The argument regarding the economic impact of AI, particularly in the realm of white-collar jobs, is evolving. Many experts, including those at Citadel Securities, challenge the narrative that automation will lead to widespread job displacement and economic collapse. Their analysis suggests that while AI technologies are advancing rapidly, the pace at which companies adopt these systems is not guaranteed to match this speed.
A key point made by Citadel's economists is the concept of the compute-cost ceiling. If companies attempt to automate en masse, the resulting surge in demand for computing power could lead to increased costs, making it more economical to retain human workers. This perspective offers a counter to the prevailing notion that AI will swiftly render many jobs obsolete.
However, this view may underestimate a critical factor: the rapid decline in computing costs. While traditional metrics like Moore's Law have slowed, the costs associated with AI, particularly in large language model inference, have been decreasing significantly—by as much as a factor of 10 annually over the last two years. This trend, termed LLMflation, suggests that the economic landscape surrounding AI automation is more complex than previously assumed.