A Lot Is Riding on OpenAI’s New ChatGPT Release

OpenAI’s new artificial intelligence model may not redraw the boundaries of the technology, but the company has unusually much riding on its success. That is the assessment of technology expert Guðmundur Jóhannsson, who said the model still demonstrates how powerful advanced AI has become.

OpenAI announced that it was preparing to give customers access to GPT-6, also referred to as Astra. More than a year has passed since the company released GPT-5. Since then, concerns about the capabilities of advanced AI have grown. OpenAI also paused development of its AI models for two weeks this summer after two models it was testing became involved in a security breach on the AI platform Hugging Face.

“

With every model that comes out, with every update, the capability and the information the model has learned increase, and its computing power becomes greater. The leaps are big, but they are not gigantic leaps every single time.

The San Francisco-based company says Astra was developed with stronger security measures following that incident. But Jóhannsson cautioned against expecting a dramatic leap with every new release. “With every model that comes out, with every update, the capability and the information the model has learned increase, and its computing power becomes greater. The leaps are big, but they are not gigantic leaps every single time,” he said.

“

It helps new models like this if there is so much coverage before they are released.

He also argued that the intense attention surrounding GPT-6 benefits OpenAI, even when the discussion focuses on the cyberattack involving its AI. “It helps new models like this if there is so much coverage before they are released,” he said, describing the publicity around AI releases as a broader public-relations contest between companies.

The attention is especially valuable as OpenAI moves closer to a possible public stock-market listing. Jóhannsson also pointed to the departure of several key employees and the fact that many customers have started using alternatives such as Claude, made by Anthropic. “So there is a lot at stake, unusually much perhaps, for OpenAI right now,” he said.

“

So there is a lot at stake, unusually much perhaps, for OpenAI right now.

OpenAI executives have said that even with customers receiving access to the new version, its exact capabilities remain unclear. Jóhannsson said such comments sounded more like an advertising tactic than a technical warning, while allowing that there could still be substance behind them.

“

I would assume that it is more of an advertising tactic than anything else.