OpenAI Board Chairman: We are indeed in an "AI bubble," and there will inevitably be huge winners, while many will suffer significant losses

The fervor in the AI field is sparking an intense debate about "bubbles," and Bret Taylor, chairman of the OpenAI board, provided a clear yet complex answer: we are indeed in a bubble, but that does not prevent AI from ultimately creating significant economic value.

Recently, in an interview with media outlet The Verge, Bret Taylor agreed with OpenAI CEO Sam Altman's earlier viewpoint, acknowledging that "we are in an AI bubble, and some people will lose a lot of money."

Taylor warned that, like any wave of disruptive technology, this process will inevitably produce huge winners while causing significant losses for many. He also believes that AI will change the economic landscape and create immense value, which can coexist with the fact that there is a bubble in the market.

Taylor directly compared the current AI craze to the internet bubble of the late 1990s. He pointed out that, although countless companies fell during the bubble burst, in the long run, "those people in 1999 (who judged the future of the internet) were right to some extent."

Today, companies like Amazon and Google, which emerged from that era, have become some of the highest-valued enterprises globally, proving that the foresight during the bubble can ultimately pay off:

"In fact, if you look at the world's GDP, how much has the existence of the internet actually created or influenced? Some might say that everyone in 1999 was right. It had the same impact on almost all metrics."

The Key is to Distinguish the "Directional" Nature of the Bubble

Taylor elaborated on his analogy between the AI bubble and the internet bubble. He believes that the key is to distinguish between the correctness of direction and the success rate of specific investment targets.

During the internet bubble, many business models like Webvan (online fresh food delivery) ultimately failed, but their core ideas were successfully realized by companies like Instacart and DoorDash after the internet infrastructure matured. This indicates that even if initial attempts fail, the underlying trends and demands are genuinely present.

Similarly, in the early days of the internet, many companies investing in fiber optic networks went bankrupt, but this infrastructure was eventually utilized by later entrants, supporting the prosperity of the entire digital economy.

Taylor stated that both "AI will change the economy" and "many people will lose money" can be true at the same time:

"I believe that AI will change the economy is a fact, and I think it will create tremendous economic value in the future, just like the internet. At the same time, I believe we are also in a bubble, and many people will lose a lot of money. I think both are absolutely correct, and there are many historical precedents for these two things happening simultaneously."

This means that the current massive investments, regardless of which company they ultimately flow to, are paving the way for the next generation of AI applications, but not all participants will share in the final rewards

Reasons for AI's High "Burn Rate": The Market is Still Immature

Taylor disagrees with the view that "model iteration has significantly slowed down." He cites coding tasks as an example, pointing out that new models still show "stepwise" improvements in specific areas. However, he also believes that as model capabilities mature and become more widespread, models have reached a level that is "good enough" for many tasks.

He predicts that building AI applications in the future will be more like "how to use a database" rather than "how to write a database."

Regarding market skepticism about the return on investment in AI, such as a report from MIT indicating that many companies' AI expenditures have not yielded results, Taylor believes that this is mainly because the market is still immature. Many companies are engaging in "AI tourism," attempting to build solutions themselves, a process that is complex and prone to failure.

He believes the correct path is to purchase mature AI solutions focused on specific areas, like Sierra (for customer service) or Harvey (for legal). As more "application-oriented AI companies" emerge, businesses will be able to more directly purchase AI agents that address their pain points, thereby truly realizing the value of AI.

"I think we are in the early stages of AI, and there isn't an excellent vendor to solve every problem you encounter in your business. So, you either have to wait or build it yourself."