Abstract
The speaker discusses the current state of the AI bubble and its possible future trajectory, drawing comparisons to the dot-com bubble of the late 1990s and early 2000s. While there is a lot of hype around AI, the speaker argues that the real economic impact is still minimal and that significant improvements in AI models are not likely to occur in the near future. The speaker suggests that the current Transformer architecture is likely to reach a limit, and that real progress will require new architectural innovations. They also discuss the role of governments in AI development, arguing that they are unlikely to invest heavily in AI development in the near future. The speaker concludes by suggesting that the next 20-30 years will be a period of significant change in the field of AI, but that the exact path is uncertain. – AI-generated abstract.