AI in its current deep-learning form tends to reward concentration. The hardest part of building an AI company is collecting training data; big tech companies already have oodles of that, and the products, userbase, and infrastructure to collect more.
There may be some new big tech companies as AI lets software push into niches that haven't been computerized yet, but the same market dynamics will likely play out there, leading to one big winner per corpus. It's not economical to collect the same corpus more than once, and right now the first player to collect it has no incentive to share it.
The only thing I can think of that would reverse this would be legal regulations that either make all data public (unlikely), or that grant ownership of them to the person they're about rather than the corporate entity that collects them, or a technological solution that can enforce the latter while still allowing large-scale aggregation and machine-learning over multiple people.
There may be some new big tech companies as AI lets software push into niches that haven't been computerized yet, but the same market dynamics will likely play out there, leading to one big winner per corpus. It's not economical to collect the same corpus more than once, and right now the first player to collect it has no incentive to share it.
The only thing I can think of that would reverse this would be legal regulations that either make all data public (unlikely), or that grant ownership of them to the person they're about rather than the corporate entity that collects them, or a technological solution that can enforce the latter while still allowing large-scale aggregation and machine-learning over multiple people.