Derek Thompson argues that the substantive disagreement over AI is not about its current capabilities, but about whether or not it falls within the adoption patterns of past technologies.
In this essay, Derek Thompson identifies what he believes really separates the camps in the artificial intelligence debate: not the evaluation of model performance, but a preliminary question — is AI a technology comparable to previous ones, or is it a breakthrough with no historical analogue?
On the one hand, the so-called "normal AI" thesis, defended in particular by Arvind Narayanan and Sayash Kapoor, situates AI in the continuity of electricity, the automobile or the internet: speculative bubbles, redistribution of jobs, long period of regulatory uncertainty, slow diffusion in organizations. Three years after ChatGPT, economic growth remains ordinary and unemployment contained, which this camp reads as confirmation; François Chollet observes that AI remains incapable of operating without supervision.
On the other hand, some technologists anticipate a recursive improvement in models leading to superintelligence, and therefore to an unprecedented trajectory. Deciding between these two readings, the author points out, determines almost all the public policy positions that result from them.
Read the full article on the original source
Go to the source →