AI AI... Oh!
Updated: 6 days ago
“In a time of deceit telling the truth is a revolutionary act.” - George Orwell
The other day I wrote a piece to post as my first submission to Substack. I'm enjoying the writing process and figured it would be fun to share some of my ideas and thoughts with a larger audience, and hopefully funnel a few new interested eyes towards this website. Armed with the article, I set up the page, added an image and got ready to post. Then I noticed it had an AI checker - something called Pangram - and thought it would be fun to run my text through their system and see where I ranked. Apparently, not very well. The piece was apparently only 33% human. I laughed it off at first, until I considered what this actually means. Either I write in a manner that uses AI, I write in a manner that writes like AI, or possibly I'm not a human at all and have been living a life of internal deception since my conception.
This also raises another, somewhat existential, issue. Are we now entering a timeline where machines are not only checking for AI, but potentially dismissing genuine human input because of some freakish flaw in their algorithmic design? On further investigation, it would seem that Pangram isn't a be-all and end-all AI checker. It isn't necessarily there to definitively call out the use of AI, but rather to provide an assessment that users can then make their own judgement about. And therein lies the problem. We're increasingly allowing machines to make judgements about humans while simultaneously assuming that the machines themselves are objective. AI has already demonstrated its ability to get things spectacularly wrong. We've seen problems with facial recognition and AI-driven CCTV systems, so why would text be any different? If a system can misidentify a face, it can presumably misidentify a piece of writing as well.
The potential consequences become rather more serious when you consider where these systems could be used. Education, university admissions, job applications and countless other areas of everyday life could potentially be affected by a system making an assessment that is wrong, yet carries the appearance of authority. I also take issue with the idea of being judged by a machine in the first place. What evidence sits behind the assessment? How reliable is the system? What margin of error does it have? And, perhaps most importantly, what happens when it gets it wrong? If a person makes a judgement about something I've written, I can challenge them. I can explain my reasoning, provide evidence or simply disagree with them. But when the judgement comes from an algorithm, there is a danger that we simply accept it because, well, it's a computer. The computer says so.
Whilst I call out AI for many of its shortcomings, I accept that its ability to process vast amounts of information, perform calculations and assist with complex problems can be put to great use in science, medicine and many other fields. I don't have a problem with the technology itself. I question how it is increasingly being used for front-end decision-making, and how readily we seem to accept its conclusions. We appear to be walking towards a rather dystopian environment where 'the computer says no' becomes an acceptable answer to increasingly important questions. Even more concerning is the extent to which governments appear to have embraced AI as the future, while the safeguards and regulation required to manage its shortcomings seem to be playing catch-up. To be fair, there is growing recognition of these issues, but it still feels remarkably late in the process.
Whatever the future holds for AI, I'm now left questioning my own writing. I don't profess to be Tolkien or Rankin, but I have never previously questioned whether my writing was sufficiently human to be published. Now I have a niggling doubt that perhaps I shouldn't make my writing public if an AI system is going to call me out for something that is, quite simply, wrong, and I guess that's why I take issue. The machine doesn't have to be right to make you question yourself. It only has to sound authoritative.
Somewhere, Orwell is rubbing his hands together and saying, 'I told you so.'
Comments