Few websites are more foundational to the web than Wikipedia.
The online encyclopedia was founded almost 25 years ago by Jimmy Wales, who at the time was working on an expert-led project called Nupedia.
Nupedia didn’t work out, but its successor became one of the most influential sites on the internet. Today, its English-language edition alone hosts more than seven million articles.
Yet in a world where fact and fiction are increasingly difficult to tell apart, and as artificial intelligence (AI) reshapes how we create and consume knowledge, where does Wikipedia fit in?
To find out, BBC Science Focus sat down with Wales. He discussed how the platform has evolved over the past two decades, the challenge of maintaining trust and neutrality in an age of misinformation and how AI could transform its future.
SF: You’ve called social media the place where trust goes to die. Can Wikipedia still compete with that world?
JW: I think we can definitely compete with that. I’m not sure that I think of it so much as competition – not like business competition or traffic competition or anything like that.
Clearly, there’s a broad need for a shift away from culture wars towards a more thoughtful, reflective way of dealing with each other, accepting people with differences and having more dialogue across dividing lines.
But I think we’ll always have both. Division is quite normal in human society. I actually liken it to whenever I hear people talking about young people today – saying they have no attention span and they’re just on their phones. And I think, ‘Yes, clearly there’s this addictive short-form content that young people love. But guess what? They’re also bingeing 14 straight hours of very complicated TV shows.’
Both are part of the human experience and can coexist: the fluffy social media that’s not particularly healthy for you and the serious research and thinking that happens online. It’s the same with our diets – hopefully, you don’t only eat junk food but have some nutrition as well.
SF: How has Wikipedia managed to maintain a relatively constructive culture while the rest of the internet is becoming a bit of a dumpster fire?
JW: I like that you say ‘relatively constructive,’ which does acknowledge that obviously we have our quarrels internally and people do get upset sometimes. That’s just part of the human condition.
I actually think it’s part of the design – partly software design, partly what I call community design. It’s really a combination of things.
If your business model is based solely on ad revenue, clicks and engagement, then it’s very easy to accidentally go down a dark path where you’re promoting content that doesn’t make people happy, but just keeps them clicking a little longer, getting a bit angrier and arguing with each other. And that’s very unhealthy in the long run.
Whereas Wikipedia, we don’t have any ads, and we’re supported by donations from the general public. So our business model, if you want to call it that – we’re a charity, but we do have one – is that we don’t need you to be addicted to the site and clicking as much as possible.
That drives us in a different direction when we think about the design of the site and the rules of the community.
Civilised discourse is super, super important, because if we don’t have that, Wikipedia ends up being full of nonsense. So, I think it’s really a lot of different fundamental pieces of the puzzle.
SF: Wikipedia has faced more criticism recently – Elon Musk has even dubbed it ‘Wokepedia’. What’s your reaction to that?
JW: It isn’t great. When Elon calls us ‘Wokepedia,’ he’s just wrong – just factually wrong. That doesn’t make sense whatsoever.
There’s this ‘gotcha’ question on trans issues: What is a woman? If you go to Wikipedia, the entry on the word ‘woman’ says “adult female human”. It’s completely uncontroversial, right?
And for people who can’t quite accept that very simple definition, it also goes on further down the article to talk about some of the more complicated questions of gender in society, and so on. All very valid.
But the idea that we’ve become some kind of crazy left-wing activist organisation is just incorrect – factually incorrect. That doesn’t mean there aren’t areas where we should improve. There are definitely cases where I look at an article and think that it isn’t really fair to both sides, and in those cases we need to think about that and work on it.
But the right answer is to get more people involved. I’d like more kind and thoughtful people who see a bias in a Wikipedia entry not to assume it’s caused by some crazy woke activists who’ll block you the moment you disagree, but to understand that people are just writing from sources.
They may not have seen the other side of the debate, and we can bring that in and chew on it together.
Read more:
- This is Elon Musk's daring $1bn plan to destroy the ISS
- How Elon Musk's dying satellites could hurt the ozone layer
- Elon Musk is building Starbase, a city with a spaceport to the Moon, Mars and beyond. Here's what's inside
SF: As people increasingly turn to AI for information, could Wikipedia become redundant?
JW: I wouldn’t want to say never, because who knows what the technology holds in the future. But at the moment, we haven’t seen a huge impact in that sense.
There was a study conducted by think tank Pew Research that found that, in traditional Google search results, Wikipedia appears in the first ten links about 3 per cent of the time. In Google AI summaries, Wikipedia is linked about 6 per cent of the time.
So, Wikipedia is being cited a lot more by Google AI summaries. Of course, people click through those links a lot less. We’re not super obsessed with clicks, but that’s something to observe.
I use AI a lot myself, and I find that the use cases where it’s actually quite good now are ones you wouldn’t come to Wikipedia for in the first place.
I’d have thought 40 years ago that the first AI capable of writing in English would be very boring, not creative, and very fact-based. You’d imagine it just mechanically regurgitating things it had read. But it turns out it’s not very good at that – it’s good at creative, brainstorming nonsense. It’s surprisingly good at that.
SF: Do you worry that biases on Wikipedia will be translated to AI models trained on its vast bank of data?
JW: I do. I think that’s something that really needs to be focused on, particularly by the people who are training AI models.
The two largest sources of data for most large language models (LLMs), from the stats I’ve seen, are Wikipedia and Reddit. And both are great – but they’re also flawed. They are what they are, but they’re both very heavily male-dominated areas.
Hopefully, with Wikipedia, because we have this commitment to neutrality, we’re already trying to be conscious of that sort of thing. But if you’re not careful, you could just blindly go and reinforce existing biases.
At the same time, though, because LLMs can be trained and prompted, it’s not hard to imagine an AI running constantly over Wikipedia – looking for instances of bias and calling to our attention to an entry that we need to look at.
As long as it’s used in the right way – for instance, to show that articles about Nobel Prize-winning economists tend to mention the families of the women who’ve won, but not the families of the men. That’s interesting, and it might be something nobody noticed.
This conversation has been edited for length and clarity.
Jimmy Wales’s new book, The Seven Rules of Trust, is available to purchase now.
Read more: