OpenAI Races Toward AGI With Next-Gen AI Model in Training

OpenAI revealed that it has started training its next frontier AI model. The announcement came in a post published on their website on May 28th, with no word on a planned release date yet.
The company claims that the new model will bring “the next level of capabilities” as it aims to reach artificial general intelligence (AGI) levels in the future.

AGI is a hypothetical type of AI so advanced that it could outperform humans in a wide range of tasks, even those it was not explicitly trained for, as opposed to the current AI models that have a relatively narrow scope.

While it’s not clear whether the new model announced will be called GPT-5 or something else, it is understood that it aims to build on the success of GPT-4 and its latest iteration, GPT-4o, which was recently released.

OpenAI’s current flagship model brought the ability to analyze images and perform real-time language translations. The company has also teased additional features that are expected to be released soon, including improved voice responses and the ability to use a phone’s camera to get real-time help.

Developing and training AI models takes a lot of time. According to the New York Times, it may take nine months to a year (or more) for GPT-4o’s successor to go live. ZDNET argues that the new model is expected to bring improved accuracy in its responses, increased multimodality, and even be able to act autonomously to some degree.

Users are already speculating about the new model’s potential and sharing their expectations on OpenAI’s website. According to this discussion thread, some of the features they are eager to see include the ability to produce a wider range of outputs, solve equations, override its training data, and act like an agent.