4 ms·
What is the real issue in your view? That ml-systems don’t work? In that case I strong disagree, I use multitudes of them every day (and you most likely do as w
by hallqv 6y ago
What is the real issue in your view? That ml-systems don’t work? In that case I strong disagree, I use multitudes of them every day (and you most likely do as well) and they work very well.
As someone working with building ML-systems that solve real-world problems I can say that distinction between known unkowns and unknowns unkowns is a very useful one.
- fock 6y agowhat kind of ML-systems are you using daily? (non-embedded ones) Probably stuff around search qualifies today, but this does actually not work for me more often than it does. Sometimes I'd just like the index...
- nerdponx 6y agoWhat constitutes "this" in your case? Machine learning "works" all over the world, all the time, in a wide variety of industries and problem domains. But in many cases it takes months of development and hard work by people with advanced degrees to get to that point. Despite what various AI startups want to sell you, usually you can't just upload a CSV of all your business data to some API and get something useful out of it.
- fock 6y agook, so you could you give an example on a day, where I'm only browsing wikipedia with my Linux-computer? I would tend to agree with you and while there certainly are applications, the relevancy is really blown out of proportion. Technically my naive-bayes spam filter is ML. Practically it's not as flashy as people think ML is - just imagine 15 years back and you'll see that your life could work quite well without using ML in the googly-sense on a daily basis.