Writing in a Working Paper for the Institute for New Economic Thinking Servaas Storm from the Delft University of Technology explains how and why the U.S. is inflating an AI-bubble, and concludes this will end badly.
The paper makes three main points:
- We have reached peak GenAI in terms of LLMs. Scaling will take us no further towards Artificial General Intelligence (AGI) and returns on existing technology are diminishing.
- The AI-LLM industry represents a speculative bubble about to burst due to 1.
- The bet has been fueled by the notion the U.S. has to ‘own’ AGI because it can’t concede this win to China. The bet is now souring.
For most this (below) may be the most apposite section i.e. the one that deals with corporate AI adoption.
The key point about bubble risk is in the next two charts. The first shows the staggering amount of money being spent on unsure outcomes. The second shows returns from this spending are already diminishing.
Nobody has found a way to monetize the existing technology (only 5% of users pay for ChatGPT) let alone having a viable model for monetizing future gains, if indeed these are possible?
The paper concludes bleakly that if the U.S. doesn’t change tack in terms of this profligacy to nowhere “..the U.S. economy is surely doomed.” Inflammatory conclusion aside, the work is a great catalogue of many useful facts in plain-and-ugly sight. Consider yourself warned.
Review the work via this link The AI Bubble and the U.S. Economy.
Happy Sunday.