The cost to train and run these things is going to lead to fewer players eventually. I suspect we end up with 2 or 3 big players in 10 years.

1. Open source models are already at Sonnnet 4.5 levels. For a lot of people’s needs, that's going to be sufficient.

2. Costs will come down as more efficient AI hardware continues to roll out, and once demand eventually catches up with supply in the coming years.

3. So super low cost (or free ad-supported) options will exist, and people will only pay more (in money or ads) for superior quality.

… unless training sources become pay to play?