Ask HN: Jaded with AI – Alternatives?
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.
I feel like you've assembled a group of criteria (some of which are mutually exclusive) and trying to make a job fit all of them. I wish you luck in thst, but it may help you to identify them to understand what to compromise on.
Significance. You don't want to be a cog in a machine. Since it's easy to find significance outside of work, and since almost no work is significant this may be the easiest to recalibrate.
You want job security with high pay. This comes with working at large companies. They're less likely to go under. With some luck you can avoid the regular culls. These jobs require zero creativity (and likely oppose creativity. )
Creativity is risky. Because what you create may have value, or not. If you need to be creative, that's also easy to do outside work. Take up painting, music, pottery, whatever.
You're still young. Understand that most of the work you do over the next 40 years will just be work. It can supply money. It can be made "secure". But dont expect your work to give you significance (find that elsewhere), don't expect your work to satisfy your creative urges (find that elsewhere.)
Life is more than work. Figure out what you want in life and do multiple things to satisfy those goals. Don't place the burden on work satisfying them all. That route typically does not reduce well.
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.
Thanks! While I appreciate this type of role, for family reasons, near-complete job security is a must for me, so this option is likely not suitable for me. On top of that, I'm also very much a risk-averse person and don't think would be able to survive the pressure that naturally accompanies freelancing or contracting.
> near-complete job security is a must for me
Unless we understand vastly different things by "job security", I have some bad news for you.
Recent years showed that even large, profitable companies with significant cash reserves -- that had a reputation of high level of job security -- are ready to do layoffs as a sacrifice to the gods of Wall Street.
At least FAANG and well-funded startups did a lot of overhiring. When thinking about stable jobs I have classically boring ones in mind, such as ones in utilities or defense
I guess _relative_ job security is what I'm hoping for. I realize there are few career paths that guarantee employment (perhaps medicine being the main exception), but despite the FAANG layoffs and whatnot, wouldn't it still be correct to say working at a somewhat well-established company is safer than working independently?
Not necessarily. I think that employment has a higher job security floor but a lower job security ceiling.
_In a way_, employment is like having a single client. You have smaller variance month to month but are more prone to unexpected events in the company you work for.
You can build financial resilience by having multiple (preferably diverse) clients.
Another thing to take into account is that you are not in control of being laid off or a client not extending their contract with you. But you are in control of your own savings. And this applies in both scenarios.
Though if your clients share correlated retention factors, systemic changes could still result in large drops of revenue.
I hadn't thought of it in this way, security floor vs ceiling is a very interesting way to look at the matter.
Although it's difficult to monetize, you could (for the time being) work full-time on your open-source projects. I recognize that might not be the easiest path, but if you pull it off (you can do it!) then you have - in my view - the best position in tech.
In summary, install linux, move to a cabin in the woods, become open-source developer.
>I'm getting the impression that the field is over-saturated, with new research being published on a daily basis
The 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.
I see your point - where I formerly worked, computer vision relied as heavily on more traditional CNNs as it did on more cutting-edge architectures. Yet, due to all the excitement over transformers et al., I've inadvertently lost interest in other areas of deep learning because of how overshadowed they feel.
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.