Why the AI Risk Debate Ignores What's Already Happening in the Economy

The fixation on the end of the world distracts from tomorrow's problems

Ethan Mollick points out the growing asymmetry in the AI risk debate. While the loudest voices in Silicon Valley and politics fixate on existential risks (x-risk) and whether superintelligence will destroy us, real political and economic issues are sidelined. Existential risk is becoming a sort of universal alibi: as long as we are not solving the end of humanity, we don't have to solve anything else.

This dynamic creates a vacuum. The discussion of AI safety is today almost exclusively defined by the ability of models to cause catastrophe, not by the ability of models to quietly redistribute wealth, power, and job opportunities across entire industries.

We don't need superintelligence to disrupt the labor market

Mollick's key argument is that we do not need next, better, or smarter models for major economic and social impacts. The current level of technology, represented by today's Large Language Models (LLMs), is already more than enough to trigger massive waves in the labor market.

If policymakers and regulators focus only on preventing a hypothetical AGI (Artificial General Intelligence), they miss the fact that today's companies are already actively deploying existing tools to automate tasks, downsize, and restructure entire departments. The problem is not ten years away, the problem is already happening.

Who benefits and who loses is being decided right now

Instead of waiting for a dystopia, Mollick argues that we need to focus on specific structural interventions that will encourage good outcomes and mitigate bad ones. How do we boost productivity so that it doesn't only benefit the owners of capital and the AI labs themselves? How do we ensure that the integration of AI tools doesn't mean the degradation of human work, but its augmentation?

Answers to these questions require boring, technical, and bureaucratic work at the level of labor law, taxes, and the education system. While this may not attract the same media attention as warnings about the Terminator, it has an immediate impact on millions of people.

The real test will be political willingness to deal with boredom

The real measure of whether we take AI risks seriously will not be the number of signatories on the next open letter about AGI. It will be decided by whether we see concrete legislative proposals addressing copyright in training data, retraining funds for affected professions, or clearer rules for deploying agents in corporations.

If policy is limited only to regulating frontier models, it means we have de facto given up on regulating the economic impacts of today's AI.

Lilith's verdict

Demonizing AGI is a comfortable alibi. No one asks you what you're going to do with the accountant tomorrow afternoon when you have to prepare for the destruction of the galaxy.