The OpenAI company presented the language model of artificial intelligence GPT-4o. The developers assured that the model will work in the same way as GPT-4, but much faster. About this reported the XDA.
GPT-4o also got new voice mode technology for direct communication with ChatGPT.
The letter "o" in the name is short for "omni" ("comprehensive"). The new version of the model works with speech, text and video. The GPT-4o responds to audio in an average of 320 milliseconds, which is comparable to the response time of a person in a conversation, the company said.
OpenAI always wanted to make a voice conversation with ChatGPT similar to a conversation with a real person, but the delay with the answer spoiled everything.
The company held a demonstration of a conversation with GPT-4o using voice. Not only did the GPT-4o respond almost instantly when the speaker finished talking. AI also responded with text-to-speech with a full sense of real-time communication.
In addition, OpenAI showed how GPT-4o can be used as a voice translator between two people who speak different languages.
In another demonstration, the GPT-4o language model told a bedtime story with the prompt "work and love." In the middle of the story, Mark from OpenAI intervened and asked GPT-4o to adjust the emotions. GPT-4o changed her voice depending on the questions, from a dramatic performance to a cold, robotic tone. In the end, GPT-4o even sang.
GPT-4o was also used to verify the code. The model was able to not only explain what the code does, but also tell what would happen if certain parts of it were changed.
Usually, when OpenAI announces a new version of its ChatGPT model, it is limited to paid access. However, the company has decided to let everyone use this new technology, although paid users will have much more options.
The new technology will be rolled out in the next few weeks. A ChatGPT desktop application with voice and visual capabilities will be rolled out soon.
To always be aware of the most important things, read us at Telegram
