OpenAI announced a “more humane” version of ChatGPT
The AI tool “is particularly better at understanding video and audio compared to existing models.” The function includes recognition of emotions and breathing rhythm.
The chatbot also added a full-fledged Voice Mode for voice interaction.
According to the presentation, the product can help users with various everyday issues, such as preparing for an interview. OpenAI also showed GPT-4o calling support for a replacement iPhone.
Other examples have demonstrated that the neural network can tell “dad jokes,” translate conversations in two languages in real time, judge a game of rock-paper-scissors, and respond with sarcasm.
One of the videos showed how ChatGPT reacts to the user’s first meeting with a puppy.
"Hello, Bowser! Aren’t you the most beautiful creature?” - the chatbot exclaimed.
OpenAI said the “o” in GPT-4o stands for “omni,” which symbolizes a step toward more natural human-computer interaction.
GPT-4o is "much faster" and "50% cheaper" than GPT-4 Turbo. The neural network responds to an audio request in 2.3 seconds. The average chatbot generation time is 3.2 seconds, which is comparable to the reaction time of a person in a normal conversation, OpenAI emphasized.
A version of the neural network for entering only text and images was launched on May 13. A full version with video support will appear “in the coming weeks.”
“It's like the AI in the movies. It still surprises me a little that this is real. The move to human-level response time and expressiveness has been a big change,” said OpenAI CEO Sam Altman.
According to him, conversations with the neural network now seem “truly natural.”
Earlier, Altman noted that the future of artificial intelligence will be marked by the emergence and growing popularity of personal assistants. They will become an integral part of everyday life and even replace smartphones.
Let us recall that in April, the gpt2-chatbot language model appeared on the LMSYS platform , which is associated with OpenAI.