That was my first inclination for a title (and I'm sure ChatGPT would have gone with it). It would have given me an opportunity to lever Star Trek into the discussion, and it's a familiar phrase that highlights the issue we're dealing with. According to my trusty Brewer's, the phrase was originally 'the real MacKay' in Britain. A lot of people apparently thought the US version was based on a US boxer called McCoy, but apparently it arose (not entirely surprisingly) in Scotland and was exported to the US as a way of highlighting true Scotch, as opposed to US whiskey.
I switched the title to 'the real McLegg' because it's somehow the sort of thing I couldn't imagine a generative AI coming up with - and there is a sort of reason. I have been addressed in the past as Mr McLegg. This is because one of my early email addresses (with Yahoo, I think - I can't remember why I had it) was brianmclegg@... - brianclegg@ had already gone, so I deployed my rarely used middle initial.
All of this is a long winded way of getting to a new feature from Authory, the content system I use to make my online writing more widely available. They makes a copy of everything written by me online on sources I tell it about. I've found it very useful for this. But the people behind the system have been experimenting with AI and have come up with a cunning scheme to help writers establish their non-AI credentials.
Because Authory has access to thousands of articles by me, it can build a picture of what my writing style is like. As is the case with most writers, I don't always produce the same kind of stuff, so rather than have a single 'digital fingerprint' it produces quite a few - in my case, over 1,000 of them. These are all based on items written pre-December 2022, so won't have any generative AI content. The system then compares new writing against those fingerprints, and can make a reasonable assessment of whether or not the vast majority of my writing is human generated. If this is the case, it generates a certificate.
The concept did bring up a few doubts. My content includes interviews and guest posts, where almost all the text wasn't written by me. Perhaps more worrying, I have intentionally (and obviously) included text written by ChatGPT, such as this article on a 1960s SF story about computer-generated writing that seemed to prefigure ChatGPT. In the article, I asked the generative AI to produce a story like one mentioned in the original. I'm reassured by Authory founder Eric Hauch that a few articles written by others won't damage my fingerprint, and that 'A part of an article that's been framed by you as coming from ChatGPT will not put your certificate at risk. The algorithm behind this is pretty extensive and won't be thrown off by things like this'.
Of course there are limitations on what the certificate indicates. It's only being assessed twice a year (the next time in October). And the chances are that the occasional ChatGPT written article (something I have no intention of using) might slip through the net. But it's arguably reassuring for readers to know that my blogs are not hotbeds of AI generated nonsense. I'm perfectly capable of generating my own.
Comments
Post a Comment