AMA
Yesterday
1722
I’m a professional coaster AMA
Ask Blinders
Yesterday
976
Tipping culture is really getting out of control! Waiter gave me ‘a look’ because I tipped her 10% for ‘BAD service!’
Tech Industry
Yesterday
673
Which company is best to have on your resume?
Tech Industry
Yesterday
2250
TESLA UP 14% AFTER HOURS 🎉🎉🎉🎉
Tech Industry
Yesterday
879
The new Tesla Model 3 P goes from 0-60 in 2.9 seconds
Everyone and their mom wants to slap ML on their LinkedIn now. There are of course some people who have done research or university work in the field but otherwise, is machine learning just a buzzword that people are throwing around? Especially as ML becomes a library or toolkit that someone can just throw in somewhere?
Those who don't know statistics are damned to reinvent it, poorly. Or post stupid polls about it on Blind.
As an AS I spend a lot of time fixing ML stuff done by SDEs. Most SDEs/managers think it is just pulling in sklearn or tf like any other software dependency then surprised it does not work
Machine Learning is so ridiculously easy: 1. Download ML library 2. Call ML Library For some reason everyone groups people doing ML with the people writing these libraries or doing ML research
That is a fine way to run a regression or train a simple classifier. Many simple problems are amenable to that. Most hard problems are not, at least not if you need state of the art results and small improvements in performance are associated with millions of dollars or more on the line.
Hard problems are borderline ML research tho
Do you know how the math and proof behind dynamic programming, shortest path etc works? I forgot all of it but I know how it works in code. I suspect ML algo will just be as standard as that. No point of understand beyond code and logic level.
Well same way as you call alchemy is just about jars and herbs, and anyone can make portions(models). But the experts make them work magically, the rest are just happy with their shits' color and smell
The problem I've seen in some groups/engineers/managers is that they do not realize ML expertise is useless without a good understanding of the problem domain. I've seen ML team built and deployed into solving all kinds of issues and they jump in and try all kinds of bs without actually spending time to understand importance, dependence and interaction between various features used, or which other characteristics from the problem space should be built out as features but are missing. Without that knowledge, and just looking at yield curves and trying to play with models will just be BIBO (Bullshit In, Bullshit Out).
I can do linear regression so I have machine learning on my resume
i can write hello world in for loop in python. so i have python on my resume.
If you do the right linear regression on the right data to solve the right problem, Im fine with calling that ML. If you throw garbage data into tensorflow and get garbage out, that's when I have a problem with it. It's more about the thoughtful application to solve real problems than about the complexity of the technique IMO.
Your poll options suck. It can be both a real thing and a buzzword of the moment. The world isn’t black and white.
Modern ML is non-Statisticians pretending to be Statisticians, and failing poorly. The only use cases it makes sense for are Computer Vision and NLP, otherwise Classical Stats is way more powerful. ML practitioners don’t test and control for dozens of effects in the data, and the off-the-shelf models they use don’t either.
Of course it’s the buzzword of the moment but knowing a few lines of ml can help engineers make cool stuff that they otherwise wouldn’t.
Cool stuff like...?
Anything with a recommender system, search engine, personalization, speech recognition, TTS, and spam filters in them involve machine learning. The quality of these products is directly related to statistical algorithms, I.e. machine learning. I am sure you can come up with a few names of well known products made by large companies that rely on some of the above types of systems, yes?