The problem with this calculus is that the AI exists to benefit their owners, the economy itself doesn't really matter, it's just the fastest path to getting what owners want for the time being.

This a late 20th century myopic view of the economy. In the ages and the places long before, most of human toil was enjoyed by a tiny elite.

Also "rendering such a future impossible". This is a retrocausal way of thinking. As though an a bad event in the future makes that future impossible.

Your first premise has issues:

>In such a future, people will have minimal income (possibly some UBI) and therefore there will be few who can afford the products and services generated by AI

Productivity increases make products cheaper. To the extent that your hypothetical AI manufacturer can produce widgets with less human labor, it only makes sense to do so where it would reduce overall costs. By reducing cost, the manufacturer can provide more value at a lower cost to the consumer.

Increased productivity means greater leisure time. Alternatively, that time can be applied to solving new problems and producing novel products. New opportunities are unlocked by the availability of labor, which allows for greater specialization, which in-turn unlocks greater productivity and the flywheel of human ingenuity continues to accelerate.

The item of UBI is another thorny issue. This may inflate the overall supply of currency and distribute it via political means. If the inflation of the money supply outpaces the productivity gains, then prices will not fall.

Instead of having the gains of productivity allocated by the market to consumers, those with political connections will be first to benefit as per Cantilion effects. Under the worst case scenario this might include distribution of UBI via social credit scores or other dystopian ratings. However, even under what advocates might call the ideal scenario, capital flows would still be dictated by large government sector or public private partnership projects. We see this today with central bank flows directly influencing Wall St. valuations.

If I may speculate the opposite: With cost-effective energy and a plateau in AI development, the per-unit cost of an hour of AI compute will be very low, however, the moat remains massive. So a very large amount of people will only be able to function (work) with an AI subscription, concentrating power to those who own AI infra. It will be hard for anybody to break that moat.

no the AI doesn't actually need to interact with world economy it just needs to be capable of self-substence by providing energy and material usage. But when AI takes off completely it can vertically integrate with the supply of energy and material.

wealth is not a thing in itself, it's a representation of value and purchasing power. It will create its own economy when it is able to mine material and automate energy generation.

Alternatively:

-- In such a future, people will have minimal income (possibly some UBI) and therefore there will be few who can afford the products and services generated by AI

-- Corporate profits drop (or growth slows) and there is demand from the powers that be to increase taxation in order to increase the UBI.

-- People can afford the products and services.

Unfortunately, with no jobs the products and services could become exclusively entertainment-related.