And while the company is still working to make the additional features from the ChatGPT-4o demo a reality, its CEO is already looking ahead to what's next.
Sam Altman is speaking at the Aspen Ideas Festival (via Decoder) about ChatGPT-5.
Altman did not want to go into details about the next evolution of LLM, but Altman says it offers "a big leap forward."
"I expect it to be a big leap," he added, "because a lot of what GPT-4 gets wrong is that it doesn't reason very well or sometimes goes completely off the rails and makes dumb mistakes that not even a six-year-old would make.
In May, Altman told a Stanford University lecture that "GPT-4 is the dumbest model any of you need to use," and even called the typical LLM "embarrassing at best."
"It is important to ship early and often, and we believe in iterative deployment. If we build AGI in a basement and the whole world is walking around in a blur with blinders on, I don't think that makes us a very good neighbor," he said at the time.
GPT-4o was then unveiled in a very impressive fashion, with the model solving written equations and showing emotional and conversational responses. In fact, the demonstration was so impressive that Google's Deep Mind had Project Astra react to it.
Comments