The real improvement is the multimodality. Processing Image, sound and text all at the same time. That alone might be able to upgrade its intelligence but we dont know yet.
We do not have access to what we saw in the demo. the only thing that got released is a gpt4o that is limited to text only which feels like a more refined version of gpt4 but not more powerful (more frequent succes but not higher scores)
If you use image input/dalle voice then it defaults to normal gpt4 which uses a transcription of your words as input rather then true audio.
Funny how the increasingly underwhelming ai releases still get hyped to oblivion like each one is gpt2 to 3 again.
This is an incremental at best improvement, if not basically the same thing but people assume it will be better and see what they want to see.
The real improvement is the multimodality. Processing Image, sound and text all at the same time. That alone might be able to upgrade its intelligence but we dont know yet.
We do not have access to what we saw in the demo. the only thing that got released is a gpt4o that is limited to text only which feels like a more refined version of gpt4 but not more powerful (more frequent succes but not higher scores)
If you use image input/dalle voice then it defaults to normal gpt4 which uses a transcription of your words as input rather then true audio.
I’ll believe it when I see it. Openais track record is lying about capabilities and letting hype inertia smooth it out for them.