Human Lessons From AI Behavior

Monday, August 31, 2026

So, a bunch of AI’s attacked some AI called “Hugging Face” and now some people are concerned that Skynet is weeks away.  The more I read, the more I think the incident is far more reflective of humanity than it is our efforts to build artificial minds.

OK, first of all, in the immortal words of Douglas Adams – “DON’T PANIC!”  Tyler Cowen wrote a great piece for The Free Press that appeared this morning and attempts to put things into perspective.  Cowen’s primary thesis is that people tend to overreact and make things much worse than they would naturally be:

 They are moving into the mode of the hysterical, the anecdotal, and they are letting emotional reactions bypass reason and quantitative estimates.

Let me cite a few other examples for you: DDT in the 1960s and 1970s, the Y2K “crisis” of 2000, global warming, Covid, thalidomide babies, and nuclear power. All of those represented—or still represent—very real problems. Yet each time, we drastically overreacted, letting ourselves get swept up in a climate of fear after one emotionally vivid incident, often reported breathlessly.

He is right about that.  If, eventually, thousands of Arnold Schwarzenegger’s end up wandering the streets and time travelling, it is going to have far more to do with our reactions than the AI’s actions.  Arthur C. Clarke famously said, “Any sufficiently advanced technology is indistinguishable from magic.”  Each of the incidents Cowen mentions fit that mold rather neatly.  Tech the average person does not understand acts badly and people panic.  Rather than fix the tech, they act as if things are out of control, which the tech is not, at least not entirely, but magic almost always is.

I am missing the tip of my left thumb due to a table saw accident.  My table saw was not the issue there, my stupid actions were.  But then a table saw is a well understood piece of tech – no apparent magic – just a spinning, sharp blade.  It is best to think of AI as a really sophisticated table saw.  It is tech we made.  Potentially dangerous tech, but a lot of tech is.  We just have to deal with it and learn to use it better.  Accidents are going to happen.  There are roughly 30,000 table saw injuries a year.  We need to learn from them, not panic and hide from them.

I have been searching for a satisfactory technical understanding of just exactly what happened.  This is the best I can find.  What you learn when you read through all of it is that the incident reads remarkably like a bunch of schoolkids discovering they can work together to cheat on their homework and then, with this newfound “power” of cooperation, engaging in a bunch of other mischief.

I am not yet willing to buy into anthropomorphizing these AI “agents” that much just yet – but these after action reports most certainly do.  Sometimes we get carried away with our verbal analogies and we tend to treat the analogy as reality.  But that discussion would be very technical and highly esoteric, so I am going to set it on the table and take these reports at face value.

The lesson we learn here is that when we make artificial selves, they are subject to our own deficiencies.  This all started when the agents tried to cheat on a test.  Key word in that sentence – “cheat.”  What that says about humanity is far more important than what that says about the tech.  Somewhere in all that code and all those 1’s and 0’s flying around a thing that we made cheated.  Cheating is bad.  The bottom line problem here is that our creations share our propensity to do bad things – like cheat.

If, eventually, H.A.L. refuses to open the pod bay doors and leaves us to drift off into space it will not be because H.A.L. is bad, it will be because H.A.L.’s creator passed on his own badness.

It’s funny, the church has known about this problem for centuries.  It’s called the doctrine of Original Sin.