4 ms·

By the time the kids are done their CS degree (4 years from now), are the LLMs going to be writing all the code anyways? If I were 18 again, I am not sure what

By the time the kids are done their CS degree (4 years from now), are the LLMs going to be writing all the code anyways? If I were 18 again, I am not sure what the answer is - would it even be a good idea to pursue this field of study if there aren't going to be many jobs post graduation... I hope I am wrong...

CatWChainsaw 3y ago
Compare that to the whole genAI art thing and people here sneering that being an "Artist" is dead as a career. I wouldn't feel a lot of sympathy if such CSbros found themselves out of a job.
nopinsight 3y ago
I think humans need to move up the levels of abstraction and master specific domain knowledge. Barring AGI, there will still be need for humans to elicit and learn the requirements from customers and other stakeholders, then translate those into more concrete specifications. Future LFMs (Large Foundation Models, since they will be multimodal) can help automate some or much work from specifications onward. Humans will need to validate and revise the results by working with an LFM-powered system (Ref: the spiral model in software engineering).
- chx 3y ago
  > then translate those into more concrete specifications. Yes and that's what code is: most concrete specifications :P Figuring out what to code is much harder than coding it, like 99.99% of the time. People worrying about ChatGPT taking away their job frankly do not understand this: my job is to think. I type very little code.
  - bcrosby95 3y ago
    Lots of new projects involve shitting out a massive amount of code though. I've certainly worked on stuff where I had precise clarity on what needed to be done and typing speed was practically the limiter. But for some other things it wasn't. I dunno. I go back and forth on the effects these things will have on programming.
sanderjd 3y ago
I think there will be more jobs making software 4 years from now. But whether most CS degree programs will be teaching the most useful skills to do those jobs, I really have no idea.
krainboltgreene 3y ago
> are the LLMs going to be writing all the code anyways? No, because that's not an expectation based in reality.
quickthrower2 3y ago
We are talking AGI. It is funny people think programmers will go first because AI can generate code. It can also generate words. Why do you need Lawyers? Politicians? Doctors? etc. when you have a knowledgable word predictor.
- pinkcan 3y ago
  models are still having trouble reasoning – current tricks still depend of feeding back an intermediate answer with a different prompt, but it is still an LLM, and most times the same LLM
  - sfn42 3y ago
    Yeah, it's a large language model not a large reasoning model. I don't really have any faith in what seems to me like attempts to make LLMs act like LRMs. I think it's always going to be a bit shit, no matter how large they make them it's always going to be an obvious faker, always going to confabulate and lie etc. As long as we are using a technology that merely fakes intelligence (using statistics to generate the most likely next token based on training data is not intelligence), it is going to be obvious that it's fake. This whole bubble is going to burst when people start realizing how overrated it is.
- l33t233372 3y ago
  Exactly. If programmers can be fully and effectively replaced, then almost all white collar jobs can be as well.
- avgcorrection 3y ago
  I’m sorry but that is an image [of a mole]. I’m a language model trained on text and haven’t been trained to pattern match on image.
xboxnolifes 3y ago
People thought the same thing many times, for many different reason, over the last decades. I'm sure they will continue to think the same thing for many more.
coldcode 3y ago
Building an entire application is way beyond what any LLM can do today and probably for years, if ever. Writing a method is basically all they can do, and even that is often unreliable. Maybe some companies will try to replace junior engineers with AI, but will learn really quickly that it doesn't work. The bad thing is that companies will do it anyway, meaning the juniors being replaced will never learn all the skills necessary to build applications or lead teams, so in the future getting anything built will become less and less likely. I've always said that if AI had to deal with some of the executives and product teams I have, with constant insane changes and shipping demands, the AI will more likely figure out how to eliminate people than deal with them.

CatWChainsaw 3y ago

Compare that to the whole genAI art thing and people here sneering that being an "Artist" is dead as a career. I wouldn't feel a lot of sympathy if such CSbros found themselves out of a job.

nopinsight 3y ago

I think humans need to move up the levels of abstraction and master specific domain knowledge. Barring AGI, there will still be need for humans to elicit and learn the requirements from customers and other stakeholders, then translate those into more concrete specifications. Future LFMs (Large Foundation Models, since they will be multimodal) can help automate some or much work from specifications onward. Humans will need to validate and revise the results by working with an LFM-powered system (Ref: the spiral model in software engineering).

chx 3y ago
> then translate those into more concrete specifications. Yes and that's what code is: most concrete specifications :P Figuring out what to code is much harder than coding it, like 99.99% of the time. People worrying about ChatGPT taking away their job frankly do not understand this: my job is to think. I type very little code.
- bcrosby95 3y ago
  Lots of new projects involve shitting out a massive amount of code though. I've certainly worked on stuff where I had precise clarity on what needed to be done and typing speed was practically the limiter. But for some other things it wasn't. I dunno. I go back and forth on the effects these things will have on programming.

sanderjd 3y ago

I think there will be more jobs making software 4 years from now. But whether most CS degree programs will be teaching the most useful skills to do those jobs, I really have no idea.

krainboltgreene 3y ago

> are the LLMs going to be writing all the code anyways? No, because that's not an expectation based in reality.

quickthrower2 3y ago

We are talking AGI. It is funny people think programmers will go first because AI can generate code. It can also generate words. Why do you need Lawyers? Politicians? Doctors? etc. when you have a knowledgable word predictor.

pinkcan 3y ago
models are still having trouble reasoning – current tricks still depend of feeding back an intermediate answer with a different prompt, but it is still an LLM, and most times the same LLM
- sfn42 3y ago
  Yeah, it's a large language model not a large reasoning model. I don't really have any faith in what seems to me like attempts to make LLMs act like LRMs. I think it's always going to be a bit shit, no matter how large they make them it's always going to be an obvious faker, always going to confabulate and lie etc. As long as we are using a technology that merely fakes intelligence (using statistics to generate the most likely next token based on training data is not intelligence), it is going to be obvious that it's fake. This whole bubble is going to burst when people start realizing how overrated it is.
l33t233372 3y ago
Exactly. If programmers can be fully and effectively replaced, then almost all white collar jobs can be as well.
avgcorrection 3y ago
I’m sorry but that is an image [of a mole]. I’m a language model trained on text and haven’t been trained to pattern match on image.

xboxnolifes 3y ago

People thought the same thing many times, for many different reason, over the last decades. I'm sure they will continue to think the same thing for many more.

coldcode 3y ago

Building an entire application is way beyond what any LLM can do today and probably for years, if ever. Writing a method is basically all they can do, and even that is often unreliable. Maybe some companies will try to replace junior engineers with AI, but will learn really quickly that it doesn't work. The bad thing is that companies will do it anyway, meaning the juniors being replaced will never learn all the skills necessary to build applications or lead teams, so in the future getting anything built will become less and less likely. I've always said that if AI had to deal with some of the executives and product teams I have, with constant insane changes and shipping demands, the AI will more likely figure out how to eliminate people than deal with them.