Wikipedia Founder Says AI Is a ‘Mess’ Now but Can Become Superhuman in 50 Years
ChatGPT, the wildly popular generative artificial intelligence (AI) tool from OpenAI, is currently a “mess” when it’s used to write articles on Wikipedia, the platform’s founder Jimmy Wales tells Euronews Next.
A Wikipedia article written today with ChatGPT-4 is “terrible” and “doesn’t work at all,” he says, because it “really misses out on a lot and it get things wrong, and it gets things wrong in a plausible way and it makes up sources and it’s a mess”.
He even goes as far as to predict that superhuman AI could take at least 50 years to achieve.
But while he believes it is possible for AI to surpass humans in the distant future, it is more likely that AI tools will continue to support intellectual activities, despite being in an early phase at the moment, he says.
The most valuable start-up in the United States, OpenAI catapulted onto the scene with its chatbot ChatGPT last year.
The technology takes instruction and questions and answers them with an eerily human-like response based on sources it gathers online. It can be used for writing essays, song lyrics, or even health advice, though it can often get the information wrong, known as “hallucinating”.
But even the most powerful chatbot AI start-up was thrown into chaos with the ousting of its CEO and co-founder Sam Altman last week and then his rehiring just days later after employees threatened the board they would quit en masse.
Wales said it is “worrisome” that this occurred for such an influential company but that it will “probably pass as if nothing happened”.
If anything, he said the company will likely get its house in order and that it is “a good lesson to start-ups of all kinds that you really do have to think even at a very early stage about governance, about the stability of decision making”.
Despite his criticism of current generative AI models, Wales has not ruled out AI being used for Wikipedia.
He said if a tool were built to find mistakes in a Wikipedia article by comparing it to the sources it uses, it could help to iron out inaccuracies.
He even told Euronews Next that he would consider a Wikipedia venture with an open-source AI company that is freely usable to match Wikipedia’s principles, but clarified there is nothing specific in the works.
However, he says that this would be a decision that would not be taken lightly.
“Most businesses, not just charities like us, would say you have to be really, really careful if you’re going to put at the heart of your business a technology that’s controlled by someone else because if they go in a different direction, your whole business may be at risk,” he said.
He would therefore think carefully about any partnerships but added that he was open to pilot programmes and testing models.
Wikipedia is still essential to generative AI as it sources information published online to produce content. Therefore, the online encyclopedia must be accurate and not produce bias, something that both AI and Wikipedia have been accused of.
