It’s been two years since the phrase “generative AI” started buzzing through my inbox. It wasn’t a new term (it appeared in one of Gartner’s famous Hype Cycle reports in 2020), but as the summer of 2022 drew to a close, the influx of messages and pitches I received clearly indicated a rapid rise in buzz around AI-powered tools that can generate content like text, images, and computer code. And when OpenAI released ChatGPT in November 2022, generative AI catapulted into mainstream culture and has been going strong ever since.
But something about that upbeat story has changed in the past few weeks.
A series of media headlines have been eager to pour cold water on the whole thing, with Goldman Sachs calling generative AI “over-hyped” and “too expensive”, venture capital firm Sequoia Capital saying the “AI bubble has reached a tipping point”, and the “AI hype machine is running on empty”.
Why? Generative AI chatbots struggle to answer basic questions or hallucinate false information. The most sophisticated generative AI models constantly need data and computing power. Generative AI startups with little to no revenue always have to desperately seek large funding rounds to survive. Fortune 500 companies can’t productize generative AI use cases due to concerns about accuracy, liability, and security.
And after the S&P 500 suffered its biggest sell-off in two years on Monday, there’s growing speculation that the generative AI bubble is starting to deflate.
According to Gartner’s Hype Cycle, generative AI has passed the “Peak of Inflated Expectations” and is heading straight for the looming “Trough of Disillusionment.” If that’s true, what will come next will be pain and disruption. Investment capital may dry up. Startups may go bust. Staff cuts may occur.
The pain of a market correction will be unfair and brutal for the many startup employees, founders, and investors who took on the effort and risk necessary to advance the generative AI sector, but demoting generative AI from its lofty perch is also necessary for the long-term sustainability of the AI industry, Kjell Carlson, a former Forrester Research analyst who is now head of AI strategy at enterprise data platform Domino Data Lab, told me.
“I’m pretty confident that people will realize that Gen AI isn’t the be-all and end-all of AI,” he said, referring to a range of other artificial intelligence techniques, such as predictive AI and machine learning, which were already delivering substantial return on investment before generative AI came along. “Gen AI is one in a set of technologies that is part of a broader toolkit of different technologies that require work,” he explained. “There’s no magic button. It’s about leveraging the technology for the right use cases.”
Don’t be afraid of the valley
To be clear, generative AI isn’t going away. From ChatGPT and Microsoft Copilot to Google’s Gemini, Anthropic’s Claude, and Meta’s Llama, these models and tools are already part of our lives for productivity, efficiency, or just fun. Just as we’ve become accustomed to getting the information we need in seconds with a Google search, we’re also becoming accustomed to getting easy-to-read summaries of work meetings, writing notes for colleagues, and creating images and presentations by speaking a few words.
But let’s be realistic: A massive amount of investment in generative AI, estimated at $1 trillion, has yet to pay off. While much of it may not be as ridiculous as, say, the dot-com bubble of UrbanFetch or Pets.com (we all remember ice cream deliveries and doll giveaways), it’s hard to argue with the idea that generative AI is getting the reality it deserves.
“Ironically, I think I was one of the first industry analysts to jump on the Gen AI bandwagon,” Carlson says. “Even though it was a success by any standards, the expectations for how quickly it would impact the bottom line of major organizations were not based in reality.”
Gartner Global Research Director Chris Howard said in a recent video that the so-called “disillusionment phase” is a critical stage in any technology development. The premise is simple: after an initial period of excitement and enthusiasm from early adopters, the new technology makes it into the hands of mainstream users, who realize it doesn’t live up to their inflated expectations. This is followed by a period of decline, during which the technology improves and expectations are reset.
“This isn’t a dark and dangerous place,” Howard explained in the video, “This is a place where we think about whether something will work or not.”
For generative AI, the valley is a stage marked by small, incremental advances in applications that bring real benefits to businesses and users, and while OpenAI CEO Sam Altman’s declaration that artificial general intelligence (AGI) will create “the most powerful technology humanity has ever invented” may not make for very sexy headlines.
Even Wedbush Wall Street technology analyst Dan Ives, who is bullish on AI stocks, says now is a critical time for tech companies to put their money where their mouth is when it comes to generative AI. “They need to show use cases and monetization to justify the AI revolution,” he told me in a text.
Ives said he thinks Microsoft, AMD, NVIDIA, Palantir and Oracle have proven they can deliver real value, but with a plethora of generative AI startups backed by multi-billion dollar valuations, the industry as a whole still has a lot to prove.
While there are no guarantees, AI techniques have matured and have a long history of contributing to other new areas of AI, such as computer vision, which has become a key part of today’s multimodal generative AI (AI that can generate not only text, but also images, videos, etc.).
So perhaps generative AI, driven by other emerging technologies such as agent AI (AI systems designed to act like autonomous agents to pursue complex goals and workflows), may still be able to realize its full potential.
Perhaps now is the time for the real ground work on generative AI to begin. “I think this will be a false AI winter,” Steve Jones, executive VP at technology consultancy Capgemini, said in a LinkedIn post. “Hopefully, the hype will die down and we can focus on the work.”
