The difference is emblematic of the difficulty in getting attention for climate mitigation. AI succeeds because you can sell a service to an individual human which will give them advantages over other humans. Climate change mitigation fails because you are trying to sell a service to humanity which will result in a better end state over some other hypothetical imagined future. Humans make decisions, not humanity, and many of them are pretty bad with both hypotheticals and imagination. It's no wonder that a product designed to make them do better at what they do, right now is more successful than one designed to make everybody do better than what would otherwise have resulted, 50-100 years in the future when they'll likely be dead.

Any kind of workable solution to large, societal-level problems needs to deal with the principal agent issue. Society doesn't actually exist; humanity doesn't actually exist. These are abstractions we use to label the behavior of individual people. You need to operate on the level of individual people to get any sort of outcome.

(FWIW, this is a major reason why concepts like markets, capitalism, democracy, rule of law, and federalism have been successful. They work by aligning incentives so that when one person takes an action that is good for themselves, they more-or-less end up benefitting the people around them too.)