This. The above said, the learning algorithm has flaws that can make Pareto optimality impossible (economists call this market failure). This is where we inje

This. The above said, the learning algorithm has flaws that can make Pareto optimality impossible (economists call this market failure). This is where we inject aspects of central planning into the modern economy, with debatable results. For example, it presumes a level of information symmetry through the price mechanism. Markets begin to fail when people don't/can't understand what they're buying - the "market for lemons". This is why we have an SEC and mandatory , standardized, information filing for publicly traded companies for example. Another is high levels of uncertainty about the sudden need to buy something, which is so expensive it can wipe you out - eg. Major surgery. Insurance markets cover this. But we all have observed the debates and challenges of managing a sizeable and representative risk pool. Intellectual property and digitized products (music, movies, TV shows, software) also has a fascinating role here in that it is a public good that is open to free ridership unless (a) the good is a secret and is accessed only as a service or (b) the government enforces a temporary monopoly on an aspect of the product. The latter may be past its prime given the abuse of patents and copyrights leading to non-Pareto outcomes.