AI companies are soaring on promises they can revolutionize society while making a profit. What if they're flying too close to the sun?
The fear: The latest models generate publication-worthy essays and award-winning artworks, but it’s not clear how to make them generate enough revenue to both cover their costs and turn a profit. The bubble is bound to burst.
Horror stories: During the dot-com bust of 2000, internet stocks tumbled as their underlying weaknesses became apparent. The cryptocurrency crash of 2022 evaporated nearly two-thirds of Bitcoin’s value. Some observers believe that, similarly, today’s hottest AI bets are overhyped and overvalued.
- ChatGPT’s base of active monthly users ballooned faster than that of any application in history. But it lost users steadily through the second quarter of this year.
- Serving models like ChatGPT to a mass audience is expensive. Microsoft, which supplies infrastructure to run ChatGPT and other OpenAI innovations, is trying desperately to cut the cost, primarily by distilling OpenAI models to reduce their size and thus the processing power they require.
- An ongoing shortage of AI processing chips is limiting server capacity. Some providers of cloud computing may be overcompensating by spending to build processing capacity that they won’t be able to sell at a profit.
Bad omens: Generative AI accomplishes new marvels with each passing month, but that doesn’t necessarily translate into profitable businesses. Investors and analysts are throwing up red flags.
- Investors poured $14.1 billion into generative AI startups in the first half of 2023, compared to $2.5 billion in all of 2022 and $3.5 billion in all of 2021, according to CB Insights, which tracks startup funding.
- While some venture investors have been betting on AI startups, others have urged caution. “Companies are extremely overvalued,” one investor told Financial Times in March.
- The market analyst Gartner recently published a graph that projects expectations for generative AI over time. Gartner’s Hype Cycle graph places generative AI at the “peak of inflated expectations.” A descent into a “trough of disillusionment” follows.
Facing the fear: No one knows what the future will bring, but generative AI’s usefulness, which already has attracted billions of users, continues to evolve at a rapid pace. No doubt, some investments won’t pay off — but many will: The consultancy McKinsey estimated that generative AI could add between $2.6 trillion and $4.4 trillion to the global economy annually. Already generative models form the foundation of conversational assistants, image generators, video effects, and automated coding tools. An avalanche of further applications and refinements appears to be inevitable as the technology continues to advance.