I'm in the basement of a hospital standing over a vacuum pump used for suction in surgery.
It's an emergency service call to fix it.
There's no cell service or wifi down here in the mechanical room. There's no access to Claude or ChatGPT.
And I go blank.
I know I'm capable of fixing this. But I can't come up with basic troubleshooting steps. I've spent too long training my brain to assume Claude will walk me through this machine based on its model number and the manual.
The surgery gets delayed. The patient dies.
Did AI kill that patient?
To be extremely clear: this is hypothetical.
In part 1, I wrote about how AI has a new type of scale. In this part I am pondering a more human aspect of AI.
AI is great at doing, and I delegate that often. When I start delegating the thinking to the agents will I be able to think in that basement with the vacuum pump?
What I fear is that the more I delegate my thinking, the more I atrophy.
It reminds me of Wall-E when the people are too weak to stand properly after being in their hover chairs. They didn't lose the ability in a day. They sat down, and the chair was easy, and they stayed.
What about Google? Or books? Wasn't Gutenberg seeing the same shaking of fists?
This time is different.
Writing and Google externalized memory. I don't need to hold a phone number in my head, because I can look it up, and looking it up is fast and reliable. AI can externalize thinking. At least a type of it. We are able to give over the part that decides what's wrong, what to check first, and what matters. The ability to think is something you have to have on hand, in the moment, in the basement, with no signal.
I catch myself asking Claude how to troubleshoot a portable air compressor before I've even taken off the panels to look inside. I've grown so used to having it that my brain is changing. I can feel it.
When my dad was starting his compressor business, I remember going to the library with him. Business books, sales books, electrical books. He would pore over them. He's an encyclopedia for these things now. How electrons behave is fluent in his brain. He knows things in a way I'm not sure I ever can.
In a few decades, the people who know how to think natively will be retired, and it'll be on those of us from this AI age to hold the world on our shoulders. I question how much of my thinking I'm renting from Altman and Amodei.
If I don't have my agents will I get the hospital's vacuum system back up in time to save the patient?