Skip to main content

ChatGPT creator seeking to eliminate chatbot ‘hallucinations’

Despite all of the excitement around ChatGPT and similar AI-powered chatbots, the text-based tools still have some serious issues that need to be resolved.

Among them is their tendency to make up stuff and present it as fact when it doesn’t know the answer to an inquiry, a phenomenon that’s come to be known as “hallucinating.” As you can imagine, presenting falsehoods as fact to someone using one of the new wave of powerful chatbots could have serious consequences.

Close up of ChatGPT and OpenAI logo.
Image used with permission by copyright holder

Such trouble was highlighted in a recent incident in which an experienced New York City lawyer cited cases — suggested by ChatGPT — that turned out never to have happened. The lawyer may face sanctions as a result of his action.

Another incident received widespread attention in April when ChatGPT apparently rewrote history by saying that an Australian mayor had been jailed for bribery while working for a bank when in fact he’d been a whistleblower in the case.

To make its chatbot technology more reliable, OpenAI engineers have revealed that they’re currently focusing on improving its software to reduce and hopefully eliminate these problematic occurrences.

In a research paper released on Wednesday and picked up by CNBC, OpenAI said that chatbots “exhibit a tendency to invent facts in moments of uncertainty,” adding: “These hallucinations are particularly problematic in domains that require multi-step reasoning since a single logical error is enough to derail a much larger solution.”

To tackle the chatbot’s missteps, OpenAI engineers are working on ways for its AI models to reward themselves for outputting correct data when moving toward an answer, instead of rewarding themselves only at the point of conclusion. The system could lead to better outcomes as it incorporates more of a human-like chain-of-thought procedure, according to the engineers.

But some experts expressed doubt about the work, telling CNBC it’s of little use until it’s incorporated into ChatGPT, which in the meantime will carry on hallucinating. OpenAI hasn’t said if and when it might incorporate its work into its generative AI tools.

While it’s good to know that OpenAI is working on resolving the issue, it could be a while before we see any improvements. In the meantime, as OpenAI itself says, ChatGPT may occasionally generate incorrect information, so be sure to confirm its responses if they’re part of any important tasks.

Editors' Recommendations

Trevor Mogg
Contributing Editor
Not so many moons ago, Trevor moved from one tea-loving island nation that drives on the left (Britain) to another (Japan)…
ChatGPT is violating your privacy, says major GDPR complaint
ChatGPT app running on an iPhone.

Ever since the first generative artificial intelligence (AI) tools exploded onto the tech scene, there have been questions over where they’re getting their data and whether they’re harvesting your private data to train their products. Now, ChatGPT maker OpenAI could be in hot water for exactly these reasons.

According to TechCrunch, a complaint has been filed with the Polish Office for Personal Data Protection alleging that ChatGPT violates a large number of rules found in the European Union’s General Data Protection Regulation (GDPR). It suggests that OpenAI’s tool has been scooping up user data in all sorts of questionable ways.

Read more
Google Bard could soon become your new AI life coach
Google Bard on a green and black background.

Generative artificial intelligence (AI) tools like ChatGPT have gotten a bad rep recently, but Google is apparently trying to serve up something more positive with its next project: an AI that can offer helpful life advice to people going through tough times.

If a fresh report from The New York Times is to be believed, Google has been testing its AI tech with at least 21 different assignments, including “life advice, ideas, planning instructions and tutoring tips.” The work spans both professional and personal scenarios that users might encounter.

Read more
ChatGPT may soon moderate illegal content on sites like Facebook
A laptop screen shows the home page for ChatGPT, OpenAI's artificial intelligence chatbot.

GPT-4 -- the large language model (LLM) that powers ChatGPT Plus -- may soon take on a new role as an online moderator, policing forums and social networks for nefarious content that shouldn’t see the light of day. That’s according to a new blog post from ChatGPT developer OpenAI, which says this could offer “a more positive vision of the future of digital platforms.”

By enlisting artificial intelligence (AI) instead of human moderators, OpenAI says GPT-4 can enact “much faster iteration on policy changes, reducing the cycle from months to hours.” As well as that, “GPT-4 is also able to interpret rules and nuances in long content policy documentation and adapt instantly to policy updates, resulting in more consistent labeling,” OpenAI claims.

Read more