Hacker News new | ask | show | jobs
by taylodl 611 days ago
You Rob Yourself of Learning Opportunities

Not really. When I'm doing something new to me, I do not let AI write that code, otherwise how can I evaluate the code that it produces?

Skills You Already Have May Atrophy

Of course they will! Just like they do when you use a library of routines. Nobody writes merge sort outside of their algorithms class. Could I sit down right now and write merge sort? No. Do I care? Not really. Would I let AI write it for me? No - see above: I don't let AI write code I'm not familiar with. I'd prefer to use a library routine and move on. Failing that, I'm going to have to re-implement merge sort. The upshot is since I've implemented before, it'll all come back to me, and I can quickly write it again.

Afterward, I might even let AI take a crack at it and compare its code to my code. I'll use whichever one is best.

You May Become Dependent on Your Own Eventual Replacement

People said that about Intellisense, too. People said that about Code Wizards. I remember when MFC was first released in the early 1990s and with one click of a mouse button it could create a Windows Hello World! program. I remember spending days going through Petzold's book learning how to write a basic Windows program. Now all I have to do is click a button!

On a side note - when is the last time you ever bothered yourself with a windows message pump? Processor caching and branch prediction optimization? How about loop unrolling optimizations? There are times when such knowledge is needed, but I'd wager 90+% of today's programmers have never implemented such things outside of the classroom: if they even did that! And they're not using AI.

Your Code Will Not Be Respected

Are you saying AI-generated code can't pass the Turing test? That's - interesting. You talk about solutions, but in my mind, coding isn't solutioning. That's what architecture and design is for. Coding is implementation. That's why coding is referred to as a craft.

Can AI get good at the craft of coding? That's a question of automation and to answer that question I'll point out there aren't too many black smiths or white smiths around today. Or weavers, for that matter. These are all trades requiring a high degree of knowledge and skill, like coding. And they're mostly gone.

But, mostly gone isn't the same thing as completely gone. People still enjoy learning these trades for fun. Imagine something like Williamsburg for Silicon Valley 100 years from now. The tour guide will explain how people used to write programs. The guests will ooh and aah that people could ever have done such sophisticated tasks. Maybe one of the guests will be intrigued and learn the coding craft.

This has been going on for millennia. The only reason you're scared now is because there are fewer and fewer crafts that can't be automated, and we've built an economic system around scarcity, especially the scarcity of labor. We know from history when such tectonic shifts happen its very disruptive to civilization. Stability is lost and wars are fought. I believe that's what you actually fear, and not AI per se. You fear the change that comes with AI.

On a positive note, you're alive at what will likely be one of the greatest inflection points in human history! How lucky is that?!

2 comments

> On a positive note, you're alive at what will likely be one of the greatest inflection points in human history! How lucky is that?!

That remains to be seen. Not all inflection points are good things. If we're at a major inflection point with generative AI (and I doubt that we are), it's possible that will end up being a bad thing, not a good one.

I'm not saying that this tech will end up being good or bad on the whole, I'm saying that it's far too soon to be able to predict that.

I do agree, though, that it's not the technology itself that presents risk, it's people developing and using it that could go very very wrong.

Absolutely! Hence the fear. Technology is a double-edged sword - there are always upsides and downsides. With most technologies it seems as though we only discuss the upsides. Rarely are the downsides discussed, or even mentioned. When it comes to automation technology though, the focus always quickly shifts to the downsides. I think that reflects a deep distrust of civilization and the status quo.
I think with genAI specifically, a whole lot of the distrust comes from two things: the movement on this front is coming from SV, which has earned a lot of distrust from the general public, and the major players in genAI (particularly OpenAI) have themselves been very clear that they believe the tech is dangerous and they're intentionally gambling with the lives of everyone else.

That, and when people have expressed their natural fear that genAI and LLMs will adversely impact their abililty to make a living, a common response -- especially early on when the fear was much higher -- from the pro-AI crowd is (paraphrasing) "too bad, you can't stop progress". That's not a reasonable response, and it's one that effectively says "you don't matter, sucks to be you." It's not far from saying "AI is your enemy."

I like to think we are simply at the next step of abstraction that will help explode faster development across the board. It is well overdue. Think of the step from compiled languages to scripting languages - so many engineers (myself included) know fuck-all about about garbage collection and pointers, but we can still get useful work done. For some things, you want or need to use a compiled language for greater control/speed.

Even with the rapid rise of AI, I think we will be stuck at this stage for quite a while. Yes, AI will be able to made a button for your app or make your app completely, but what people always want is greater control. Someone needs to be making business decisions about how things work and interact, or how things are architected for anticipated load. So a lot of easy and repetitive work goes away but demand shifts to more deep tech engineers and more business/design engineers.