ChatGPT pioneer Sam Altman warns against blind trust in AI: ‘It hallucinates’

OpenAI CEO Sam Altman says uncritical reliance on AI tools like ChatGPT could have major risks; he was speaking on the company’s official podcast.

OpenAI CEO Sam Altman, among the most prominent innovators behind the surge of Artificial Intelligence, has warned of AI’s hallucinations. Referring to OpenAI’s marquee product ChatGPT, he warned users to not trust AI with “almost everything”.

Talking about AI's deceptive nature, he stressed on how AI often gives fake yet well-crafted responses to prompts. (Getty Images via AFP)

Altman said that he finds it interesting when people put a high degree of trust in ChatGPT.

Speaking in the first episode of OpenAI’s official podcast, he said, “People have a very high degree of trust in ChatGPT, which is interesting, because AI hallucinates. It should be the tech that you don't trust that much.”

Drawing from his life, he explained how he has been using ChatGPT to find details on everything, from what to do for diaper rashes to nap routines for babies after he became a new parent.

Altman also addressed OpenAI's privacy concerns as he spoke about recent updates, such as a potential ad-supported model. He said such advancements have raised fresh concerns.

AI hallucination that Sam Altman fears

When an AI model starts generating inaccurate or, simply put, misleading data with reassurances, one can say the AI tool is hallucinating. Some of these outputs have no relevance to the actual thing in question. As Altman reiterated, blind reliance on AI like this could cause major risks for individuals who trust it with anything and everything.

“We need to be honest about that… “It’s not super reliable,” he said

This statement from Altman comes also in light of lawsuits filed by media organisations against OpenAI over infringement of their intellectual property. According to a Reuters report, The New York Times sued OpenAI and Microsoft in 2023, accusing them of using millions of its articles without permission to train the large language model behind its popular chatbot ChatGPT.

Altman changes stance on hardware

While earlier Altman had stressed that the advent and widespread use of AI wouldn't require new hardware, he contradicted himself on the podcast. He underlined that current computers were designed for a world without AI.

He suggested that users will need new devices as AI becomes more prevalent.

With inputs from Reuters