Wikipedia cofounder Jimmy Wales says ChatGPT is “pretty bad”

Jimmy Wales, the cofounder of the online non-profit encyclopedia Wikipedia, highlighted the limitations of generative artificial intelligence models such as ChatGPT during his session at the opening night of the Web Summit 2023 being hosted in Portugal. Speaking to a full house, Wales highlighted how he found ChatGPT an amazing thing to play with but also feels once you get serious with it, the realization sinks in that it is “pretty bad”. The opinion of the Wikipedia cofounder echoes many other researchers and entrepreneurs who are still not sold on the abilities of the large language model (LLM) and believe its mistakes make it an unreliable work assistant.

During the event, which can be watched on the official YouTube channel of the Web Summit, Wales was asked by Ryan Heath, senior journalist at Axios, “Are generative AI companies getting the scrutiny they deserve?” To this, Wales says, “One of the things that we have seen is that once a new technology comes up, there is a lazy alarmism that happens”.

Highlighting an example from the early days of eBay, Wales recounted how people would raise an alarm whenever a controversial ad came up, but slowly people understood that it could just be reported and taken down and no illegal business was being carried out. Similarly, he also gave an example of when Anna Nicole Smith, a US model and TV personality, died in 2007, and someone vandalized her Wikipedia page briefly, which led to multiple news publications calling Wales asking about the flaw in the Wikipedia model. Relating it to the original question asked, Wales says the scrutiny of generative AI is part of this alarmism for an emerging technology.

Jimmy Wales thinks ChatGPT is just not good enough

Further, addressing the AI chatbot and not shying away from sharing his opinions, Wales said, “When we look at the emergence of ChatGPT, clearly it is not good enough. It is an amazing thing to play with but when you really start to use it, what initially seemed fantastic seems pretty bad. And it is getting a lot of scrutiny. I don’t think there is anyone in this room who has not heard of even one story (involving its mistake) of ChatGPT”.

Wales also highlighted that chatbots could be made more responsible and factually accurate, which can then be deployed into various fields if the data is heavily filtered and low-quality sources such as social media are avoided.