i added another comment about my background in this field, what concerns me about ML music is the lack of extrapolative synthesis i foresee being implemented, as people focus on 'relatable product'.hth :)
I really like the new way you are writing text on the screen kind of color-coded, also the subtle bubbly sound effects when a text comes up on the screen.
Thanks for these videos. AI is not my main area of interest, but I regularly watch your clips to get a quick bird's eye view which is essential in my line of work.
Tester Mail he is not a gay u lil boy. If u consider his behavior as a sign of gayness then you haven't understood science. Siraj is one of the most inspiring teacher his love for science beams in his eyes.
i've made procedural music generators for 25 years. i've always focused on lyrical music :) i'm glad to see people finally interested in this via ML technologies. all i ought to say i guess is, when you get it to start singing, or it starts talking some other way (eg. noise parametrically fit to speech) you all in for one wile ride :) i think my most well known is breathcube, which you can find videos of, its very simple. there's a track on soundcloud from earliest work in '94, 'care for that i should only have your houses' - sounds and sequences determined by program, then hand assembled.. sounds terrible because i had a mirage sampler with phonemes (3 keybanks for one take) and shifted pitch using a pitch shifter, very hard to understand without text..
(but, for whatever it is worth mentioning, my work in procedural music was for memetic deculturalisation/depatterning.. procedural semantics basically makes you more intelligent but less intelligible to others, which is understandable really)
We recently conducted an experimental project to generate music with extensive kinds of advanced models. Feel welcome to browse our final report at www.deepsymphony.com.
Again a great video. You rock! Been wanting to start in music with AI, but thought, I'd have to write it myself.. This is amazing. So many programs.. We've had band in a box for a while, but with a neural net will be a LOT more interesting..
They went to make "AI music"...indeed they did ;) Okay seriously interesting video! Can you make a deep learning video about its uses in cfd/fem? I am slowly getting interested in deep learning but I just love fluid/thermodynamics since I have a B.Eng in Aerospace.
Siraj I think that AI music is the most exciting subject in AI.. There are a lot of ways to research sice genre recog till music generation... Tks for the video.
Fabulous! Today, through the magic of current technology, no-talents can make bad music which no one cares about, which is even worse than traditional bad music which no one cares about? That's brilliant! That's progress!
we've been friends for a while, met at The Next Web conference in Amsterdam, she invited me to this United Nations event in NYC during general assembly. Just friends
Taryn's album is really exciting! We're building an artificial intelligence powered songwriting assistant for those in need of inspiration for melodies. Asking for signups for early beta testers here 👉 bit.ly/2ysHdCH
By the way, engineers do no create the interface but designers do. This is like saying sound engineer does the composing. A very common mistake made...
Why are there so many comments about Taryn being Siraj's girlfriend? I mean he is already taken. (He has that whole trust-video ^^) Well I guess it is better then the "She's so hot, I would do with her" comments that you get on some other videos.
I know I'm going to get a bunch of hate for this, but... As a musician, this just breaks my heart. Nobody is willing to learn to do anything challenging anymore, and now we're taking art away from ourselves, because it's easier to have a program (which someone else created) "mimic" music (which someone else created). The techie part of me thinks it's really cool that AI can create art, but the more philosophical approach suggests that nobody would actually want to live in a world all the things which separated us from the animals in the first place, are done for us by AI.
Yea, we better embrace our robot overlords. Soon there won't even be human artists they will be entirely ai. Same with everything else. Then we will all feel even more redundant than we already do.
I'm not completely sure what you mean by "different", but I wasn't talking about just composition. We know what happens when people who don't play instruments compose music - it's called EDM, and in terms of composition, it's very primitive. What I'm saying specifically is - With current tech, it's too easy to sit in your bedroom and "make beats". Why spend the 10,000 hours and get good at playing an actual physical instrument? What I'm saying more generally is - seems like people don't want to do anything challenging anymore, unless there's instant gratification. Aside from just being an embarrassing trait, it's also likely the very door that AI will eventually roll through.
You are right, more and more crafts will become optional if we develop strong AI. Just like exercise used to be integrated into our lifes when humans lived in the wilderness, and now exercise is optional (and a health risk if you don't do it), the same will probably happen with "mental exercise". People who don't put careful dedication into training their minds may find themselves unhealthy mentally. It is worrying.
Hi Siraj thanks for the video, so I have a couples of problem I would like to ask you. I am struggling to learn AI using your RU-vid videos as well as working on challenge at the end. but one thing always confuse me, is to know which video to go to since all are different and interesting. as now I am watching Learn Python for Data Science where I am on episode three and planning to enroll your Math of Intelligence Video and I don't know if it is the next step. right now I don't have money to pay at Udacity where you teach deep learning very good so I use your Channel. so I would like to know the best way I can learn efficiently by using your video(with challenge at the end). #is it possible, I give my self 7 month to learn up and down at least to have the knowledge about AI, I have experience in Math, so only the topic I don't understand well is probability. I am third year students.#Tip here in Africa most students, people don't know the power of AI. we need skills to prove them, that's why I choose your channel.Thanks Siraj
Hey Niyongabo Eric you can learn AI from many source Not only this channel..No doubt this channel is good , but there are other channels that are better in teaching AI and coding stuff . One who want to get really deep into Machine Learning and Deep learning should read blog as many... In AI not only coding matters , there are other parameters that is so helpfull to learn AI and help to understand what is really underneath in AI
to be honest Siraj's channel isn't good for learning if you're a beginner , Watch the course i recommend after that go through Abu Mostefa videos ( available on RU-vid ). In my opinion deep learning book is a great book as it covers basic required for understanding ML and DDNs.
Taryan we are in the same path and edge of NOW and that edge has an amazing view of the possible futures. Need Taryn for a new team into the sea, no, the universe of AI entertainment. please check your email from your web site, and IN invite
hi i have a question. 1. what is the difference between tflearn and keras (i heard that both are same) 2. what is prettytensor and why/when we have to use that
All these are wrappers makes it easier to build and work with TF. Don't use pretty, its old and abandon. Both TFlearn and keras are good. TFlearn is 'closer' and more similar to TF than Keras, but Keras is more used and it can be easier to find help on forums. I use Keras, lots of good features like multi-gpu and visualization.
Depends on your goal. If you want to do more complex things you should have an Nvidia GPU (most frameworks use Cuda). If you want to learn how machine learning works and start with simple stuff, every CPU is suitable. Even with an RPi you can get started.
PC-Gaming desktop hardware with Nvidia card is quite good for beginners, even if it looks unprofessional, good Linux compatibility is a big plus - if you are familiar with cloud services they are maybe a good alternative, but if not you get possibly confused with yet another technology. Usually for the most beginner tutorials you don't have to care about hardware at all focus on you learning progress and don't care about hardware until things getting slow.
