Everyone is an idiot in some capacity. I’m an idiot, you’re an idiot, so are all the people you see every day. Everybody has at least one facet in life where they are colossally stupid.
Most of us are fortunate enough to recognise where they are most likely to be an idiot and tend to mitigate for it ahead of time. Give me an Ikea flat pack and I’ll immediately be on the phone to friends to save me because I know things will go badly if I try anything.
Yet what about when our own idiocy is staring us right in the face and we don’t see it? John Carpenter’s rollicking action flick They Live posed just this question in 1988. The ruling class in the movie were in fact camouflaged aliens that had been subjugating the masses.
Nada, the lead, finds a special pair of sunglasses that allows him to see the truth in front of him. What follows is a one-liner-filled romp but the crucial point is an over-the-top fight sequence with Nada’s friend Armitage, where the latter resists Nada’s efforts to make him wear the sunglasses to see the truth for himself.
READ MORE
Perhaps we all need to check our prescription because the ease with which people have fallen for obvious artificial intelligence (AI) training models lately is alarming.
There have been two clear examples of this in recent weeks on Instagram. The first was a new-year trend to share a photo of yourself from 2016 and 2026. The benefit to AI is pretty clear in that it led to hundreds of thousands, if not millions, of data points of people demonstrating ageing processes.
That one, in isolation, was forgivable. It was the new year, people have little to do in January, and a bit of whimsy seems harmless. It was the second one that truly broke my brain.
The trend was for users to ask ChatGPT to make a caricature of them using all the information it knew about them. Many of you reading this took part in the challenge because it was inescapable. Indeed several elected officials did as well, although some later took them down.
All of this despite the repeated warnings since ChatGPT and its ilk arrived about the importance of protecting your own personal data.
[ Can machines think or are they just fooling us?Opens in new window ]
To understand why these trends aren’t a bit of harmless fun it’s important to look not at what they ask for but what they give in terms of information.
On the surface, it’s all light and a bit of craic. There’s nostalgia in looking back at how you used to look compared with how you look now, and something light in seeing an exaggerated version of your current self.
Upon closer inspection, this is effectively labelling something as saying it looks like you or, if you offer up corrections, telling the system where it went wrong. The system is testing both your distortion tolerance and your vanity.
You have demonstrated behaviour and reaction to the machine and it has learned from that, all while gathering heaps of more bland but useful data. You gave it up for free just for that light moment of amusement.
Now multiply that by an enormous number, for all the people that participated in said fad, and you’ve got a data set of real value that can be used by AI companies. Again, all of this with no compensation to the user.
[ It costs a fortune to feed the AI monster, but is it worth it?Opens in new window ]
Warnings about such matters have been shouted loudly for a long time yet people who are both smart and informed fall for them. It’s a failure to mitigate individual idiocy on a global scale.
We love to be flattered, we love things that are easy, and we love to try things that our friends have done. This cocktail of ease and low-effort social interaction can catch us all out.
Think of the alternative. I’ve just written the guts of 700 words saying you are an idiot and that you shouldn’t do an easy, fun thing. That is obviously far less appealing, and the people putting together AI training models know it.
Here’s the thing: AI systems are extraordinarily stupid, at least compared to the human brain. They need far more assistance than any of us to learn. That’s why they require such enormous data sets to be able to adapt and, essentially, manipulate the user into action.
Viral fads are a relatively low-effort way for AI training models to hoover up swathes of information to get smarter. AI systems are decades away from matching human intelligence but, right now, they’re winning by getting just smart enough to bring us down to their level of stupidity and then beating us with experience.
Their models extract as much information as possible without really explaining why. Does that sound equitable to you?
Keeping the glasses on didn’t exactly save Armitage or Nada in They Live. The duo managed to cause some damage to the establishment but paid the ultimate price in the process.
You alone eschewing fads that are AI training models won’t stop them from learning. The more of us who do, however, the dirtier the data sets become and the less useful they are to AI systems. It begins by accepting the truth in front of you.
