What happens when Grokipedia writes about you (I’m lucky, but there are cautionary lessons)

This morning, quite inadvertently, I stumbled upon the article about me on Grokipedia, the xAI / Elon Musk alternative to Wikipedia. More about Grokipedia on Wikipedia here.

And – much to my amazement – it is reasonably accurate, and much more detailed than Wikipedia in English* (German is even more outdated, and there are translations into Macedonian (how?) and Arabic).

The things Grokipedia gets wrong – like the subject my father taught (it’s geography and not history and politics) – are likely AI hallucinations, extrapolating from all the blog posts I have written and drawing a categoric conclusion where nothing as clear ought to be said. Most interesting is how it has tried to make an assessment of what is important, and does that quite well – it goes into a lot of depth about the Brexit diagrams for example, and I did spend a hell of a lot of time on those diagrams, and got a lot of reaction. And that whole body of work does not even feature in Wikipedia.

But the thing is, ultimately, has what I have spent my time doing actually really justified real humans using real time to write up something on Wikipedia?

It’s not really that anything since the Atheist Bus Campaign is of much note, and it is only thanks to the early 2010s there is anything about me on Wikipedia anyway. Although amusingly it is thanks to existence of the page on Wikipedia that there is even the page on Grokipedia. And what Grokipedia has put up is a public version likely roughly equivalent in quality to what Copilot or ChatGPT would spew out, only on the public web. It is also worth noting here that everything on my blog is free to access – it is not as if Grok is scraping and pinching copyrighted texts.

The challenge of course is this: is what Grokipedia publishes actually true?

In my case it has a hell of a lot of information to use, because this blog has been running a very long time, and – for example – it can work out I have moved from Berlin to Bourgogne thanks to this. And I can confirm a lot of it is true, because I know what I have and have not done. But then you have to believe me if you are reading this blog post, just as Grok “believed” the hundreds of other posts it obviously scraped to pull together the entry.

But that is when it gets a whole lot more messy.

Because were there less public information about me, and were there hordes of people trying to undermine me and flood the net with incorrect information, and were I someone more likely to be attacked by Musk allies (I am a straight white middle aged man who writes about railways mostly, so I am not really in the firing line – and the Brexit stuff was too long ago) it is easy to see how this can go very wrong. And we know very well it does go wrong by design – I am just not a person in the firing line. And the danger – were you to try to fight back against inaccuracies – of creating a Streisand effect is very great.

So what conclusions can I draw from all of this?

First, I am not going to even try to correct the errors in Grokipedia – it’s not worth my while. It lends legitimacy to something that doesn’t deserve it. Second, were someone to want some more detailed biography of me beyond the much smaller text than Wikipedia can totally understandably produce, AI might be the basis for that – it most definitely is an easier to read than ploughing through my blog and cross referencing. Third, I am breathing a small sigh of relief this is not worse than it is, but I can of course see I got lucky – it is very easy for it to turn out a hell of a lot worse. Solidarity with those in that situation!

* – amusingly English Wikipedia has had the label “This article contains promotional content. (October 2021)” and I have no idea how, because I have never edited my own Wikipedia page (because that’s against the rules and I am not that vain anyway) and don’t know who it was who pasted in some biography from somewhere that sounds somehow like promotion.