TL;DR: It's bad it's not GPT-5. If AI could get substantially better, we'd have GPT-5 right now. If they had GPT-5 right now, they would have released GPT-5. They didn't. Therefore it's bad.
I'm a little surprised that's the Gary Marcus hot take on it, guess there's not much room for complaint.
n.b. this continues to follow every public roadmap statement OpenAI has made since GPT-4: they'd be focused primarily on making GPT-4 faster and more efficient for a while, while setting up GPT-5 training, intending release for late 2024/early 2025.
I'm a little surprised that's the Gary Marcus hot take on it, guess there's not much room for complaint.
n.b. this continues to follow every public roadmap statement OpenAI has made since GPT-4: they'd be focused primarily on making GPT-4 faster and more efficient for a while, while setting up GPT-5 training, intending release for late 2024/early 2025.