TechEcho

Hello HN,Since a young age, I've been interested in machine and deep learning. I’m currently in the second year of my Computer Science BSc (Toronto, Canada) and already have almost 2 years of experience in industry (computer vision + NLP) and over a year in academia doing AI research (both full time). Additionally, I have quite a few open-source projects (all DL-related) that have garnered over 1,000 stars in total, and some are very well-known in their respective niches. Lately though, I'm getting the impression that the field is over-saturated, with new research being published on a daily basis, and I identify as nothing but a cog in the machine as an AI developer. I understand that all domains are affected by this phenomenon to some degree, but in AI in particular, my work doesn’t feel personal at all, and to myself, I ironically seem like a robot that trains a vision transformer to do classification, fine-tunes an LLM for certain types of documents, makes architectural changes for a tiny improvement in performance, etc.What are alternative branches in CS that you suggest I consider? I have two chief priorities:* Creativity: I'm not seeking a typical software development job such as full-stack developer. Instead, I'm interested in opportunities that require creativity, almost like puzzle solving, and don’t become “routine” after a while. * Industry: My goal is to work in industry, not academia. This is not because I don’t enjoy research (in fact, I prefer it to applied work), but as reluctant as I am to admit it, salary does play a role in my decision making, and I’m aiming for six figures.To give you a concrete example: I love work in logic, programming language research, theoretical computer science, and so on because they satisfy my first criterion, but it sadly appears that employment opportunities are mostly confined to academia?I really appreciate your thoughts and feedback.

3 comments

throwup238vor 2 Tagen

First the good news: There is a relatively well beaten path to becoming an independent consultant/contractor who gets to pick and chose their clients and projects. You probably won't work on intellectually challenging or stimulating projects 100% of the time, but you can schedule them so that there's usually an interesting contract running in parallel to the ones that reliably pay the bills. There will be dry spells where you have nothing but boring work but as you get better and have a proven history of delivering, that work improves.Now the bad news: getting there takes a lot of work - most of it unrelated to technology - and requires a certain personality. You need to get out there and network till your ears bleed, get good at sales and marketing, manage the stress of not having a W2 salary and health insurance, learn how to bill clients and convince them to actually pay, maintain relationships with current and past coworkers, and a hundred other yaks you have to shave. It's death by a thousand cuts but if this lifestyle fits your personality, you can achieve a very interesting and varied career.If you spend your college years and summers building out that network, you can even start this process shortly after graduation but it is very risky. Depending on how the job market looks in a couple years, it might even be the optimal thing to do (it's what I did after the 2008 GFC), but this is not a path for the feint of heart. It is hard and much more stressful than collecting a paycheck. I was personally relieved to get a FTE position after years of consulting.

评论 #43909157 未加载

pizzavor etwa 17 Stunden

If you want a hard problem that's far from predictable with tangible utility, why don't you work on something that helps humans with human problems, but whee you can still use your talents? Some research that could help understand a disease, or a new kind of medical technology, or something like that. I don't know how well it aligns with your salary goals but it might be meaningful.

ActorNightlyvor 2 Tagen

>I'm getting the impression that the field is over-saturated, with new research being published on a daily basisThe research being published is mostly about optimizations of transformer models.There is a lot more research in other areas. For example, everyone seemed to forget about MuZero, but that was one of the most important breakthroughs in the last decade, because it was a model that could recursively search.So if you think about tasks like driving a car, a model that can predict the evolution of a physical system (much like MuZero can predict the evolution of the game board), and make a decision, is going to perform much better than any of the vision based transformer models that are in cars today.But there are other areas of research as well, namely in hardware with specific instructions for general compute.

Ask HN: Jaded with AI – Alternatives?

3 comments

Ask HN: Jaded with AI – Alternatives?

3 comments