Ask HN: I don’t want an LLM career | Hacker News Reader

Ask HN: I don’t want an LLM career

11 pointsbosch_mind12 comments

ChatGPT is interesting. It’s a decent utility for my work, which is networking and distributed systems stuff. However, I am not particularly interesting in a career in AI or ML.

Does that mentality seem detrimental? I wonder if people who used to be expert Fortran developers or COBOL developers felt the same when new technologies started coming out like Java, the web and C++.

View on HN

> It’s a decent utility for my work (...) However, I am not particularly interesting in a career in AI or ML.

I don't get your point; there's two different things

1. Using LLMs to assist you with your work. (you're saying that LLMs are useful for this)

2. Working on AI/ML things directly (being a researcher, data scientist, ML engineer etc.) But this isn't your job anyway, so why the post?

I don't want a mobile app dev career, that doesn't prevent me from using apps on my phone.

1 for me is real real nice.

2 is no go, ill leave it to people smarter than me in that field haha.

Just like fortran and COBOL themselves, the pre-AI / ML stuff will be around for much longer than we will be.

I've been having this weird feeling lately which I figure must also be what the fortran / cobol / mainframe folks felt... I guess I learned to code in the late 90's / early 2000's with perl, linux, C / C++, shell, etc... which was basically just the mainstream way to do stuff at the time, nothing terribly esoteric.

Nowadays, I work at BigCorp places that have a lot of systems that have probably evolved from similar times, and it's easy for me to troubleshoot / diagnose / improve them... but the younger folks in their 20s seem pretty lost with really basic shell stuff, how to grep, how to ssh / rsync / copy files around the place, how to #! + chmod +x scripts, etc. Even after a few years on the job, they're not proficient - it's weird, because the systems still exist and probably will for decades to come, but I feel like we're only a few years away from this stuff being as esoteric (but also, still relied upon) as cobol + fortran.

> Even after a few years on the job, they're not proficient - it's weird

It's because they focus on high level stuff and the latest frameworks.

I think it is fine. There will always be work for people to program stuff, until the ML becomes sophisticated enough to talk to people, understand their needs and create programs (or indeed a bunch’o’weights) to solve a given problem. By then most jobs will be in jeopardy because why do you need say an experienced family lawyer when an AI could do it better?

Machine learning is mainly linked to knowledge discovery, data mining, and related applications. While anyone can learn it, it's most relevant for jobs like data analysts, data engineers, and data scientists. But computing has many more branches. There will always be a need for people to work on infrastructure and develop new technologies, showing that the computing world is vast.

If you're not interested in a career in AI or ML, that's not necessarily detrimental. The tech field is always evolving, so it's important to adapt and learn new skills. Continue to thrive in your area of expertise/interest, while staying updated on the latest advancements.

I once knew a guy who worked on a very specific old technology. He would move from city to city, company to company as they shuttered working on this thing no one else did anymore.

I think we've got space left without LLM's for a while. Possibly 20 or 30 years even on the outside. Segments of the economy that ban them etc, companies that are slow to adopt the tech.

Innovations take time. Even now, I don't know of many companies that are encouraging tools like GitHub copilot yet and that has been out for over a year now.

Lots of people will move to LLM's (personally I love them) but I think we'll have a long transition period where it's optional to use them.

There are hundreds of thousands of C++ applications out there running everything from factories and warehouses to airlines, airports, large enterprises, cars, aircraft etc. It’s like COBOL but times 10000. C++ developers will be needed for a very long time. Not to mention that new C++ projects are being started every day. And yes you will probably use LLM as a tool in your tool box working on C++ code. But it’s just a tool. There are hundreds of other tools already being used by C++ developers to automate things. LLM is just another one.

There were plenty of COBOL devs that had no interest in learning newer tech.

But did they go out of business ?

i think it is only a matter of time (5 to 10 years?) before we have robots that can interact with physical envionments and communicate with us.

its not detrimental, maybe just different tastes.

i dont feel like studying it but i love reaping the good parts of it as an end user.

Reply on news.ycombinator.com