Jornal Do Brasilia - OpenAI releases faster model to power ChatGPT

The update to OpenAI's flagship product landed a day before Google is expected to make its own announcements about Gemini, the search engine giant's own AI tool that competes with ChatGPT head on.

"We're very, very excited to bring GPT-4o to all of our free users out there," Chief Technology Officer Mira Murati said at the highly anticipated launch event in San Francisco.

The new model will be rolled out in OpenAI's products over the next weeks, the company said, with paid customers having limitless access to the tool.

The company said that the model could generate content or understand commands in voice, text, or images.

Murati and engineers from OpenAI demonstrated the new powers of GPT-4o at the virtual event, asking questions and posing challenges to the beefed-up version of the ChatGPT chatbot.

"We know that these models get more and more complex, but we want the experience of interaction to actually become more natural, easy," Murati said before the demo.

The demo mainly featured OpenAI staff members asking questions to a human-sounding ChatGPT in Italian and asking the bot to interpret facial expressions or make complex math equations.

The company said that GPT-4o had the same powers as the previous version when it came to text, reasoning and coding intelligence, and set new benchmarks for multilingual conversations, audio and vision.

In the demonstrations, ChatGPT successfully interpreted an employee's surroundings, speaking in a friendly, feminine voice.

"Hmmm from what I can see it looks like you're in some kind of recording or production set-up with lights, tripods… you might be gearing up to shoot a video or make an announcement?" the ChatGPT bot said.

Hinting at GPT-4o's powers, OpenAI CEO Sam Altman told the All-In podcast last week that he wanted AI interactions to resemble a relationship with "a great senior employee" who could push back and wasn't a "sycophant."

- 'Take our time' -

Anticipation was high in recent weeks that OpenAI would release an AI-amped version of an online search tool to compete with Google search engine.

Observers were also waiting for the launch of GPT-5, but Altman said that his company would "take our time on releases of major new models" and could choose to just update the technology more gradually.

The event is just the latest episode in the AI arms race that has seen OpenAI-backer Microsoft propelled past Apple as the world's biggest company by market capitalization.

OpenAI and Microsoft are in a heated rivalry with Google to be generative AI's major player, but Facebook-owner Meta and upstart Anthropic are also making big moves to compete.

All the companies are scrambling to come up with ways to cover generative AI's exorbitant costs, much of which goes to chip giant Nvidia and its powerful GPU semiconductors.

Making the new model available to all users, may raise questions about OpenAI's path to monetization.

Until now, only less performing versions of OpenAI or Google's chatbots were available to customers for free, with doubts that everyday smartphone users are ready to pay a subscription to maintain access to the technology.

The AI makers are also feeling pressure from creators, who are demanding payment for the content used to train the models.

OpenAI has signed content partnerships with the Associated Press, the Financial Times and Axel Springer, but is also caught in a major lawsuit with The New York Times.

AI companies have also been confronted with separate lawsuits from artists, musicians and authors in US courtrooms.

L. de Freitas--JDB