I feel so validated by this article. I took two semesters of machine learning electives for my CS masters and feel nearly as ignorant and mystified as when I started. I worked so hard to create something useful and at the end of the day, my work felt like it was 96% example code with modifications hacked in to make it work. And in the end it was still terrible! At least now I know what people are talking about when discussing neural nets and their inner mechanics.
For now, ML research and development is too complicated and frustrating for me to dedicate the time and energy to become skilled in it.
Well that's because ML isn't really software engineering. Unfortunately, it is also software engineering as I often have to remind my colleagues coming from algebra / econometrics / statistics sides who are happy to shove all kinds of horrible code in.
What I've found in reality is that machine learning is 99% data cleaning scripts and 1% the part you're talking about. I've also seen the heavy duty statistics people writing data cleaning python scripts which probably leads to a lot of frustrations :)
I think what may be understated here is that while it’s true that ML is mostly date cleaning, data cleaning is not easy. There are a million little decisions made and it’s rarely clear which ones are most effective. Experimenting with various techniques is great but the iteration times and cost are usually too high to try more than a small handful of approaches.
> 96% example code with modifications hacked in to make it work.
This is 96% of how ML is used in practice by companies. Most parameter optimization should be automated by whatever library you’re using, beyond basic sanity checking. The challenging parts are creating high-quality training data and deploying the models efficiently at scale.
I'd be surprised if it doesn't suffer the same fate as graphics programming. The lower level stuff is what brings in a lot of talent but the producers often have little knowledge of how things work but just wire together some libraries in a GUI.
I don't think it's as complicated as it might seem, if you break it down.
I think the real blocker is time; modern software devs are expected to be across the whole stack. You can't have one person write your backend, frontend, infra, db admin, build ml architecture, train model, etc. It's just too much.
It's just specialising really, wide and shallow vs narrow but deep domains. Forefront of ML stuff requires researchers specialised in that domain, same as any I guess.
It goes without saying that I could easily Google this myself, but I’ll ask anyway for those who are also wondering: what is the significance of HTML5 for this implementation? I’m not a web guy have gotten by with simple html css js and occasional templating when needed. Is HTML5 supporting some native programming?
Are howto articles considered spam? I found it useful and wanted to share it as well as save the link so I could find it later. If this post violates some code of etiquette I’ll not do it again.
Yeah it’s just a tutorial. I don’t really have any kind of platform for saving links I find useful so figured I’d post it here to save it for myself and maybe someone else would like it. I expected it to be completely ignored.
Perhaps this holds true more to fields that stem from hedonic experiences themselves like the arts, sports, games etc. I felt this way about music after graduating from music school. I became bored and jaded to music and had to spend several years not consuming music at all. Eventually I regained my love of it.
Conversely I don’t think this applies as much to sciences. After I pivoted away from music to become a software engineer, I discovered a world that never ceases to captivate me and elicit curiosity no matter how much I grow. In fact the more expertise I’ve developed, the more intense my interest becomes.
I work in tech and the arts and my experience is that some of the most emotionally disengaged people are in STEM, not everyone of course, but comparing the two, it’s not even close. It’s actually pretty rare to find anyone serious in the arts emotionally dimmed, a large part of the job there is dialing into emotion. That said, seems people at the top in STEM (shaping the field) and on the fringes (doing experimental, cutting edge work) are usually experts and highly emotionally engaged. Take it all with a grain of salt, I guess.
This post says everything I’ve been feeling. Ive been doing a lot of work with the Dronecan/UAVCAN protocol and nearly all their documentation is relegated to a Discord channel with no discussions about problems I’ve been trying to solve. A LOT of people have asked the same questions I’ve been asking and not one person has responded. Probably because their questions are ignore for a short time by busy people and then the questions are buried in a deluge of other questions from users. No doubt people have solved the same problems I’ve been coping with but have neglected to post what they’ve learned because discord does not allow you to make focused threads on a very specific topic without the content being buried. I’m getting more frustrated that projects are ditching the tried and true old school forums of yore where conversation can be focused and easily found.
The first time I took a rather complex library, learned it inside and out. I read documentation, did examples, wrote my own examples, implemented quick and dirty working versions, refactored, sometimes started again from scratch and eventually clicked with the material. It started slow, but once I really began to understand things, the code started to write itself quicker and quicker until I had stable production code that I’ve been maintaining for several years now. The beginning stages was like solving a big puzzle and built a lot of confidence.
Also in school we had to do some big projects like make a virtual machine from scratch and implement things likes call stacks, threading and memory management with our machine op codes. Doing a big long project that really pushes you out of your comfort zone is a BIG help and also can be an opportunity to do something really fun.
Edit: I left out the important role of asking questions, talking with others and even taking the time to compose a forum post or issue on github when it seems that I’ve truly exhausted all my options. Bottom line is, when I took my time and really tried my best to learn something new/difficult there was always an eventual breakthrough and consequential boost in confidence
I was big into Tintin as a kid as well in 2000. I checked out library books and eventually began to purchase my own volumes every time we went on a family road trip. My nephews recently found them at my parents and I’m happy to see they are enjoying them as much as I did.
For now, ML research and development is too complicated and frustrating for me to dedicate the time and energy to become skilled in it.