Q: How easy is it to create machine learning models nowadays?

= TensorFlow

That's ML commoditization.

Why is that a bad thing?

unless you use linux and it isn't ubuntu !

@MadMadMadMrMim dunno what you're on about, I'm happily using TF on CentOS 7

@RememberMe watering down the field making it easily accessible to masses, getting more and more low quality projects with no effort. At last that's what I found out from reading candidate resumes.

@rantsauce ...so? Making ML easier is beneficial for everybody. The real problem here is the resume process, not Tensorflow, and said pretenders should be easy enough to weed out with basic questions. If a candidate with trash projects gets through the selection process, that's on the selection process.

If nothing else, TF (and PyTorch etc.) is a massive productivity boost because it frees you from thinking about how to implement a model architecture. It's just a library for working with a kind of computation. For that matter it's not even the worst "offender" when it comes to people pretending to know ML, that would be scikit-learn.

@RememberMe I think rantsaucr believes the commercials about machine learning experts being a thing in private industry

@rantsauce and btw I’m still confused by the idea that tf building on red hat means any of the shit your poser ass is talking about

Oh you fucking bots!

@MadMadMadMrMim I'm not a believer but, what in the name of Jesus Christ our lord and savior and the name of the skies, heavens, hell, angels, fairies and Peter Pan are you talking about?

@rantsauce even though i haven't got tf working because it doesn't fucking build on fedora. for some reason. and i'm not using docker, and i'm not gonna switch distros since fedora works 99.9% of the time for me without error. and its more up to date than debian. and it defaults to gnome 3 unlike mint. etc.

anyway..
I'm going to use a lib to do machine learning. especially starting out and if my problem is a STANDARD one.
libsvm is actually matching my specific interest right this second.

anyway.

I was making fun of those commercials that play everytime they loop history around and the bots come out to play. in between my being exposed to weird ass people mistreating strippers and just being little bastards in general for some reason because a bad touch isn't enough unless you're an overexercised looking moron in a suit jeering them, who is likely the same kind of trash.

point is. the commerical makes ml specifcally as being a specialty or field needing people.

@rantsauce its like advertising to be a pentester.

@MadMadMadMrMim but ML is a field needing specially trained people. Anybody can run models, which is all that TF etc. will do for you. But to actually do useful stuff with it in production needs a lot more knowledge and training. Using libsvm to run a SVM doesn't make you a ML person, just like using a driver doesn't make you a kernel developer. But it does let you play around with it with zero knowledge of how stuff actually works, which is cool.

@RememberMe i get it, I'm just saying there doesn't seem to be a point in training a whole discipline of people to this instead of just training existing programmers, and to me it also seems like another hump. it will hit a breaking point in demand because of the limits of tech.

@RememberMe I want to start hearing 'we need more researchers and developers to fix what is presently broken'

@MadMadMadMrMim there is a point to training a whole new bunch of people for this. Machine learning/data science etc. is a field on its own. It just happens to use programming, which is its only link to development.

@RememberMe then add it to theory i suppose. point is they seem to be advertising it as another false hope. specifically they mention 500k jobs in the repeat ad.

that's nothing.

@MadMadMadMrMim I have no idea what you're talking about.

@RememberMe the whole point of me indicating there likely isnt a need for a rush of individuals to learn ML is for one its probably not really to a point of usability, which means it needs more research. that research will happen in certain specific high powered high income labs like google to specific purposes, the rest will happen at universities. there is not likely enough applications that are trustworthy enough to be usable presently hitting the average company, so advertising, which has been happening and has happened in the USA to fill the 'growing need for ML professionals', is not helping us. I'm learning because I'm curious and I have some books left to read because I"m an absolute begginer, but they are advertising 500k jobs hitting the market. Now we have 370 million people in the usa, if 5% of them are programmers out of work, thats 18 million or so people presently with more on the way into the field...

@RememberMe they'd be better off learning light and varied skills in various technologies or just studying other technologies than something which is still kind of iffy and has very limited usable applications because our country doesn't fund non specific research like it should. so using application frameworks like tensorflow for example is likely more useful if the purpose of learning is pragmatic, not personal interest.

@RememberMe thats just my take on the subject I was talking about. but I've seen 'the jobs of the future' being advertised most of my life.

@RememberMe to illustrate we still don't have any real self driving cars yet. and object detection on the roads has been a subject of interest for awhile.