By Lisa Eadicicco, CNN
Photo: AFP / JONATHAN RAA
This week, OpenAI has launched GPT-5 - an upgraded version of the AI model behind ChatGPT that the company claims is significantly faster and more capable than its predecessor.
The AI giant has faced increased competition, amid growing concerns about AI's impact on mental health and future jobs.
GPT-5, which launched across OpenAI's free and paid tiers, will make ChatGPT better at tasks like writing, coding and answering health-related questions, OpenAI claimed. It also promised to prevent the popular chatbot from hallucinating as often and deceiving users when it could not answer a question.
New models like GPT-5 are important, because they determine how services like ChatGPT function and the new capabilities they will support, making them an essential part of its future direction.
OpenAI chief executive Sam Altman said GPT-5 was a "significant step" on the path toward artificial general intelligence, a hypothetical point at which AI can match human-level thinking.
"GPT-4 felt like you're kind of talking to a college student," Altman said. "GPT-5 is the first time that it really feels like talking to an expert in any topic, like a PhD-level expert."
In addition to ChatGPT, the new model was also available for developers who built tools and services based on OpenAI's technology.
Software development was a big area of focus for GPT-5. Altman said the model could generate "an entire piece of software" for you - a practice colloquially known as "vibe coding".
In a demonstration, OpenAI showed how ChatGPT could create a website for learning French, after typing in a prompt asking it to do so. Altman said use cases like this would be a "defining part of the new GPT-5 era".
GPT-5 had arrived, as coders increasingly used AI to handle certain parts of their job. Meta chief executive Mark Zuckerberg previously said he expected about half the company's code to be written by AI next year.
Twenty to 30 percent of Microsoft's code is written by AI, chief executive Satya Nadella said. Anthropic chief executive Dario Amodei sparked fears over AI impacting jobs in May, when he said he believed the technology could lead to a spike in unemployment.
OpenAI's new model also hit on another concern around the use of AI - it could be deceptive, as research from Anthropic and AI research firm Apollo Research had shown, or provide incorrect information.
Previously, the system would say it could do a task that it could not complete or refer to an image that was not there, said OpenAI safety research lead Alex Beutel.
With GPT-5, the company said it had trained the model to be honest in these types of scenarios.
GPT-5 would also be more careful about how it answered questions to queries that could be potentially harmful. While the company did not provide a specific example of how that would look in practice, it said the model would aim to give an answer that was helpful "within the constraints of remaining safe", which may involve giving high-level answers that are not too specific.
The update came after concerns were raised about people becoming too reliant on AI assistants, particularly emotionally, raising questions about the technology's impact on mental wellbeing. A man in Idaho, for example, told CNN that ChatGPT sparked a spiritual awakening for him, after he began discussing topics like religion with the chatbot.
His wife said it put a lot of strain on their family.
OpenAI is widely considered the frontrunner in AI, thanks to ChatGPT, which is now on track to hit 700 million weekly active users, but the competition continues to grow, especially among younger users.
Research from web analytics company SimilarWeb suggested AI search app Perplexity, DeepSeek, Anthropic's Claude and xAI's Grok all had higher app usage among 18-34 year olds. Zuckerberg is attempting to snatch up top-tier AI talent with reported multimillion dollar pay packages to get ahead in the AI race.
The launch also came at a period of growth and expansion for AI. The tech giant now played a bigger role in education and the government in the United States, recently striking a partnership with classroom software provider Instructure and launching a study mode for ChatGPT.
OpenAI had also worked with the Trump administration on projects like the US$500 billion ($NZ840 billion) AI infrastructure project known as Stargate. Altman had appeared in Washington several times and OpenAI recently confirmed planned to open its first office in the US capital city.
- CNN
