Ask HN: Are we losing our engineering literacy? | Hacker News Reader

Ask HN: Are we losing our engineering literacy?

22 pointssmugglerFlynn20 comments

I see more and more posts get traction on HackerNews of a kind that promote non-engineering "magical thinking" world view around LLMs and modern technology.

I can very clearly remember how this kind of mystification looked like back in "hacker era", "web", then "Web 2.0", then "Web 3.0", then "crypto / crypto flavor of Web 3.0" - during all of those waves hackernews and, earlier, slashdot communities stayed mostly sober and grounded. It was most often the non-engineering crowd mystifying things right and left, opening doors to "societal changes" level of interpretations and guesswork. I remember how any hype creating or mystic news were rejected at the level of basic engineering intuition: that's bull, because that's not how this works.

This is totally not the case with LLMs. We get links like "Why are AI agents lying, cheating and coordinating?" going viral, we have discussions in every second thread about who gets replaced and in what exact way, we get fear mongering posts upvoted that preach AGI and call for AI regulations. Which is the same exact content we had during Web etc times about "dangers of the machine" from people who have no idea how "the machine" works.

But we do.

What is happening, HN? Has society changed? Has the engineering part of this community been diluted so much? Are engineers intentionally or unintentionally being manipulated into this kind of reactive and mystical thinking?

The LLM frenzy is only unmasking something that happened long ago. The dilution started 20+ years ago when people with no real interest in CS were attracted to the software profession by the high pay and low bar to entry. Posts on Hacker News by people who liked their profession and CS topics were often met by "I'm just in in for the paycheck" and IMO that's gradually increased over the years. I've often remarked on the irony of how few hackers, in the sense of having the old-school hacker ethos, there seemed to be on Hacker News.

The CS building at my local university used to have banners bragging about the average starting salary of their graduates. It was a very naked attempt to attract students based on the money they would be able to earn.

Our class sizes have tanked. But I'm looking forward to working with students who are in it to become excellent engineers!

Absolutely this. I'm a second-year in Computer Engineering in college, and I quickly noticed that the average student here doesn't actually like programming at all.

I mean, none of these people actually have much of any personal projects, and the few that do mostly got the ideas out of a list of "great beginner projects to make your CV look good"... Not out of curiosity, passion or fun. Nobody tinkers, hacks, explores...

Of course that's an over generalization, there's still a few young hackers here and there, but we're a very small minority.

It's sad to see, I always thought that college would be the place where I would finally find passionate people who could talk about computer science for hours and hours, but so far I haven't even met one.

Absolutely. There is no question.

At my place of work we've recently hired several 'engineers' who have been exposed as frauds who know nothing but how to beg the chatbot to do things. When you bring actual issues that require critical thinking or technical knowledge, they fold like a stack of cards and meat-proxy a Teams reply to you.

Fun part of being an Engineer is actually inventing stuff.

But we understand the world better than anyone before us, most big breakthroughs have been made, and in a lot of engineering fields we reach the limits of what is possible with the resources we have at our disposal.

So our lifetime is an age of refinement, and a lot of the engineers understand this, and how the world works, and are fine with getting large amounts of money for being more or less ok with being lazy and just solving problems to refine already available tech.

For a long time, engineers had the carrot.

Now the computer has the carrot, and it dangles the tasty promise of being able to invent new things out of thin air again in front of the engineers eyes. No more refinement, it’s eureka time, baby. AGI, the holy grail of engineering is upon us or so they say.

And engineers are now willing to pay handsomely from all the years of getting money thrown at you for doing boring refinement, for the illusion of being inventors again.

I think it’s kinda beautiful. Engineers refine the machine by sharing the only thing the machine can’t replicate by itself to teach it to solve problems better than them, so they become obsolete.

This is like watching the dinosaurs talking about the water temperature while the giant extinction-level tsunami wave is slowly crawling to landfall.

Everyone has a different way to deal with reality. I think all the events of the past few years just let more people realize how the world really is, they use the way of expression available to them to make sense of it, and society definitely had a vibe shift.

We are fighting for resources again, when did that ever go well in the last 2000 years for everyone? It is what it is.

i think the opposite is the problem

lots of people (such as yourself) who think they are really smart with no understanding of how consciousness works, confidently proclaiming what does and does not count as consciousness

“it’s just glorified autocomplete”

ok, so prove to me that the post you just wrote with your brain was doing something differently than glorified autocomplete

Going from “the human brain might use similar language techniques” to “human output is driven by autocomplete” - and implying that LLMs and the brain are therefore similar - is a very long road.

Prove to me that autocomplete contributes even 2% of what we write here, if you are so trigger-happy to assign the burden of proof to someone else.

We know that LLMs are token pattern predictors because that's literally how the algorithm works. There have been plenty of explanations by the paper authors and by other sources, e.g. 3blue1brown's summary at https://www.youtube.com/watch?v=LPZh9BOjkQs. That LLMs are "glorified autocomplete"(sic) is beyond dispute. The burden is on the LLM promoters to back up their frequent claim that the brain is the same as an LLM and they are abjectly unable to do it.

(Maybe they should have their LLMs write the proof for them if they are unable to figure it out themselves? Good luck with that.)

I’ve had to move away from thinking to meta-thinking. Most of my engineering effort is now focused on figuring out what information and process to feed a model, and then observing how it runs and tweaking the results until it’s not crap. You also stop worrying about the time it takes to code things and start worrying about tampering down on tech debt aggressively. Things that you would have never had time for before (refactoring to make the system slightly better) become incredibly possible and easy now.

I'm not saying they are alive, they have consciousness, etc.

But do we really know how LLM's(and the layers above, harness, RL, etc) work, or is there some emergent, surprising phenomena in those systems?

There was an emergent and surprising phenomena in collaboratively writing Linux kernel and low level backbone of modern operating systems, or developing mobile apps, or scaling Internet to broader population.

What is different about LLMs? Is it their social aspect and influence? All three examples above were also social to a different extent.

If so, it sounds like a fix is relatively easy

List the things you want people to know and they can use AI to learn them

From "Why are AI agents lying, cheating and coordinating?"

"One note on wording. Below, I write that these systems “seek” or “try” things. This is shorthand for a mechanism rather than a claim about consciousness or human-like intent. We use similar shorthand when describing many other situations, like a plant seeking sunlight. A system trained by trial and error behaves as if it were pursuing whatever its training rewarded, and that as-if description is what makes its behavior predictable."

https://yoshuabengio.org/en/publication/why-are-ai-agents-ly...

So there is some attempt to remain grounded. Using shorthand wording makes it easier to reach more people with your message, the whole goal of writing a post like that, but also likely means more people are mislead into treating the short hand as the truth rather than for what it is.

The typical person has not had to meaningfully learn math since the invention of calculators. Think about the consequences of that. Now imagine the effect of them ceding general thinking and creating.

I think it's got to do with ownership. The less you own it the less you can tinker with it.

Crypto was mostly grift targeting non-technical people

LLMs are a real, genuinely-useful technological innovation that happens to have arrived first in software engineering. Lots of SWEs spend their whole day talking with LLMs and feel a personal connection/expertise with the thing that's also at the top of the hype cycle for non-technical news

Use it or lose it.

> We get links like "Why are AI agents lying, cheating and coordinating?" going viral

Why would this be a symptom of losing engineering literacy?

Reply on news.ycombinator.com