I just finished a PhD in stats, and before that, studied computer science. I had planned to go back to industry to do software engineering or be a data scientist afterwards.
However, it seems that in the meantime, vibe coding has taken over the software industry and swallowed most data scientist roles. Few open positions, and the ones that do exist are just pushing a button instead of thinking and solving problems. Is this a correct assessment? Should I try to get back in or just find a new career? I am not too keen on staying in academia either.
AI poses one major risk to SE in general: the dropped need for low/entry level employees.
It used to be, a senior would have a team of mid- and junior engineers, senior would be in charge of the big picture while the team learned, and delivered the actual low level tasks.
But with AI, companies started hiring only seniors. Which is an issue - now the market is full of “entry level” graduates who simply can’t find a job because nobody’s hiring. And seniors are left wrangling AI agents that don’t learn, don’t work autonomously, don’t ask questions, don’t bring ideas.
Skill issue, in the most literal sense possible.
oh yeah? when was the last time your LLM asked you a question, unprompted? when was the last time your LLM had a line of thought and came up with something groundbreaking, something unique, on its own?
The coding agents will do that. Ask which of a couple of approaches you’d like to go with.
I think they meant questions unrelated to the current tasks.
AI agents and Junior engineers both have their pros and cons. But this column definitely goes to the human engineers.
I had no idea at the time that I’d miss those annoying questions once all the juniors are gone.
Yeah, that’s a good point if that’s what they meant.
Woosh, apparently. Let me explain.
LLMs are virtually nowhere used without some sort of harness these days, most prominently agent harnesses.
Skills are a harness artifact for reusable prompts where we encapsulate small terminology or large processes. People who use them quickly come up with or reuse a skill that will lead the agent to lead interviews to get a common understanding of key issues before going to action.
As for no 2, idk why that’s so relevant for automation. Often after an extensive interview the agent proposes solutions like I’d envision them, but much more thoroughly written out, in a quality and consistency that maybe a much reviewed and discussed RFC reaches, but I typically not. Sometimes it suggests solutions that I didn’t think of, but when poked it’s usually an elegant one. I don’t suppose the agent came up with these things itself, but it certainly has read many more books than I did, so that’s not objectable.
In my experience the AI agents do if you ask them. Rather than giving them a problem and telling them to solve it autonomously, I ask them to give me feedback on my plan and propose potential alternatives before starting any coding, and often I get good suggestions from them this way.
If your idea of an idea is “concept that I personally am not aware of but which has been conceived, elaborated on and documented by another human being”…
Then yes, totally!
The AI can extrapolate, but even if it couldn’t, ultimately I need an idea that works, not necessarily a completely original idea that works (and most humans don’t come up with completely original ideas very often either).
Yeah, I was gonna say, most engineers are gonna be taking a lot of code from the internet anyway. Most problems modern companies face have already been solved in one way or another, its just figuring out what fits your specific product and needs. That being said, I’m pro “ai as a tool in non-creative fields”, but it certainly doesn’t replace engey level engineers. I see it closer to a calculator or like factory machine. It can cut out the grunt work and summarize things well, but it takes a looooot of steering to actually work. And the engineer can maybe learn some new concepts or techniques along the way
Sauce for the first sentence?
Well, I have my personal experience, but there are also all those math problems that AI has been solving recently… Here’s one mathematician’s discussion, with links to blog posts by other mathematicians who had spent years unsuccessfully working on some of those problems. He says:
That’s what’s gotta sting the most about this. Imagine you worked a problem for a long time… and then The Giant Pattern Matcher With Zero Fluid Intelligence actually finds something.
If it did find something, that implies the solution already existed in bits and pieces in the published literature. You would “just” have had to have read about it, and remember it, for successful application.
But suppose you had succeeded in that. You would have published a paper that enumerates a bunch of results some other people had reached, then plugs them together for a supposedly new theorem. Sort of like importing a dozen Python libraries, adding around fifty lines of glue code around them and then going, yeah, here’s a “new” thing, Fields Medal plz!
How much remaining interestingness did your chosen problem really have if the new math you had to create for it fits on the placemat buttressing the campus cafeteria sandwich you had for lunch?
What I’ve seen appears to indicate some genuinely new approaches by the AI, rather than putting together existing information. That blog post, for example, quotes some running commentary from a mathematician studying the proof of the theorem that she had been working on for years:
Note that this particular proof is correct, according to that same mathematician.
The issue with longstanding problems such as these is that yes, they can be used to establish evidence of plagiarizing if your LLM is simply regurgitating stuff, but suppose you cannot come up with any… you don’t really know if it’s because this particular line of thought is new, or because you just hadn’t read about it.
That’s true even if you’re an expert practitioner; LLMs have been trained on All Of The Text and have superhuman recall after all.
Sounds like a pretty smart mathematician
The blog post was written less than a day after the proofs were released, and apparently they’re very hard for humans to follow. Now that a month has passed, there are human mathematicians’ explanations of how the proofs work up on the internet.
I think the blog says exactly the opposite, the AI is very bad at explaining its solution and core idea.
I think the interesting part is that explaining ideas is something that don’t seem to have an obvious reward for RL to train on, unless mathematicians keeps handing them training data.
(OMG of course this guy is a rationalist 🫠
Both can be true at once - it is coming up with original ideas but explaining them poorly. I’m not a mathematician but I find that in my field (software development) the AI is much better at writing code than it is at writing documentation - it will write documentation if you ask it to, but the documentation will be unfocused and difficult to read, with explanations of things that don’t matter and garden path sentences. I got this sentence from it recently:
That is actually not gibberish - it’s technically correct, but it certainly isn’t clear.
Maybe I miss understand what “extrapolate” means in your context. To me that means “extract the core idea and explain/apply them”, as AI do not learn by producing output (unlike human), applying core idea requires it to be explained in clear language.
I was arguing that AI, at present, cannot effectively extract core idea which is what I would characterize as “extrapolate”.
“one mathematician” who happened to work at openAI, btw.
You can look at Terence Tao’s blog instead if you’d like. Just from the last week or so, there are the following headlines:
I looked at Egoless Mathematics in the AI Era, and it was AI slop. BTW, his blog is full of guest posts now
And that alone is worth a lot
No one is arguing that an “idiot savant in the Hyperbolic Time Chamber” cannot be economically useful, but does it make plain old intelligence obsolete? I’m not buying it (yet).
Uh, idk, I’m not the one arguing for it either.