AI Winter: Why we don’t care if they “fail” and lose money

🌈 Abstract

The article discusses the massive investments being made by Silicon Valley companies and venture capital firms in training ever-larger AI models, and whether these investments will yield a return on investment. It explores the possibility that the current approach to AGI may not be the right path, and how this could impact the broader adoption of AI in businesses.

🙋 Q&A

[01] The Massive Investments in AI Models

1. What are the key concerns raised about the massive investments in training large AI models?

• Goldman Sachs and Sequoia Capital are not convinced that the billions being spent on training ever-bigger AI models will yield a return on investment.
• The author suggests that even if the investments in AGI do not lead to the hyped returns, it may not matter for the successful adoption of AI in businesses.
• The author proposes that the investments could lead to new technologies and capabilities, similar to how the Apollo space program led to the development of fly-by-wire technology, which became ubiquitous in aviation.

2. How does the author view the potential outcomes of the current approach to AGI?

• The author suggests that the current approach to AGI may not be the right path, and that the investments may only result in incremental improvements rather than a breakthrough to AGI.
• However, the author believes that these incremental improvements could still be valuable for businesses, similar to how the Apollo program led to the development of technologies that benefited various industries.

3. What is the author's perspective on the risks associated with the current approach to AGI?

• The author suggests that the current approach is focused more on "getting there" (to AGI) rather than "getting there safely and back again," which could pose risks for businesses that rely on these AI systems.
• The author emphasizes the importance of businesses focusing on the risk management and systemic management of AI, rather than just on training larger models.

[02] The Potential Impact on Business Adoption of AI

1. How does the author view the potential impact of diminishing returns on the adoption of AI in businesses?

• The author believes that as the investments in training larger models lead to diminishing returns, there will be a shift towards cost-effective solutions that can provide significant improvements in specific use cases.
• This, in turn, will lead to an increase in business adoption of AI, as more use cases become affordable and accessible.

2. What is the author's perspective on the potential "AI winter" scenario?

• The author suggests that the upcoming "AI winter" may be different from previous ones, as businesses will still have access to the compute capacity to use AI, but the justification for spending billions on marginal improvements in large models will diminish.
• The author likens this to having a "pristine powder" of AI capabilities that businesses can leverage, even if the investments in training larger models slow down.

3. How does the author view the role of businesses in the context of the current AI landscape?

• The author emphasizes that businesses should focus on how to effectively use the AI capabilities that are being developed, rather than on training larger models themselves.
• The author suggests that the key challenge for businesses will be in managing the risks associated with the use of these AI systems, rather than just on the performance of the models.