I wrote about it here too:

https://blog.nilesh.io/post/llms-and-jobs

Just because people have a vague idea of what someone's job is and a deep understanding of their own job, it feels like AI can replace everyone but you.

It's a tool not a replacement.

Job losses due to productivity may be inevitable, but that's not the same as being replaced by AI, that's being displaced.

Comments URL: https://news.ycombinator.com/item?id=48366813

Points: 3

# Comments: 4

As long as AI cannot dynamically generate a true world model on the fly, it will continue to fall short.

Humans can think abstractly, parsing a problem space and condensing it into a solution. That is, I can take 10 different seemingly unrelated problems, and determine they're all the same class of problem, and then create a novel (ish) solution for that problem. LLMs simply cannot build that world model AND even when explicitly told the solution, they can't actually hold the model while authoring code.

They're fake thinking, and no amount of harnesses, or increasing the number of models is going to change that.

There was a post here before: tech companies have become adept at taking a thick thing (like friendship) and boxing it up to give you the same surface level feelings but it's not _real_ it's a thin version (following an influencer, social media). And you feel like you're friends with these people, but real friendships are not so simple, they're not so messy.

LLMs are robbing us of thick thinking and convincing us that thin thinking is something real. I don't think anyone can legitimately claim that LLMs are improving their product's actual problem modeling space, beyond just being a randomness / entropy generator where a human can actually choose.

Human written books, code, art are fundamentally shaped by having a true understanding of the abstract problem space and compressing it down it a way that makes sense to other humans. LLMs are just word salad masquerading as intelligence. Letting them into our world is a mistake that will take years to correct.

People said the exact same thing about web searches, and I think there's a lot of devs who would instant search for every issue they hit.

Isn't this just better web search?

On the other hand, it definitely feels like it might be too big a step in the spoon feeding direction.

Writing code without AI feels like art, and writing it with AI feels like painting a wall: get it done quickly, cheaply, and good enough that people don't see issues.

It's the art part of engineering that's being lost, AI has no appreciation of elegance. It has no empathy for cognitive overhead of bad code or poor-fit design patterns.

Cognitive Debt is the phrase to Google btw.

What does your persistent storage layer look like on Talos? How have you found it's hardware stability over the long term?

The hype and zealotry remind me of a cult. And as I go higher up the chain at my big tech company, the more culty they are in their beliefs. And the less they believe AI can do their specific jobs, and the less they have actually tried to use AI beyond badly summarizing documents they barely read before.

AI, as far as I can tell, has been a net negative for humans. It's made labor cheaper, answers less reliable, reduced the value we placed on creativity and professionals in general, allows mass disinformation, and mostly results in people being lazier and not learning the basics of anything. There are of course spots of brightness, but the hype bubble needs to burst so we can move on.

LLMs will spit out responses with zero backing with 100% conviction. People see citations and assume it's correct. We're conditioned for it thanks to....everything ever in history. Rarely do I need to check a wikipedia entry's source.

So why do people not understand that: this is absolutely going to pour jet fuel on misinformation in the world. And we as a society are allowed to hold a bar higher for what we'll accept get shoved down our throats by corporate overlords that want their VC payout.

I'm beginning to become disillusioned with these things. We're replacing like 1000s of jobs with a system that will almost certainly do a worse job than before. And the money is split between hospital shareholders and VC.

I get that there's an efficiency (market) gain here. But these AI startups that target existing sector automations seem like they're most just attempting to drive wealth inequality in a period of already terrible westh inequality.

I've been burned by a small SBC that had poor support, but on laptops/desktops never felt limited.

But people always sound so excited to libreboot their personal computer... Am I missing out or is it just nerd cred?