44 Comments

  • KJ

    Spotify's music suggestions are the best algorithmic suggestions I've ever seen, I've found so much good music that I absolutely love from Spotify. I can't wait till there's technology for full on ai suggestions or even ai created music that's based entirely on what you'd like to hear.

  • Feakos

    I found out that ML scratches very different itches than web development. As a web dev it was very interesting to learn it, but I wouldn't like to make it my job.

  • Роман Кривцов

    It seems like only Spotify could manage to build a really useful ML recommendation service. Do you know any other such successful examples?

  • TikKu-UkKO

    have you tried JavaScript ML libraries such as https://github.com/BrainJS/brain.js
    I would love a video (serie) about this one. or similar, but js anyway..

  • Lxender

    It put me off that ML seems to be Python only. However, recently I found out that you can do it in Node with brain.js! I hope knowing that ml is available in this environment exists, helps you too

  • Chris Dillon

    It's a ton of math to really get to the core of it. There are a lot of libraries to help or "cheat". Sort of like studying algorithms (you could "cheat" and use a library). Spotify's recommendations are pretty bad for all-day radio. You start to see the cracks. But I don't disagree with your point. As far as career though, ML is one of the highest unemployment job categories (see Stackoverflow survey). It's just so specialized and hyped right now. But then it's a huge field so there are a ton of interesting sub-categories. Also, it employs ACTUAL SCIENCE so you can actually measure something whereas a lot of software engineering (ie: best way to do something) is more subjective.

  • shatley123

    Lately it feels like Youtube's machine learning as taken a poop on me. It's suggestions seem to be based off whatever the last two videos i've watched or a bunch of random videos I have no interest in.

  • Seieukos

    Do you think we'll come to a point where machine learning is a fundamental topic in computer science, just as data structures and algorithms are?

  • Seieukos

    For someone like me (mobile / web dev) with a bit of free time, what would be the more wise choice career wise: learning how to create machine learning algorithms, or learning how to use them?

  • Tom Conway

    Lately I have been having fun with micro controller programming. Playing around with an Arduino and random sensors really makes you look at the world differently. You start to notice just how much stuff is powered by micro controllers and sensors.

  • Jose Ayerdis

    Being a while since I did not watch fun fun functions today you seem a bit tired, are you ok? Development can sometime drain our energy sometimes is good to take some rest!!

  • med talel

    you don't need to know the math behind ML , it's so easy with tensorflow , you can even use a pre-existing modals , plug your data and you get a new modal , the hard thing here is how to prepare your data

  • Ernest G. Wilson II

    Can you please consider doing a real-world interaction including dynamically updating and interactive data with an API connecting to A-Frame? REF: https://aframe.io/

  • bitcoinTroy

    I would love to see you take on Machine Learning! The videos would be awesome. If you're serious, check out www.lambdaschool.com

  • J R

    Proper project management by experienced developers, along with proper training, team leading, and mentoring. These are things that are not exactly coding, but, go hand in hand in order to build a better platform for each generation to stand upon.

  • HardwareAddiction

    I said this two years to my colleagues. So obvious. This is also where non-academics may have an issue, because it involves real math and algorithm ability. I say that carefully though. I believe machine-learning will be abstracted in such way that is made approachable to pretty much anyone.

  • Kavi Ramyead

    The problem I have with spotify is that it serms to only care about metadata about a song like its so called "genra" or said "influence" but that is subjective, a lot of "rock" song are not rock to me, for my définition of rock, and anyway I might just not like it. I would much prefer something that goes deep into the Fourier analysis of the audio, tempo, beats, temperature of the song to provide similar sound with similar signature, something that is not metadata but thr actual data of what makes the song. I don't think this is implemented yet or if it is it quite naive then because I must admit I rarely find good recommendation on Spotify, I have to listen to a lot of playlist randomly to rind something and the system only seems to care about some info about the artist and some written classification on that particular song. It does not extract actual data from the song (or even a small portion of a song on which i can loop a few time, like a particular riff etc.) So there is a LOT of improvement in that field to be done and only machine could learn how to do that.

  • Joey Eng

    Check out the documentary http://doyoutrustthiscomputer.org/watch if you haven't already. It's only free to watch for the next 1-2 days.

  • Rory At Work

    I just started doing a bit of Machine Learning. I started with Microsoft Azure Machine Learning Studio. I found it quick to get into. So many examples and a lot of fun. Plus it created easy to use web services.

  • K T

    Did any of you guys ever use Limewire? It was P-2-P sharing service that had suggestions. It's from about 18 years ago but I discovered music I would not have otherwise from it.

  • asd qwe

    I listen to a lot of different genes. That is why discover weekly doesn't really work for me, the contrast is to big. The daily mixes are great though.

Leave a Reply

Your email address will not be published. Required fields are marked *