OpenAI’s newest update to its artificial intelligence model is capable of mimicking human cadences in its verbal responses and can even attempt to figure out people’s moods. While some could find the new model a minor update, OpenAI claims that it does work faster compared to its precursors and can reason across text, video, and audio in real-time.
GPT-4o, short form for “omni,” will power OpenAI’s popular ChatGPT chatbot, and will be available to users, including those who do not have a subscription, in the coming days. The company made this announcement during a short live-streamed update. CEO Sam Altman, who wasn’t present among the presenters at the event, posted the word “her” on X (formerly known as Twitter).
During a demo with Chief Technology Officer Mira Murati and other executives, the AI bot chatted in real-time. It also helped me walk through the steps required to solve a simple math problem without giving the answer straightaway and assisted with a more complex software coding problem on a computer.
Also Read: Microsoft Developing New AI Model To Rival Google, OpenAI
OpenAI On Its Way To Challenge Bigger Rivals
The new AI model also tried extrapolating a person’s emotional state by looking at a selfie video of the person’s face and translating English and Italian to show how it is capable of helping people who speak different languages can hold a conversation. Gartner analyst Chirag Dekate said the update suggested the impression that OpenAI is playing catch-up to larger rivals.
“Many of the demos and capabilities showcased by OpenAI seemed familiar because we had seen advanced versions of these demos showcased by Google in their Gemini 1.5 pro launch,” Dekate said. “While Open AI had a first-mover advantage last year with ChatGPT and GPT3 when compared to their peers, especially Google, we now are seeing capability gaps emerge.”
Google has plans to hold its I/O developer conference soon, where the company is expected to take the wraps off of its own Gemini AI model.
Also Read: Analyst Claims GTA 6 Will Not Release In 2025 As Promised
Did OpenAI Use YouTube To Train Sora?
OpenAI created a stir across the AI industry when it introduced SOra, its video AI model, earlier this year. However, the company still hasn’t confirmed whether the videos that were used to train the AI model were from YouTube or not.
During an interview at the Bloomberg Technology Summit, OpenAI’s COO Brad Lightcap talked about the potential business applications of artificial intelligence. And, undoubtedly, Sora is right there among the potential use cases. However, Lightcap did not give a clear answer when asked about whether YouTube videos were used to train the company’s Sora.