OpenAI may debut a multimodal AI digital assistant quickly

OpenAI has been exhibiting a few of its prospects a brand new multimodal AI mannequin that may each speak to you and acknowledge objects, in accordance with a new report from The Information. Citing unnamed sources who’ve seen it, the outlet says this might be a part of what the corporate plans to show on Monday.

The brand new mannequin reportedly affords sooner, extra correct interpretation of photographs and audio than what its current separate transcription and text-to-speech fashions can do. It could apparently be capable of assist customer support brokers “higher perceive the intonation of callers’ voices or whether or not they’re being sarcastic,” and “theoretically,” the mannequin can assist college students with math or translate real-world indicators, writes The Info.

The outlet’s sources say the mannequin can outdo GPT-4 Turbo at “answering some varieties of questions,” however remains to be vulnerable to confidently getting issues mistaken.

It’s potential OpenAI can be readying a brand new built-in ChatGPT potential to make cellphone calls, in accordance with Developer Ananay Arora, who posted the above screenshot of call-related code. Arora additionally spotted evidence that OpenAI had provisioned servers meant for real-time audio and video communication.

None of this might be GPT-5, if it’s being unveiled subsequent week. CEO Sam Altman has explicitly denied that its upcoming announcement has something to do with the mannequin that’s imagined to be “materially better” than GPT-4. The Info writes GPT-5 could also be publicly launched by the top of the yr.