I find this (somewhat) treatable problem a good way to start thinking about alignment issues. Here’s a bit of my bloomberg column:
more to the point, they are now autonomous AI Agents, who in turn can create their own autonomous AI Agents. Therefore, it will not be possible to allocate all AI revenue to its human or corporate owners, as in many cases there will be none.
And to continue with the analysis:
One option is to let AI bots run untaxed, like bees do. At first, that might make life easier for the IRS, but a tax arbitrage problem will arise. Tax-free AI work would have a pronounced competitive advantage over its taxed human counterpart. Also, too many AIs will be released in the commons. Why have an AI and pay taxes when you can program it to do your bidding, relinquish ownership and enjoy your services tax free? It seems easy enough to give up ownership of autonomous bots, especially if you are producing your own autonomous bots. At the very least, he could sell them to shell corporations.
The obvious alternative is to tax the work of the AI. Working AIs would have to file tax returns, which they might do in the very near future. (Can they claim deductions for their AI babies? What about their investments?)
Since AIs don’t enjoy leisure like humans do, their work should arguably be taxed at a higher rate than humans. Still, AIs shouldn’t pay too much tax. With prohibitively high tax rates, AIs will have less stock of wealth to invest in improving themselves, which in turn would reduce long-term tax revenue from AI work. Yes, they are AI, but the incentives are still important.
Some people may fear that super-smart, super-patient AIs will accumulate too much wealth, either through investment or work, and thus have too much social influence. That would create a case for a wealth tax on AIs, in addition to an income tax. But if AIs are such good investors, humans will also want the social benefits that flow from such wisdom, and that again means tax rates well below the confiscatory level.
And here is one of the deep problems with AI taxes:
The fundamental problem here is that AIs can be very good at providing in-kind services: improving organizational software, answering emails, etc. It’s already a problem for the tax system when neighbors exchange services, but AIs will take this type of relationship to a much larger scale.
Forget hiring AI, really: What happens if you invest in them, tell them to do your bidding, repudiate your ownership, and then let them run much of your business and life? You could write off your investment in the AI as a business expense and subsequently receive in-kind services tax-free, in what would amount to a de facto act of exchange.
Here’s a general problem:
A major theme in AI circles is “alignmentnamely, whether humans can count on AI agents to do our bidding, rather than mount destructive cyberattacks or destroying us. These investments in alignment are necessary and important. But the more successful the humans in the alignment become, the bigger the problem with tax arbitrage becomes.
Is not easy!