TLDR;
Sabine Hossenfelder discusses the potential bursting of the AI bubble, highlighting five bad omens: companies not seeing returns on AI investments, AI coding not meeting expectations, overinvestments needing correction, public talk of an AI bubble and investor retreat, and physical barriers like data center and energy supply shortages. While optimistic about AI's long-term potential, she expresses concern about the short-term overvaluation and unrealistic expectations surrounding current AI technologies.
- Companies are not seeing returns on AI investments
- AI coding is not meeting expectations
- AI overinvestments are about to be corrected
- There is public talk of an AI bubble and investor retreat
- AI is hitting physical barriers like data center and energy supply shortages
Lack of Return on AI Investments [0:24]
Reports indicate that companies are realizing their AI investments are not yielding expected returns. The US Census Bureau reported a decrease in AI adoption among US companies for the first time in two years. An MIT project found that 95% of companies implementing generative AI have not seen measurable return on investment. Similarly, a McKenzie report revealed that 80% of companies have not experienced a positive impact from using AI. Gartner has announced that the AI field is past peak hype and is currently in a trough of disillusionment.
AI Coding Underperforming [1:16]
The initial excitement around AI coding is fading as people realize it's not meeting expectations. Studies have shown that while AI can produce code quickly, it often introduces mistakes and security vulnerabilities that require fixing. A recent study from Meta found that large language models struggle with completing long tasks and can actually slow down developers. The need for "vibe coding cleanup" further indicates that AI coding is not panning out as expected.
Correction of AI Overinvestments [2:00]
Signs suggest that AI overinvestments are on the verge of being corrected. Major firms like Google, Alphabet, Meta, and Amazon have heavily invested in AI, driving positive stock trends. However, a Goldman Sachs report indicates that these companies have tied up significant capital in AI, which is unsustainable. The report predicts a slowdown in AI investments by these large companies, which will likely diminish optimism for AI and negatively impact the stock market.
AI Bubble and Investor Retreat [2:41]
Public discussions about an AI bubble and potential investor retreat are increasing. Open AI CEO Sam Altman acknowledged that the AI sector is in a bubble, with investors being overly excited. Goldman Sachs analyst Cash Rangan noted that the problem is more pronounced in private markets than in public software companies. Martin Fimoski, a former Reserve Bank of Australia economist and AI researcher, divested all AI investments from his pension fund, citing concerns that business leaders misunderstand AI capabilities and that inflated expectations resemble past bubbles.
Physical Barriers to AI Progress [4:15]
AI development is encountering physical barriers, specifically data center and energy supply shortages. A Goldman Sachs report indicates that data center vacancy rates are at a record low, with new power at scale not expected until 2028 or beyond. This shortage will drive up prices for model training, potentially cutting off newcomers from AI progress and worsening the overall outlook.