Jessica Wade: belief

Transcript context

…Will GPT models take significant market share from Wikipedia? Because if you type in the name of an obscure female — or just an obscure scientist — into ChatGPT or its successors, it’ll give you a lot of information. It’s not perfect, but actually, the more obscure the person, the more likely it is to be accurate. I don’t think it will take away from Wikipedia. I think the beautiful thing about Wikipedia is you can interrogate it quite a lot. You can see who wrote what, when they wrote it, and when that reference was added, what that reference is. You have that kind of talk or dialogue behind a Wikipedia page. Most people just land on the read aspect of Wikipedia, but if you go to history or go to talk — the tabs across the top — then you see what was contributed to and when, and you have a discussion around what is and isn’t on the page. I think things like ChatGPT are very useful at giving you a very quick answer, or if you were looking up a person and said, “Make me a short PowerPoint on this,” it might give you that starting-off point from which you go on and find out more things. But I think the ability to interrogate it and to look at — again, coming back to my earlier points really, where it looked for that information. Has it been broad in where it’s collected that from? Is it using actual appropriate citations and literature? You can’t do that. Wikipedia is still important for showing people this person really did the homework on pulling this very nonbiased page together, and that’s why you can trust it. You can keep on interrogating GPT models, but does it at all induce you to rethink what you’re doing? Because obscurity per se is no longer the barrier. Literally everything, everyone is in ChatGPT in some sense.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.