[lazystone]

I think we all been fooled by the sentence: "It's yet another automation, it's like horses were replaced by cars". It is not. Industrialization and automation is about manual labor. LLM/AI is about outsourcing thinking. And while I'll give two thumbs up for using ML(there is not 'I' in 'AI') as a technology for some tasks, outsourcing thinking is an evolutionary dead-end.

[joe_mamba]

>LLM/AI is about outsourcing thinking.

No it isn't. I still do the thinking on how to solve my problems, I only outsource the tedious part, which is typing the code and fixing the syntax errors till it all compiles and does what I want.

If you also outsource thinking to it, that's your choice though. Or the company's choice. But ultimately the free market will deiced with products made using LLMs outcompete those made without.

[cryo32]

It shouldn't be used for that either. The problem is our programming languages and tools are shit so we made another expensive tool to drive them.

I've said this elsewhere before but I single-handedly produced more actual tangible business value with Microsoft Access than anything else since. What was an hour's work is now a procurement process and thousands of lines of tedious configuration and boilerplate that involves pipelines and tens of services all coordinated and hosted by someone who has created a moat to extract money out of me.

All I want is a fucking report.

The LLM makes us blind to the gigantic fucking shit show we built.

[jonatron]

Recently, I had some data for which I wanted some graphs. I uploaded the .jsonl file, and prompted "make a html page and graph this data using plotly". I wanted a report, and got a report, quicker than I could have made it myself.

[joe_mamba]

Not sure why you got downvoted, but I agree with your take. It's now so much easier to one-shot such small task that are supposed to be one-and-done, instead of needing a small army of interns or juniors with would typically handle such non-business critical tasks.

[ethin]

> I still do the thinking on how to solve my problems, I only outsource the tedious part, which is typing the code and fixing the syntax errors till it all compiles and does what I want.

Do you really though?

Here's a question: how many times do you visit claude.ai or open Claude Code (or whatever harness you use) (or use whatever model you prefer) to help you solve a problem, ask it a question, etc.?

One thing I've noticed, which seems to go completely unmentioned, is how these LLM tools are like drugs. It can pump out thousands of lines of code ---> it can write my entire program for me ---> I've written quite a few programs with it and I don't write a single line anymore ---> I go to it for what would normally be things I could do on my own. The problem is that this isn't some immediate thing: like a drug, it sneaks up on you and you don't even realize it until it's far, far too late. I've been programming since I was 13 and I've just now started to notice the deskilling that's been happening to me. I've just now started noticing how often I'll visit Claude Web for something I should be able to do on my own. Nobody really seems to mention this, or it's repharsed as a good thing. And I don't get it: how is undergoing cognitive surrender a "good" thing by any metric other than the metric a beancounter would use? What worries me is that I fear this is happening to way, way more people than those who actually bring it up, potentially yourself included, and you just haven't yet realized it because you haven't really thought about it. That is ultimately what this "AI revolution" is going to bring. It's what the billionaires want, and what they want is usually what they get because the systems we've built are set up to not constrain them.

[joe_mamba]

> I've been programming since I was 13

Congrats, you're in a small minority of people.

>What worries me is that I fear this is happening to way, way more people than those who actually bring it up, potentially yourself included, and you just haven't yet realized it because you haven't really thought about it.

People probably said the same thing about using pocket calculators when they started getting cheap and mainstream in the 1970s.

Do you still use your brain to calculate sales tax or the square root of pi just to keep your brain sharp on knowing calculus, or do you always use a calculator because it's a mainstream commodity that will never go away?

[ethin]

> People probably said the same thing about using pocket calculators when they started getting cheap and mainstream in the 1970s.

Except a pocket calculator is very different from an LLM. It isn't a tool you can go to and ask it literally anything under the sun and get an answer (even if incorrect). It's cold, logical and uninterested in your life or what happens. These LLMs are trained to be interested and empathetic. They are, IMO, way more like drugs than most other things.

[joe_mamba]

>It's cold, logical and uninterested in your life or what happens.

The entire universe is cold, logical and uninterested in your life or what happens to you. But that's exactly what I want from a machine that's supposed to be my personal slave. I don't want a lecture from it on current day political correctness and sensibilities, I just want it to do exactly what I ask it no matter how rude or machiavellian it may be. Whoever will deliver this type of unhinged obedient slave to the masses(most likely China), will win the AI race.

> These LLMs are trained to be interested and empathetic.

Luckily there's already like a dozen LLMs at this point, many free and open source, especially from China, and I and many others will just pick the one that isn't trained to be interested and empathetic, but the one that's my personal slave and does exactly what I want from it, because of the competition between them will allow this freedom of choice.

In 5-15 years when semiconductor shortages ease up and hardware advances a bit more, there's a chance offline open LLMs in your pocket will be as obliquus as pocket calculators were in the 80s.

>They are, IMO, way more like drugs than most other things.

So was music, TV, video games, porn, etc. when they became mainstream. We adapted and survived.

I'm more scared about the state of world peace, economy, CoL, jobs and housing market than of LLMs being addicted.

There's millions of people on the planet who are a couple of meals away from hunger and rioting and migrating to other countries, and a climate catastrophe or one more wars could destabilize the whole planet for good, than LLMs being addictive. If LLMs are your biggest concern, you're in a very privileged position.

[rrvsh]

> > I've been programming since I was 13

> Congrats, you're in a small minority of people.

Why did you respond to this? I fail to see how it relates to your point whatsoever.

[joe_mamba]

Why did he mention it? How does it relate to the topic? Why do you think people are not allowed to reply to it?

[ethin]

> Why did he mention it?

Because it's this thing us humans call context. Or background. There are a number of terms for it. I was providing that so the rest of my post would make a bit more sense.

[dbalatero]

Thank you for writing what I want to scream every time a comparison is made to some archaic technology change.

[satisfice]

It’s worse— it’s seeking to replace every single aspect of what it means to be YOU in the world. Some people are literally trying to “fire themselves” and be replaced with digital twins. Perhaps those people are independently wealthy and also have no need of human connection? For the rest of us, it is a sickening prospect.

AI is automated irresponsibility, and it is nothing like any earlier transition.

When a technology trend means people literally won’t be able to tell if you are living or dead, and also stop caring about the difference— that’s unprecedented in the history of humans.

[jvanderbot]

Its advertised as outsourcing thinking, but I doubt many serious people making serious things actually outsourced their thinking very much. I definitely outsource my typing, search, and LSP interaction!

[wolvesechoes]

Moreover, historical events and processes are unique, even if there are some similarities. Nothing that happened in the past can give us certainty on what will happen now.