Page 1 of 1

GPT-4o is out

Posted: Tue May 14, 2024 1:29 pm
by Kraut
OpenAI Releases World's Best AI for FREE (GPT-4o)*

free at
You can interact with the AI voice over a text and study it together
9:20 real time conversational speech

Re: GPT-4o is out

Posted: Tue May 14, 2024 2:54 pm
by emk
FYI, GPT-4o looks pretty impressive, particularly the improved speech recognition and the video input.

Right now, OpenAI's Whisper is pretty solid for automatic transcription of clear Spanish and French TV shows. This looks like it might be a significant step beyond that in some ways, because the speech recognizer is integrated into the main AI, and because it can use videos for context (though it's not clear whether it would be able to produce subtitle timing).

You can see some demos here. The most interesting two videos for language learners are "Real time translation" and "Point and learn Spanish."

Depending on how strong it is in different languages, and how soon they make it available via the API, I'm pretty sure it will be useful for something. But as my Spanish log shows, the big tricks are being very aware of the model's weaknesses, and using the model to remove certain kinds of human tedium in an already successful learning technique. You can't just slap "AI" on something and expect magic.