I suspect the Flynn effect is a measure of an education system more than actual intelligence. Will AI improve education? It is my opinion that it won't, at scale, because of the opportunities it provides to game the score based education system in place today. If we switched to a more application based education system, it might prove to be a fantastic multiplier. But the major gains wouldn't really be about AI, then.

AI could potentially be a menial assistant in the search for a better definition of intelligence, its key mechanisms and improved metrics. For example, ADHD medication improves working memory by increasing neurotransmitters involved in filtering the noise that interferes with the content stored in short term working memory "slots". That doesn't increase the number of working memory slots, but is a crude method to allow more efficient use of those slots, increasing processing capacity and speed - potential facets of intelligence, but light years from being the whole of intelligence. Could AI theoretically help more quickly sift through studies and data looking for potential paths of inquiry that might lead to more sophisticated increases in efficiency or increasing the overall slots in working memory? Or help understand the preconscious processes where a piece of new data is connected to multiple pieces of stored data to create an Aha! at the moment the composite enters consciousness? Possibly, but it might need better human data to learn from.