Ask HN: What is the outlook for a new career in ML or DS?
I've been slowly hacking away at a math, coding and ML knowledgebase in concert with efforts to increase my work endurance, with the intention of accruing hard, demonstrable skills that could serve me well in any capacity. I assumed this was the most robust use of my limited energy because these skills could be used in almost any white collar position which uses computers to perform repetitive tasks or intersects with structured and unstructured data.
At least, that was the idea x years ago. Now it seems we're on the verge of a centralizing and commoditizing revolution in ML and UX where entire swaths of knowledge producer skills will become obsolete.
Where will that leave people trying for either ML engineer or more data analysis focused DS roles? Or even just everyday utility scripting/automating powers? Can even the latter two can be largely replaced by hybrid finetuned multi modal language/code models. I feel a bit lost and like I have wasted my time.
Will only elite scientists in the top percentiles of skill and resources be needed or will the future have room for more pedestrian aspirants?
22 comments
[ 83.2 ms ] story [ 1365 ms ] threadThere is still dumb money in Proof of Concept kind of work.
And I honestly enjoy doing this kind of stuff.
This is the number one thing that doesn't make me want to even try the switch. I am trying to learn some machine learning for educational purposes, but I have free choice to use what I want: F#, Python, and Racket.
If you aren’t one of those, being a regular swe is much better. A lot of ml roles are using canned models, and increasingly so. Many data science insights are ignored by the business, or the statistics don’t work on such a biased sample.
It's often even less in my experience. Despite having a "unicorn" skillset (soft-skills, advanced degree, domain experience, and SWE experience), I make about as much as a vanilla SWE. There are a huge number of inexperienced PhDs that want into the field, and we are flooded with resumes every time a DS leaves. Also, most of the time, models don't really matter. What makes or breaks most DS projects is soft-skills, stakeholder management, and data cleaning / feature engineering.
I have the same impression. I did my Master's degree in data science, but I quickly realized that coming up with ideas and running the models is the easy part. Doing the engineering work + synthesizing everything such that value creation occurs is more difficult.
I'm happily doing mostly data engineering + stakeholder management instead of hyperparameter tuning.
Agreed. I actually like that in DS you can have a job where you are involved in the end-to-end of a business problem and that you need to have a mix of skills (e.g. Act like a Product Manager and an Engineer) to succeed. And it's not just "Today I get to crank out yet another tile on the Kanban board."
I'm talking more about "Data Science/ ML Department at a typical company". You won't see salaries above comparable SWE roles there. Most likely, the SWEs will be better off.
What I'd say is since everyone is shedding employees that the next 6 months - year or longer, however long the recession is bad lasts, will be bad for all tech people. But as the economy improves and the recession recedes, good companies are going to want to know where to spend money to capture that growing market.
> Now it seems we're on the verge of a centralizing and commoditizing revolution in ML and UX where entire swaths of knowledge producer skills will become obsolete.
this is kind of vague and uncertain, I certainly wouldn't plan my career around it. It's not a good enough reason to change career paths IMO.
A lot of the time you could train model prepare pipeline and test enviroment for 3-6 months before you get good enough result to push it into production. And it can get extremely stressful rly fast if you care about that, because not every model is good enough for production so once per year you could have as low as 1 or even 0 models that are working fast and good enough for proper usage and this can burn you after just a 1-2 years of work in the field (I know at least 4 people who just drop ML and go for SWE after 1+ year of ML work and all of them are a lot happier with classical backend/devops jobs).
In SWE after 2-3 days you can have small stuff working fine and after few weeks push your small code into production codebase fixing some stuff or optimising smth as there is a ton of potential in almost every codebase for "easy" upgrades in ML space everything is extremely competitive your results are "state of the art" or they are not if you want to upgrade model it better be sota or you would get asked "why we don't just implement/use ...?".
I still love my job but for sure but I prefer working with ops, classical SWE and deployment then model training, optimising and learning/collecting new datasets (I have 5 years of commercial exp in ML/DL maybe it gets better after 10 years or I'm just boring out who knows)
>A lot of the time you could train model prepare pipeline and test enviroment for 3-6 months before you get good enough result to push it into production. And it can get extremely stressful rly fast if you care about that
Some of the the major job challenges in ML/DS are that the field is new, the number of jobs are far fewer than SWEs, the teams are small, and the people who can exert control of you have wildly unrealistic expectations of what it takes to build a useful model. So it's easier to land in a crappy job where stakeholders add to your stress because you are not "delivering" according to their definition of "I thought ML would solve all my business problems in 2 weeks."
I personally prefer ML/DS to what I see a lot of SWEs do. Yes, it can take months to get something working (or maybe it will fail after all that effort), but I'm also involved in the end to end of what I'm building and I get be part of defining the the "what", "why", and "how" and not just crank out yet another story on a Kanban board. What really helps me in my job is having a management chain that pushes back and educates stakeholders on why things take time and why that time is required to have a certain level of quality.
There are plenty of SWEs (on HN and elsewhere) who say they feel like factory workers. Somebody else makes a bunch of decisions and the SWEs simply churn out features and stories. So the quick wins turn into an endless flood of drudgery, just like how when you learn to fix a toilet you feel great, but if you fix 20 toilets a day, it probably becomes monotonous pretty quickly.
IME most people have a bit of culture shock when they first get an ML/DS role. When you learn ML/DS, there's a huge focus on the coding and mathematical parts of the job, and not on much else. When you get a job, suddenly you're exposed to everything else and you realize why every data scientist says that ML is only a tiny part of their job.
There's a lot of variety in DS jobs, but one of the things I try to explain (when somebody asks me) is that the real skill of every DS job I've had is "Taking a vaguely defined business problem and working with stakeholders to come up with a solution that happens to use code, math, and charts as the path to that solution." In reality, to succeed in as a Data Scientist, you have act like a Product Manager+Product Owner+Data Analyst+Software Engineer+Data Engineer+Data SME.
This is why so many attempts at "We can make your Business People Data Scientists" products haven't taken over. You might be able to take over some of the boring parts (e.g. AutoML, doing the parts that Data Scientists hated doing anyway), but I've never seen a piece of software that could tell the business users that they don't have the right data to measure what they are trying to measure.
I've even seen commercial AutoML solutions lead to business people realizing they need to hire Data Scientists. This is because once you use AutoML, you realize you need somebody who actually understands the data and the process to really trust the results.
If "hybrid finetuned multi modal language/code models" can replace what competent people in ML/DS can do, that technology is going to replace a lot more professions than just ML/DS. There's going to be a job apocalypse for lots of professions, including SWEs.
I think from a career standpoint, ML/DS is a bit of a mess because it's a new field and businesses are still trying to figure out the best way to get value out of it. So there are a lot of pain points for people working in the field. Compare that with Software Engineer, which is older and a bit more mature. But I still think ML/DS is a field worth getting into if you can sort through the noise of which jobs are good and which are crap.