giant AAEON integrates AI software into computer systems to improve urban infrastructure, industrial processes, manufacturing and healthcare. Young was at the Reuters Momentum AI conference in San Jose this month with hundreds of other tech workers and executives. “We’re not seeing any real organic revenue yet,” he said. “Even with the best talent in Silicon Valley, it’s going to take a while.”
Jim Covello, chief global equity researcher at Goldman Sachs, said people in the tech industry are “massively overestimating” generative AI’s current capabilities and questions remain about how much it will improve.
“The technology is still a long way from being useful,” Covello wrote in a June newsletter from the bank focused on generative AI. “If there are fewer use cases and adoption rates for AI technologies than current consensus estimates, it’s hard to imagine this wouldn’t be problematic for many of the companies currently investing in the technology.”
David Kahn, a partner at Silicon Valley venture capital giant Sequoia, wrote in a blog post in June that the “speculative fever” around generative AI is spreading a “delusion” from Silicon Valley that “we can all get rich quick.”
Generative AI, affectionately known in the tech industry as “genAI,” is definitely forming a bubble that will burst, and the pain is on the way, but it’s nothing new to Silicon Valley, says Steve Blank, an adjunct professor of management science and engineering at Stanford University, who likened the technology to the bubble-like birth of the World Wide Web and the subsequent collapse of the dot-com bubble.
“It’s not that we were wrong about the Web, it’s just that we needed to iterate and do more thorough research to separate the good from the bad,” Blank said.
For Bay Area startups, it’s true that if you don’t have AI in your title or story, you won’t get funded, Blank says. “It’s funny. That big sucking noise is the lemmings putting money into the next big thing. One or two of them will strike it rich. The rest of them will lose all their money.”
Goldman Sachs’ Covello said the bubble’s inevitable collapse may not be as damaging as the dot-com collapse “simply because a lot of the companies investing today are better capitalized than the companies investing back then.”
The business transformation touted behind huge spending on generative AI has not materialized, with the expensive technology doing little to solve the complex problems that would enable widespread automation of tasks and functions.
Meanwhile, the technology continues to be plagued by what are, depending on who’s talking, either growing pains or fundamental flaws. Leading developers of generative AI are in court fighting artists, photographers, authors, programmers, music labels and newspapers (including this one) who they say have stolen their copyrighted material by scraping the internet to “train” AI models.
Training and running generative AI has reversed progress toward Google and Microsoft’s climate and sustainability goals. Both companies reported dramatic increases in electricity and water usage due to AI-related data processing and storage last year. Chatbots and generative search continue to generate errors and lies. Propagandists use the technology to spread disinformation, students use it to cheat, and bad actors use it to defraud and harass. State legislatures introduced nearly 200 bills last year to oversee and regulate AI.
Still, believers in the potential of generative AI see most challenges being overcome through innovation, and point to important early uses and powerful potential applications.
“This industry is just taking shape,” Blank says. “The Earth is still molten. The outlines of the continents are just beginning to become visible.”
The technology already helps companies quickly identify high-performing and displaced employees by simply inputting company emails and texts, prompting them to see who is asking whom important questions or issues and who is providing answers or solutions, said Cong Leong, CEO of Milpitas-based data management firm ZL Technologies. Generative AI is “extremely powerful” at making sense of chaos, Mr. Leong said.
“Right now, we’re just getting off our knees,” Leon said. “When we get to walking and running, we’ll be astounded at what that means.”
Shomit Ghose, a lecturer at the University of California, Berkeley and venture capitalist, noted that generative AI is being used to develop drugs for treating lung diseases that can lead to cancer, and is currently undergoing clinical trials. The technology is also quickly beginning to make inroads in weather forecasting. Ghose believes there is too much investment in the AI technology underlying generators like ChatGPT, and too little investment in other types of generative AI that are beginning to revolutionize science.
According to the International Energy Agency, energy companies are beginning to use the technology to improve the efficiency of power grids and integrate wind and solar power to get the most out of it.
Shoubhi Ramakrishnan, chief digital and technology officer at pharmaceutical giant GSK, told attendees at the Momentum AI conference that the company had increased production of its shingles vaccine by 1 million doses by using generative AI to create a “digital twin” that replicates factory operations in software.
“This is an underrated and overrated technology at the same time,” Ramakrishnan said.
