It seems text based GPT from openai can no longer go forward, only sideways. 
The level of thinking (I know it doesn't really think but let's assume what it does is thinking) has been at same height it has been for a long time. 
Sideways improvements include better context window, better memory, more specific models (primed to a certain context eg. coding model, research model, etc). However the underlying tech seems to not be improving much (or even at all).

I predict the next breakthrough in the AI space will not be from them. 

That side audio and especially video models seems to have a large space for improvements (for everyone not just openai). Looking to see how far we can go with those models.

@blockbot Steward. Frames Builder. Pro Chess-er ♟️


I meant image and video models. I guess audio counts as well just not much demand compared to those two