How will AI change government policy-making?

Linkpost from my substack: https://mkbaio.substack.com/p/ai-will-make-regulation-much-less

Excerpt:

One possible answer is that AI will vastly expand the production-possibility frontier of net-beneficial government policies.
For many, the most persuasive argument for libertarianism is avoiding the enormous deadweight loss that arises from government overregulation.
I claim that AI will change the cost-benefit trade-off in favor of more regulation. When AI can correctly fill out a complicated tax-form for just a few cents, the regulatory burden of complicated tax rules drops significantly.
When humans are the ones filling out forms, tax rules need to be (and too often fail to be) simple and fast to fill out. If AI can fill out the forms, adding a hundred edge-cases to internalize various positive externalities is now a reasonable idea!
Most can agree it would be insane to add a minuscule tax-benefit for those who opted to plant the most co2-reducing bushes their gardens. The time spent on claiming and administrating the benefits would far outweigh the social benefits of the slight co2 reduction encouraged.
With AI, complicated policies filled with quirks and edge-cases now look far more attractive. The inane policies that left me shrieking, all of a sudden don’t look so bad! As administration costs decrease, the possible policies that are net-beneficial radically increase.