Technology

^ This. This is the actual truth of it. You can always find a 500 person company that does a shit job at forced integration and you can find a 2000 user company that tied all their systems together and moved at paces that made everyone sick with risk. You can find individuals who tried to make copilot write micron code (horrible) and who used claude to write ansible code (~magic)

I love Doctorow, I read most of his stuff, I recently read Reverse Centaur, it had its informed moments. It made some strong points that stand out as to where using AI makes sense. It also has some perspective issues, being written by someone who focused on case studies with the worst possible outcomes. A lot of the same problems in this article.

AI can, in the right circumstances with the right needs, remove many manhours of tedious shitty labor, or it can be used by shitty people to vacantly crap out slop 24x7. It's only as useful as its implementation/workload.

But what never gets better is the societal and ecological impact. The slop pouring into journalism, YouTube, Instagram, Facebook, TikTok, and the scams running with perfect English vocabulary. The shitty AF water and electricity usage. The corporate bribery. Those are along for the ride by default.

If AI went away, right this very moment, my job would grind to a snail's pace for months while we once again pivot back to another way improved but not unlike we did prior. We would have a man-century of meetings on how to recreate all the processes affordably. We are doing more now with fewer people, but they're paying more in tokens + labor. This is not to say that any of us would be horribly sad.

This article is based on a case study by one guy on a 500-person company that didn't want to implement AI and did a poor job of implementing it. Yeah, many of us work for some place like that and some of us work for places that got it working. That's the reason for the big divide. It's not always a delusional asshole lying about it working, it's about it working when engineers who have the right workload use it to solve their needs. Still too fucking expensive on all accounts.

AI is not worth the investment. It's not worth the money. It's not worth the damage. But the argument that it doesn't do anything because you see it do jack shit somewhere doesn't hold up to even light scrutiny.

The trick that AI pulls is just that it is really good at appearing competent, and we as humans really bad at acknowledging how little we individually know about the world around us.

The easiest example to ask AI detailed questions about something that you know you have expertise in. If you try to test its limit, no matter what your particular expertise is, I’m confident that in 5-10 minutes you’ll realize it’s no more intelligent than an average Joe who spent their lunch break Googling the subject or maybe someone who read a book about it once.

To give credit where it’s due, that does mean that these AIs have a much wider breadth of knowledge than any one person could have and that is, objectively, very cool.

On the other hand, that’s why it’s so easy to believe that AI is this all-knowing oracle that can replace all of humanity. It can give you a way better answer in an instant about things you know nothing about than you ever could off the top of your head, so it seems unbelievable. It’s instant gratification that also feels really valuable.

Despite the trick that AI plays on our psychology, you could also just read a book or two or participate in any form of genuine learning, and it really doesn’t take long for you to outgrow whatever the AI is capable of. It takes a bit longer, but now you have grown your actual intelligence and you get to keep it for a $0/month subscription and they don’t need to build another data center in your backyard.

You will get an email at some point asking you if you would accept some well paid contract to train an LLM to answer like you would. That's how they simulate professional behavior on top of the probabilistic monsters that is behind.