During the opening night of Web Summit in Lisbon, Portugal, Wikipedia founder Jimmy Wales told an auditorium of tech enthusiasts that ChatGPT's frequent errors make it "pretty bad". This is despite the AI-dominated discussions at the summit. Wales further conveyed that large language model aritificial intelligence (AI) is approximately three decades away from being capable of rivaling human capabilities, says a report by Tom's Guide.
This comes amid the growing poularity and concerns around OpenAI's ChatGPT that was launched about a year ago. Big Tech firms like Microsoft and Google are also betting big on AI and the latter has also launched its own conversational AI service Bard. Wales also mentioned that Microsoft's collaboration with ChatGPT parent OpenAI will be a problem for Google. He exhibited less enthusiasm for the near-term potential of large language models. He told the audience that, for many purposes, it falls short, describing it as enjoyable to experiment with but ultimately "pretty bad" when used seriously.
Wales also took a swipe at X, formerly Twitter, and SpaceX CEO Elon Musk and his social media platform saying that X is not a credible source for correct information, according to a report by Fortune.
At the summit, Wales further emphasised that generative AI won't be employed for writing articles outright. Instead, it could serve as a tool to enhance existing articles. For instance, AI could search for relevant information across various sources to identify statements missing from Wikipedia entries. He also suggested using AI to find statements present in specific articles that should be added to a Wikipedia entry, noting that even if it's 70 per cent accurate, it could significantly boost productivity. The problem with AI, according to Wales, is that the stuff it is good at and gets correct includes content already widely written about on Wikipedia and by humans.
“If you ask ChatGPT questions about someone famous or a historical topic it is much better but that isn’t where Wikipedia needs help,” he was quoted as saying by Tom's Guide.
“We have tonnes of people writing about the Battle of the Bulge or World War 2, what we need is help on more obscure topics, and on those, it makes a lot of mistakes.”
Notably, amid concerns of artificial intelligence impacting jobs and replacing humans, a study revealed last month that ChatGPT may be better at following recognised treatment standards for clinical depression than doctors. The study's findings were published in the open access journal owned by British Medical Journal, Family Medicine and Community Health.