At PolymathMind.ai Adam Riley and I have been examining the threat that comes with the AI medium luring us into accepting outputs that should have been more thoroughly checked, double-checked and rigorously evaluated.
But the bigger threat is the almost sinister way in which this deference to AI authority sneaks up on us, slides under the door and by stealth becomes the norm.
The worry is how easily we slip into accepting AI’s outputs without the clear, deep, forensic evaluation to which we would normally have subjected such information.
So the big risk is that people stop noticing that they have stopped checking. It is akin to the death of human thinking by a thousand cuts, because we have not noticed what is happening: the steady erosion of our questioning muscle.
One minute we are rigorously and intellectually checking everything. But, over time, we lazily drift into accepting AI’s convenient first draft in an unquestioning way. The opening gambit from this new medium that is AI sneaks up on us until it becomes “the way we do things around here now”.
So, the speed with which humans may be handing over authority to AI is surely cause for concern.
The practical question is not simply, “Should we trust AI?”
It is, “What would AI dependence look like if we stopped noticing, if we stopped checking, if we stopped asking, ‘Are you sure this is the best you can do?’
If you cannot answer that for your team, then you could be in trouble!
