Startups: “AI doesn’t do what Wikipedia does”, says Jimmy Wales

When Jimmy Wales founded Wikipedia 25 years ago, the idea of ​​having a collaborative encyclopedia, with entries written by anyone, seemed absurd. The central concept behind the business was precisely trust, the theme of the new book launched by the entrepreneur. In an interview with Startupshe talks about the technological changes experienced since the emergence of Wikipediathe influence of artificial intelligence on human relationships, and how to restore trust in people and institutions.

The Seven Rules of Trustlaunched in Brazil by Objective Publishingwas written in partnership with Canadian journalist Dan Gardner, co-author of bestsellers such as Superforecasting e How Big Things Get Done.

Published in the midst of a time of profound crisis of credibility in institutions and human relations, the book analyzes the mistakes and successes of Wikipedia and shows how an open project, edited by anonymous volunteers, managed to become one of the most trusted sources of information in the world.

For him, a crisis of trust has an impact on the way society organizes itself: “When there is a decline in trust, demagogue politicians appear, disconnected from reality. This is very serious. If people don’t trust journalism, they no longer know what to believe and become vulnerable to this type of thing”, he says.

Asked about the role of Wikipedia With the popularization of AI, Jimmy is blunt: “AI can’t do what Wikipedia does.” And he makes a comparison with chess: “Computers have been playing better than humans for decades. And yet chess has never been more popular. People still play because it’s fun. The same goes for Wikipedia.”

Read the full interview below.

Startups: You created Wikipedia 25 years ago. At some point would you have imagined that technology and the internet would reach where they are today? How do you see the popularization of AI?

Jimmy Wales: A lot is happening very fast with AI. Technology is exciting, but it’s also problematic because it’s not very good at fantasizing. Still, it’s a very exciting time for innovation. In the Wikipedia community, we are exploring experiments. AI isn’t yet good enough to write articles, but it could be useful in other things — like suggesting edits without wasting people’s time.

Why do you say AI is not good at writing Wikipedia articles?

The problem of hallucinations is serious. For famous topics it does reasonably well, with some errors. For more obscure subjects, it hallucinates a lot, producing nonsense that seems plausible. It’s simply not good enough.

Have you ever seen humans “hallucinate” like AI does when writing entries for Wikipedia?

It’s different. People make mistakes, they confuse things, that happens. But the AI ​​hallucinates in different ways: it invents something very plausible even without knowing what it is talking about, and when confronted it apologizes. If you met someone like that, you wouldn’t trust them.

You wrote an entire book about trust. Why does it matter so much?

When trust declines, demagogue politicians emerge who are disconnected from reality. This is very serious. If people don’t trust journalism, they become vulnerable. Within companies, a high trust environment is essential. I saw a news story about Facebook tracking employee mouse movements — this is simply toxic and very low trust.

Does Wikipedia still make sense today, in a context of so much polarization and misinformation, and precisely a lack of trust?

It makes more sense than ever. There is a huge demand for clear, quality and neutral information. When you face a health issue, for example, you don’t want ideology, you want facts. The public is increasingly dissatisfied with the current information ecosystem.

Are people trusting each other less than they used to?

There has been some decline, but not as big as in politics and journalism. One concern is that people are forming their views from social media, which promotes those who are not ordinary people, who often gain from discord. In real life, almost everyone you meet is perfectly pleasant.

You are an optimist.

I’m pathologically optimistic, as I often say.

How do you navigate the tension between engagement and accuracy on social media? Do you use it? Do your children use it?

Use with care. I didn’t delete my X account because many journalists follow me, but I use it much less, it has become completely useless. I limit YouTube Shorts to 15 minutes per day. With children, we use parental controls. And I always tell parents: before you worry about your children’s screen time, look at yourself. You lead by example. There’s no point in wanting your children to reduce their screen time when you’re using your cell phone at the dinner table.

Where do you usually get information?

Mainly on Google News. Sometimes I delete my cookies and browse offline. I like the randomness. I remember printed newspapers: as you leafed through, you saw everything, without any algorithm choosing for you.

The biggest names in technology are aligned with Donald Trump. How do you see the relationship between technology and politics nowadays?

I believe many companies are simply staying quiet to avoid threats to their business. Some actually support it, like Peter Thiel. But I think it’s important for companies to stick to their principles — even if sometimes you have to choose your battles wisely.

Have you ever found yourself in a situation where you needed to take a political stance?

Yes, when the matter impacts Wikipedia. We are strong defenders of press freedom and access to knowledge. When we were blocked in Türkiye, we went to the Supreme Court and won — I’m very proud of that. In China we remain blocked, but I will never violate our principles to access a market.

What are the good and bad sides of AI for humanity?

On the bright side, AI is revolutionizing biology. Autonomous cars will drastically reduce traffic deaths. On the bad side: the proliferation of low-quality content — now anyone can create tons of fake news at no cost. And vibe coding is exciting, but AI-generated code tends to have a lot of security holes.

How do you see the use of AI by younger people, for example, to do school work? Is it something that worries you?

Overall, I think it’s a good thing, as long as we teach how to use AI correctly. It doesn’t matter what we think: young people will use it anyway. We need to teach limits and correct uses in schools. Using AI to empower thinking is extremely valuable. Using it to avoid thinking is extremely bad.

Does Wikipedia have a future in the age of AI?

AI cannot do what Wikipedia does: the judgment, the discussion, the process of refining knowledge. And look at chess: computers have been playing better than humans for decades, but chess has never been more popular. The same goes for Wikipedia. Even if AI writes well, humans will continue doing it because we are nerds who do it for fun.

Content produced by