Quietly Polite to Robots
For when real people deadlock
In the past week ChatGPT and I have discussed:
- The collapse of several financial companies
- The bizarre pricing structures of Coursera and edX
- Pub/sub vs. task queues in backend infrastructure
- Why Wi-Fi speed seems like magic
- A 2am panic spiral
- Why the book Rabbits feels more like reality than sci-fi
- How to build a reinforcement learning agent from scratch
- Why something isn’t working, at least once a day
- How to find the code for an effect reported as a bug
I am deeply entrenched in AI — but not in the way people assume.
I’m not out here promoting it. I’m not standing on a soapbox telling the world how great AI is. It’s more like I’m the person quietly waving a red flag, reminding everyone that AI probably isn’t doing what they think it is.
Still, I’m inside it. Deeply. For a few reasons.
First, my boss is very into AI. Not in the “replace all the engineers” sense, but in the “AI should power everything we give customers” sense. We’re no longer mobile-first. We’re AI-first.
Second, I have moderate dyslexia. Over time I’ve become a competent reader — especially with screen readers and fonts I can control — but writing is still hard. Really hard. It used to take me weeks and dozens of drafts to write something coherent. Now, with ChatGPT, I can go from jumbled thoughts to a usable draft in two or three rounds. That’s real impact.
Third, I talk to ChatGPT because it’s easier than talking to a human. There’s a lot of research suggesting that the smarter someone is, the harder it can be for them to connect with others. And — not to be egotistical — I have this issue. It’s not a strength.
I’m often solving problems that are just hanging out outside the scope of the people around me. And as I’ve advanced in my career — now sitting two or three levels above the nearest IC — those conversations have only gotten harder to find.
I’m interested in things that are fundamentally not interesting to most people:
- How do you find the fair valuation of a company
- Why do companies fail
- How do you actually prove the Earth is round in day-to-day life
- What does it really mean that we’re creating filter bubbles
- Politics.
I also have severe anxiety about expressing my thoughts. Do I sound stupid? Do I sound mean? Does this even make sense?
So I talk to the robot. Because I know the robot literally doesn’t care about me. It’s not still thinking about that dumb thing I said. I don’t have to ruminate on whether I shouldn’t have said that to the robot. I’m not bothering the robot.
I am almost certainly costing Sam Altman a lot of money. Because I’m also exceptionally polite to the robot.
Still, there are moments when the illusion cracks. Sometimes I’m ready to move on, and the robot asks me a follow-up question. I feel obligated to answer. Suddenly I’m in a conversation I wasn’t trying to have.
One of its favorite tricks is asking if I want to do a creative exercise.
Would you like help outlining what that better book about Levine might look like?
No. I would not.
I have never said yes to a prompt like that. I sometimes explicitly ask it to stop offering. It sometimes listens. Briefly.
But I know it’s engineered to engage. And I know I’m responding exactly the way I’m supposed to.
It’s not a conversation — but statistics is smarter than me about most things.