One thing I have noticed (it could be only my assumption) is that OpenAI is quite secretive and generally downplay their recent releases, ChatGPT and GPT-4.
What I mean to say is, they often talk about “surprised at the popularity”, but I don’t think its that surprising at all, and they clearly knew they have found something that would change the AI landscape.
And at the same time, they can do other experiments because they keep weights and other details about the model secret.
I doubt they haven’t already found something far more potent than GPT-4, but are simply planning years ahead of how to trickle the downstream of releases so it makes sense for the public eye and their self-interest.
What I mean to say is, they often talk about “surprised at the popularity”, but I don’t think its that surprising at all, and they clearly knew they have found something that would change the AI landscape.
And at the same time, they can do other experiments because they keep weights and other details about the model secret.
I doubt they haven’t already found something far more potent than GPT-4, but are simply planning years ahead of how to trickle the downstream of releases so it makes sense for the public eye and their self-interest.