Stupidly Efficient
Modern systems are stupidly efficient.
This thought popped into my mind when I was thinking about a problem I had with a food delivery order. My delivery driver had accidentally given me someone else's order and I was unable to find a way to get in contact with them or a customer support person to ask them to come back and swap them.
The food delivery app had hidden its customer support behind enough layers that by time I found a way to reach a person so much time had passed that I gave up. Instead the app's customer support bot directed me to the "I have an issue with my order" refund form which I filled. After filling it, I went to try and order my food again but the restaurant was closed.
So I ended up eating instant noodles and getting refunded and dumping some perfectly good food into the trash (allergies).
They were probably able to cut down on hiring support staff. How efficient.
Another example is Google summarising the top web search results for me. I know it has a decent chance to be wrong (especially when searching for niche things, you end up with a word for word ripoff of some random person's Facebook comment). But if I click through all 10 links to find an actual concrete-ish answer it'll take time. It's much faster to just read the summary and hope.
Efficient.
Stupid.
The best part about this one is it's a bi-directional stupidity. Because Google is effectively preventing me from needing to go through the links. As a result the quality of their search results aren't getting better because nobody ever clicks through. So the summary gets worse (or at least doesn't improve), because it's based on the top search results.
It's efficiently making things stupider from both sides. You really gotta hand it to Google.
The primary example of stupid efficiency and the one that haunts my posts like a spectre is AI. Using and running AI is going to speed up climate change. It will do this purely as a result of the amount of electricity it uses. The argument is that as AI gets better, it'll become able to help us solve problems in ways we can't even imagine now. So we can cause wanton harm to the planet and the environment because the AI that might come someday will be able to think of cleverer ways than us to fix them (unlikely, but that's an argument for another time). So it's more efficient for us to burn the resources trying to get to that better AI than to slow down and plan ahead. Stupid.
The clever bit is, this one is once again, multi-faceted. The AI models we have now, LLMs more properly, are purely based on human knowledge. They are trained on the output of humans, chosen by humans, modified and put through reinforcement learning by humans. However, the use of AI removes the mental labor humans have to do.
Have you used AI for work?
Do you feel like you got slower at doing it yourself? (Have you even tried?)
I consider myself fairly competent but even after using it briefly I felt the friction.
How about Google Maps or Waze?
Do you think you store more or less information about travelling in your head?
Or modern IDEs?
I've definitely noticed that when the code completion is on the fritz my brain is slower at remembering what comes next.
These are all symptoms of the same problem.
Just to be clear I don't think Google Maps is bad. I don't think most of us need to have all the random routes we may ever need to take stored in our heads. Though I feel pretty disgusted when a game does that routing thing for me, so maybe I should think about that more.
The knowledge we gain doesn't go away at some point. I don't know what that point is, but I do know the process of getting there involves struggle.
If an AI just spits out the result.
There is no struggle.
If it spits out the wrong result but your strategy is to just ask it to do it again.
There is no struggle.
When you arrive at a problem the AI can't help you with, suddenly you're faced with an unreasonable mountain of struggle. AI let you defer it all, pushing the small piles of dirt ahead, further down the path. All the struggle you avoided before is now compounded before you, a mountain cutting you off from the small pile of dirt of the problem you actually need to deal with. This mountain does not exist for those who had walked the path themselves, sweeping these piles of dirt away one by one.
The result is less experts and less expertise. Juniors never get the chance to get better because the training wheels were forcibly bolted to their legs on day one.
And eventually, there is no new information to improve the training data. There are no people to tweak and improve the systems.The whole thing becomes a self perpetuating feedback loop of nonsense, until it presumably finally collapses.
Stupid.
Efficient.
Great.