Merken

Closing Keynote by Lamere

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
the and the man and the ball up this is great so and this is my very very 1st reals Conf but in a full disclosure I have never ever shipped aligner reals code in my life I am a total outsider but I came to this conference it came to the whole thing and would lots of really really great toxic to meet lots of really really great people there is really a really welcoming community I'm really glad I came and I've learned as a few keywords and a study on like something like Active Record and something Turbo at about something and
then apparently you folks are into this called test in which I have never had to do so so I I'm polymer yeah I work at Spotify music streaming service I mean I'm really excited about the stock because I get to talk about my 2 very very very things 1 is music and 1 is code and so let's talk a bit about music so when when I was in high school I was in a band it was the memorial high
school marching band they played trombone and I really loved about being a musician I love getting from people I love to uh you know share of a performance I love to maybe even be of the touch people emotionally with no power or or make him feel may be sad because the set trombone or something but I sort of realized 1st of all I was not really that good I could not make it a career as a as a trombonist so I moved on to become a programmer on this is back in the early eighties this is well look like to write code in 1984 but it's still a scope that me wearing a tie and you notice and I'm a tie actually we look a lot slimmer and the so I I spent 20 years so writing all sort of software in lots of different industries including finally enough of that industry that make this is talking about the wrote software that controls the cameras that flying in the U 2 and they're actually still flying to this day which to my shock and horror and but I'm about 15 years ago I got to combine the streams might might to very very favorite things cold and music I started to work and and that in the music tech industry and eventually that led
me to our working at Spotify and so now I mean this total heaven because I'm surrounded by all of this these are some people who are building is awesome products but have me even better and surrounded by all this great music data that lets us all learn about how people actually experience music I'm so stock is called data mining music but that's too so you can go back to your bosses and say yeah we we learned about data mining it's really are hacking on music is that's what I do I even like to explore music through code on all sorts of different ways I like to find out of how people experience music but I like to think about how people can learn about music I like to meet the music experience more interesting more interactive and all sorts of things like that now my real job at Spotify is to spend every day looking through all this data and try trying to improve the music listening experience for our listeners so that maybe things like helping them final a bit more about what the music itself or help the organizing music collection of given good music recommendations and so in this talk i'm gonna show of 14 examples of really kind hacks but I column experiments because it's better than acts and that show different ways that we can add the learned stuff about music so
1 things about Spotify's were surrounded by data and the data that we get it Spotify about how people listen to the music gives us a view of the music listening that we've really never had before I'm just gonna die right into a quick example to give you a taste of this and this is called the Arrowsmith anomaly so you guys I'm sure all familiar with Aerosmith they're the bad boys from Boston they had their 1st hit way back in 1975 but I'm sure most of you are most familiar with their and do only number 1 hit which happened in 1998 that was it's a song of of linking on the name somebody does know what is this is what it sounds like a a my head might have had for a spontaneous applause and the receiver top so that and some of you may have actually that that we have been the URI for slow dance and middle tuple and to the new trigger warning of I was so for 1 of things because we know exactly when people play I play the song Spotify we can look at a time series of the
songs so here's of the attractive of streams as a function of time for for don't wanna miss a thing over about a 2 year period of from January of 2013 to December 2014 so you notice a few things about this plot 1 is you know this is general slope up to the right that's not because the Sun's getting more popular because Spotify's getting more popular but you also see a lot of these little bumps in that's the weekly listening patterns during the week people listen more when they're at work in the listing that little bit less during which that's the pattern pretty common across all songs but the really interesting thing and this is the Aerosmith anomaly are these peaks what is causing these peaks we have our and so we say something said that the volume of listening is or tripling in in what looks like a single day on this actually is pretty where we don't see this across you know the song so what is causing this so we can start out by taking a look at the dates are right so I think there were 44 every 4 tell you thinking middle school dance rights but then what about this August 7 7 September 18th in November 13th well it turns out of these are all time because the spikes it here often same reason and it's not because of Valentine's Day on its for another reason in itself uh actually it's an on World War be reasons so take a look at this for 1st one close to with asteroids 2012 potentially hazard add asteroid this year it was at a spacecraft orbits common 67 the the way winning size selected and uncommon 1617 robotic Linda touches down 67 me they were deemed to that's and so on so yeah you guys of totally figure this out right the Sun was on the soundtrack movie Armageddon were Bruce Willis goes and saves the world from the from a common it's about it here so what we find out is this song is sort of the poster child this is the go to Sun when there's some kind of asteroid me you're coming our way so this is the kind of thing we never ever would have found
out if we looked at you know that the old traditional charter his billboard and and before the the differ music revolution but people go to billboard and build were actually call record stores and call radio stations to find out how often people playing science and so on we look at how much more data how fine of you we have the data it's it's kind of akin to with
Galileo misguided akin to in Galileo took a telescope and at the sky the 1st time you saw the moons of Jupiter or the rings of Saturn but you know it's maybe not as the heads and said to be global impact this but it's the same sort of thing certainly a big impact in the music industry of the data that we we see here and is just the opening up a eyes they're all sort of way to how people listening to music and but before you leave this example is 1 more thing I know you guys know that correlation does not necessarily imply causation so we go back to that that chart with your Smith thing just sort of think maybe 1 of the the correlation things actually reversed and increased Aerosmith listening is actually attracting I asteroids really or to think about it and really right so why
do we care why do we care that we have a lot of music data that lets us find strange things about music well
back to when I was 1st writing that code this this is our an exciting new music technology and the a Sony Walkman you could put 10 songs in your pocket really big deal the biggest challenges that to trying to get that the fit in your pocket and 20 years later Steve Jobs get up on stage and introduce the 1st iPod you about a thousand songs in your pocket the exciting technology that will with this was shuffle playing you pitch of a few thousand songs get a pretty good listening experience on and then 10 years later Spotify lies in the US lots of other streaming services did as well as the subsequent 30 million songs in your pocket so your succinctly tap away from listening to just about any song that has ever been recorded except for Taylor Swift and it to and and so on and we're going to need help figure out how to listen to music and we have 30 million songs you can hit the shuffle play but included and get some John Philip Sousa March and some Gregorian chant and some Katia any day iPod whiplash right so that we need tools to help us so we have 30 male million songs in your pocket water we didn't do so this is what I do it's modified I try to figure out how to improve the listening experience so this talk and I know so what 14 Experiments around how we can better engage the listeners and
every were dog love is all about data so I'm going to focus on 4 different kinds of music data of music metadata so this is the basic facts about music cultural data so this is what people are saying online about physics there's listed data so this is uh who's playing what and when and finally acoustic data so this is treating the music as data itself so what does the music actually sound so we start off with music metadata sometimes called the most boring newest of the music data and so this is really the basic facts are what your Spotify we have millions of tracks are artists and albums and there are all interlinked artists have albums albums have tracks track can appear in multiple albums so there's a crazy intertwining of all these things in there and have lots and lots of different fax associated with them and it turns out that this is really really hard because of the music space it's not really will contain artists can call things whatever they want they can do whatever they want and nobody can stop itself uh turns out that we spend a whole lot of time trying to get this data right and it's going to give an idea of some of the challenges here know 1st of all you get a
simple music query somebody says hey what are the songs on the album White Christmas by Bing Crosby the reason that this is hard is actually being crossing has 40 albums called white Christmas I and they all have different of a lot they have different tracks they have different performers so he had the call hold like christmas and you kept make intriguing milk at every year with with the new album and so and Sarah had told me that there's a really highly technical crowd and I should try challenge you as much as I can so I have a little math quiz here
so why is this formula troublesome for music recommendation and discovery to out the answer if you know half of Fort Hare her right right now write the answer is that this is the name of a song by a fixed wing FIL so if you're building a music service and you want to make sure that your listeners can find the son would you do with it's not easy so just and you can kind of look at artist names right back before we had global we had been called that that would be the name of stop words and I know you guys are all familiar with the band Duran Duran you did you know there's a band called Duran Duran Duran it is a DJ who calls himself DJ Donna Summer and there's glass teeth in fact there's a whole genre called which house which is is filled with a new unpronounceable names and then the final artisans as an artist there as far as I can see can the concern the burn in Hell forever is the artist various models of how to the there really is a DJ whose name is Various artists and so say you're a fan
of the band eclipse did do not exactly 22 bands Pollack that's now be having to be the the of the Utah vocal choral group in I happen to play through the Ukraine brutal death blank or personal both of us about have a bad day right so a music metadata it's
hard but it and but we can still try to have a little bit of fun with it so of 4 or we want to do for our 1st experiment we discovered cannot hole in the water a little bit of data is to enter a very important question from the Internet's in
questionnaires have been names getting longer is the real question that's been posted on Quora it so helpfully this guy named Zachary Davidson you start off by giving us his qualifications which are I Name to both my bands in e says he would say yes they are getting longer but only very gradually goes on and on and on to the justifying his things but of course that's his opinion we're just going to do this with data
and so it's pretty easy to do and it's the only thing we have to do is go through the top 500 artist for a five-year window of about calculate the average names and redundant the code is that shorter than Zack's answer and so have been a forgetting getting longer little audience participation ranging hand if you think artists names no longer than they were a long time ago OK you rare Asiatic even hands up if you end up your on the artist name for actually longest in 1955 to 1959 is a always going on then as there are quite a bit longer but these are the
kinds of band names we had back then McAulay in the Soul City Symphony Orchestra Academy of St. Martin this so essentially the the the the mean at the time was orchestra leader name + orchestra name so we end up with very long names the course while we're at it we might as well take a look at what the and with the longest artist names are ever of that were popular and the longest is the difference in the tendency and with tendency and I think tendency on both grew up with short names in the compensated the the alright so that's a quick tour through metadata nowadays to to cultural data and you probably may know who less about culture beta because no 1 else on whatever anointing the writer culturel data so so the idea of cultural data is there's a lot of data out there on the web where people writing about art music so there's music blogs there's of music review sites there's people were pretty play together and there's lots of lots of really interesting data so the idea is that go out crawl this data server Google skill crawl collect all up to some natural language processing on it statistics on it and use that data to give us a really good description of a particular artist so here is an example so we have a set of the black middle banding board year the a typical review that you'd see for b an insidious actually lots of descriptive words for the for this artists of black metal melodic platinum Middleton a unique symphonic Norwegian fast famous and so we would expect to collect all these words up we get a pretty good idea of what this band is is about this is just from the words from the 1st paragraph of 1 review so if you repeat this over thousands and thousands of reviews have been written about this be and collect up all the words find wanted a distinctive for the band from the ones that appear a lot we get a really good description of what this band is and we can use this data for all sort of things and 1 of the things that we use of 4 is for of artist
similarity so we can find out how similar 2 artists are so we Katy Perry we have the word year we can look at the top terms and here only showing a dozen terms but imagine that these lists go on and on the we didn't also degrees ways and there's very little overlap essentially the decades that that that they played on words we look at at a different artists like say Britney Spears up quite a bit of overlap with with the period so we can sort of assume that Britney Spears and Katy Perry are out right so now we're going to do an experiment with this artist similarity in it's our experiment is to expose the listener to new music by using this artist similarity but this is the extreme Edition in its extreme because our
goal is to help the Katy Perry thing and listen to the reward here I write how are we gonna do this right so we start off we have these 2 artists
we know that they're very very uh dissimilar but maybe we can find artist that that fits in between on that has some overlap with both of them right so here we have about maybe 3 song playlists that would get us there but that's a pretty big jump from the period of innocence instead is also a very big jump from evanescent Sabine word here so we have to go a little further so in fact what we do is we can take all the artist a big graph and connect them all up to the nearest neighbors is a visualization that were done by researchers at Yahoo by the way it's pretty nice and so when we see we carry Perry's at 1 end of of the space toward years at the other the evanescent is mostly in the mail what we wanna do then is to sort of fill in the gaps following a friend died stratifying to pass through the space and we have a playlist South and so there's an
app that does this is called well the fraud I write you guys we that you know the story about boiling affronty part of fraud in cold water and heated up I am very slowly or noticeable jump out the idea is to do this musically so take somebody's listening music very gradually taken from 1 artist to another so I ended by run is generated play and will listen to the very recent that's of the 10 songs and the idea is there should be a gradual print progression of between dust songs here go
that is very to here while the friday replace of 10 so let's listen to know about your help it comes along the time and the same thing in a but in the thing and I I I I I have been a this mile on that and then you know you and but you and I think I might time and I know that you have and I and thank you so you don't have that in a and I I'm going to and alright so that's where the from right that a quick tour through
cultural data and there were going into listener data so that the data this is a 0 every time you pressed Play and Spotify we get a little MDIS messages as a uh you play the song of this case a some of the state and so we get of billions and billions of of this data every day is really sort of the bread and butter of Spotify we drive derive all sorts of things from the data this data from charts and music recommendation collaborative filtering and stuff like that but and try to talk about something that's maybe a little future of looking in on is something I'm really interested in that finding
artists that have really passionate fans so the others this you know certain artists or any Hamilton fans here by any chance right so we Hamilton vandalism Alamo over and over and over and over again in my right yes right and so there and and very very passionate I think finding artists have passionately is really interesting bridge helping people find music and you can steer them to a past uh artist that they become really passionate about I think it's a big win so what we wanted to see if we can find artists that have the most passionate fans and will sort a bucket this and say is it a which which type which genre has the most passionate fans adopt steps or of metal that's right so how are we going to find out how of
how about how passion of the fans are for particular artist there's probably lots of ways to do it but 1 really simple way is to just look at the so the average number of plays profane and of for particular artist is Oregon sorta illustrate this is numbers of fake it or don't disparagingly artist here so here's to artists with a million plays of a Robin Thicke on the left he has 200 thousand fans which means every fan is playing his his son 5 times of blurred lines made into the play this they play it 5 times in as it where the other side we have Lord and as she has 10 thousand fans are each of those the interplaying her son's 100 times so I those is a much more engaged in lots of she has these high and the and the same number of plays different Murphy and so we can say about which artist has highest passions and so it is for a lot a lot of
artists is a plot here the x axis is the number of fans y axis number of plays we can sort of see you know most are as follow the same general character but we have some artists that are have really high number of average plays for listeners in their Samara have well so to answer a question which artists have the most passionate fans we can just look for for the artists that are in the the top
here out that so here's a question who
has the most fashion of his regime and if you take but it's also the choices are dub step for metal radiated think it's metal have gone into this loop to you know right yeah right right pretty good number right right so who has the most passionate so that artist with the
most pattern of into the band called in flames 115 plays Gryffindor right now there is the case of the entering class I and goeswith Stanton quality when I'm sorry the the and I was rooted you'd see why that that events so if we look at the top 20 artists 9 out of the top 20 our our middle bins and 0 r dub step so in the middle here the heads are the most passionate and according to this 1 small experiment right so who has the least suppression advances 1 being that stands out there have only been violently and the man and the way it so that they do so this 1 hit wonder band they have a son that's a thousand thousand thousands of players ever listen to Wykle guys like me with a running and and but then you never listened anything else so they have a pretty low up place for fans right so that's a quick tour the that the passion index alright
so playlists no at Spotify unless in years people of created tool billion playlists as a huge amount of data of all from sorry the I is over the question here is what can we learn from from 2 billion playlists and and so 1 thing to do is just look at the most frequent names of players so we'd look through all these 2 billion playlist find the most frequently occurring names and these are the top 10 so we have wrapped she'll Country Party house out rock Jim music trip to stratify people make places called music of lots lots of people that play this time but you know it's here you know 5 5 of them are are John related with fiber not gender related their their context so people actually making playlists around what they're doing as opposed to the type of music that in the way so if you look at the top 100 playlists 17 of the of the top 100 names of playlists our John related and 41 our context-related so this leads me to believe context is really the news genre so some people are organizing a listening around 0 and 1 for runner having dinner party or I need to get some focus music so just to give you an idea of some of the names that people use in integrating players is a war of words with the bucket bucket it into different sections like training work out a lots of different types of play this we have moved all relax motivation for 20 yeah travel road 1 flight commutes along here we have romanticism exuberance of we have time for 20 and focus for 22 because there and right the active starting to rise and those guys there from we now that appear here I we had this great idea how will sort of joint effort Spotify read mapper plea this thing no socializing their party and right so lots of different ways people are organizing their music the contact is really really important but the context is really really hard because you are music does not come predi labeled with the context that is good to listen to so I challenge here is so let's find music for a particular context by mining tracks from these 2 billion existing playlists and so well you know to put a very specifically even create a place places the mainstream tracks the good for running for a 55 + year-old male like like me and I how we don't do this while he is
apt but I will just look at what we do and we have 2 billion players we start we just start by searching for other players have running in the title and we may get about 10 thousand of those playlists but we aggregated all the tracks from them so all the tracks occur in those 10 thousand this summer when it occur a lot more than others so we aggregate them up so that the tracks that appear the most most places go the top we sort of just for popularity so that distinctive running tracks uh from the top there we can filter these by demographics so that tracks the more likely this into by 55 you old get row re-rank higher this leads us to a playlist so here's of the 55 year old male running playlist and so we can use sounds like but my and my friend I so that that will be rocky came out 1975 so the 55 rules what's that when they're 16 years old and still there they go to running song and so to be doing is nice demographic filtering we can generate the same places but target for an 18 to 24 year old female only in a much different set of tracks of knew where they sound like that he was the link here you have a high and a right so you notice that the temple is somewhat higher for the know 42 point 4 year old and as we have a few others we have rotor practice and a lot of words of a young woman with was not a so succeeds there's a lot of time in a in a more so I sort of got that myself and corner here could I have 3 18 to 24 year old daughter and so I'm just gonna play the 1st on for the next this you can read the titles I don't even wanna know hand on the they can make their own decisions the alright so after 60 times break up time you know as the and it's all about the crying uh in other words it in the 24 year-old it's all about the view when I went hand hand hand hand behind the behind but so quick tour lighting up playlists of funding is a were rolling this kind of thing I our Spotify you you really get good so recommendations based on things like that of most of my experiment number 5 perhaps 1 of
my favorite experiments of all this is using studying data to find the best parts of the song so people is a music and
they oftentimes not only just press play but they actually scrub inside the on bill of movies and play the best part are skip the worst part of the sun so what happens if you aggregators data across millions of millions of people were listening to the same sounds well you start to get a map of what are the hot spots in the and the best part of the song so here the I should blow this
up here the most common of places that people describing to In this song in the air tonight by Phil Collins now and people act you don't scrub to their their favor part to use gravel before that so freezer integrate the actual listening part we get another curve that shows us the most listened to parts of this song as the result of gravity so you guys probably figure out what is going to sound like let's that's listening where we start at the most to spot listen through the act the most listen to spot and new divide up this little bit on this example will bring in a in a scene on uh the the world with a little more in the future it might so so and so on and so on a lot of time to do the experiment again some interesting upper this these results as parties look-alike adopts that song adopts the song like this it that take that very rare shopping it's the dropped some kind of a of the value of that so it's related to do a
really awesome dropped it which is
with the world needs is 1 so it's not all that great you just listen to it to this day but you might have lost my the right up I always of ignored in and the I'm Bloom had been there is that the the user is sort of that we that we the peoples scrub the executive and if all available at the time
rights as free atom that the more so because we can look at the shape of this these peaks some peaks are more prominent than others so we can say well find me the most prominent peaks across a whole genre so we can actually find what is the most interesting wrapped in all of hip-hop and so
we have as little interface there we a search based on prominence in and find the most prominent ascribed to spot so our free part this is the most prominent peak and of your family with a song but some studies that would you need to manageable supersonic speed influence the development of the novel and unintelligible innovations that have a little bit of evidence that the original incompatible with the example of that is the of
no they're not at all right so what you think is the biggest drop in when both quote from the 19th
century and the amount of work that we we did the drop was with him
in the basis on classic rock it's all
about the guitar solo and hence the name and so what I think
is the greatest moment classic rock take a moment to think about it may be controversial so here comes but a
in a in a in my greatest given rocky notices
is right all of the i in the but so Rolling
Stones I was wondering about where closer look at the song I have in my mind when I thought would be the best parts of my favorite part of the song ways said that probably was different than everybody else is not the case if they were part of the
shelters when the backup singers voice cracks it's amazing that is is that little red does right at that moment my hand in my
so everyone loves you think of the
4 part music it's all of the words of
have and you have the the the uh right so
that's what of what we care well also 2 reasons and I'm just looking forward to the day where we have play we just drop but no and modified and coming soon
right now renamed into the last bit of data this is the acoustic data this is treating all the music like data and so how we do this so this is a we do all sorts of things we take Audio Wikinews signal processing on the audio to machine learning around this and we can extract all sorts of things like basic things like tempo key mode we can figure how danceable son user how intergenic it is or how loud it is we can also get a really good map of how I have where all the beats are in the and the bars in the data and and things like this there we can use this to do all sorts of things and all the next experiments economy driven off of that but but I would take a
little step back just started think before we had recorded music how people use to listen to music you're always in the same room with musician and this means that the listening session was much more interactive you could have eye contact with the musician you could I shall play the course again or you could sit down and join them depending on the venue and so on musical singers really really interactive now compare that to today how we listen
to music is no interactive tivity at all except for a Parzen plainly enabling the discovery and so we want to see if we can use some of this data to help make use of a single bit more interactive and some these experiments will be around there some the right so first one is trying to say here is a human or a machine setting the temple for your favorite song so how recently dramas have been putting on headphones and listened neglect tracks and they said this at the the the beat on it
so what we can do is is sort of figure out it if the ban you listening to is actually using a click track or not this is a little app called answer to the click track what this is showing is a variation in temple over the course of the song you can sort of see of course osama we get some natural variations for peaks this is because the the the band speeding up and slowing slowing down so the and and that's not necessarily because Stewart Copeland the drummer for the police the police on our is about drama he is actually using the changes in tempo to add tension to the song and releasing in over so let's take a little less to here where what that sounds like could you pick
and if it is there this that act on so in a way that I was a little about the role of the a and so we can do little man in
the in the years between natural variation in double so this is the a human drama of those with another song so I signed libraries various you can it's perfectly flat is with sounds like a good have to haul all my a so notable various and probably a machine of drumming and singing in I right the I'm going to disparities in artisans as the fungal right so finding a lot dramatic moments in music so we look at hot spots finding interesting moments where I will be I'm interested also in finding dramatic moment I really love dramatic music music that goes from real quiet to real loud in 30 seconds or so and so and that's
what this is it's got it's an app called where's the drama and the idea is is you only 1 listen to the dramatic parts to that the best parse song so you'll find the most dramatic it's just by looking at that the loudness contour highlighted in you can hit the button say I'll play you need a drummer here's what it sounds like the variance
of the a and B of a of a of a of the game and a and a lot of my mind
but your the so by the way that's on given instances that bends the our our spec guys there at the end a material lots of innocence examples right and select when this is as something of a bill that music active sitting right next assumed by the way and this is my attempt at doing something creative sodium coding but I think back to the time when I was a musician and I could you know to somebody's a most emotionally so this is my attempt at doing it so that we know where all the buyers and the siren where the Sun gets exciting and where it falls off and well the phrases are we can do a very up much better visualization echo accompanies the music much better than sort the VU meter kind of visualization you may have seen and when the upper or something like that so this an aperture in 3 GS runs in the browser and it's really just an accompaniment for a song by Ellie Goulding that is
so small excerpt but the
word problem in our are over time and in the long run we humans around the globe and we have a minus you my a in a in a in England and I know that all of the time I have I domain dominating mind
room in things
so and so I after I a build this had imposed on the web I realized and that I was a 55 year old man which is created to merely gold in the in part and so I figured in billion for pound so I found my way to an Ellie Goulding the an art site which these a quite a few of them by found my that was the biggest 1 crafted a form post and post the link to their and nobody liked it's a right so next automatically remixing music so you know a scientist or a list of beating buyers and stuff so whether you could manipulate these lists of beaten buys it's like you manipulate a new list your programming languages so that's where this remix as a library as whole a Python code and so he it it's it's goes and so on and so itself and so here's an example of doing some using our remix library there's a Python jobs bindings for manipulating music the algorithmic so the here's of 6 lines a program it's going to take a song that really is by Lady Gaga the following the 1st being every bar in out into a new song and for those who can read Python will do little visual for you that the 1st 1 of the so what was the cell like don't put my by around the last time I in time and you have to have been a some surprising when if you arrived did 3 lines code to put those beats in reverse order some like this all look like a society prior to the 1st part of the theme integrated geons any questions and I would mind I so I this theory that you can but lady gaga in any order and as a matter of a lot and so I have some have of my youngest daughter she was listening to prepare the stock conceited dead dead that you likely that the like some of the stuff we mainly the big ones the the jokes are the next experiment now so we know other bitterness on well let's see if we can improve a song by changing the drummer and that's the idea behind the Bonham so just right so imagine the idea here is that in a song could be improved and John Bonham work was the drummer so we know all the the tires on we have some outtakes from John Bonham so we just need to align those beats and let a played some tricks and because nervous and temple and the real trick is to make sure that the drum stop at the right time and start at the right time otherwise it sounds just like a machine so the idea here is that by here's a before before no 1 ones so as I can say that there are some things that I would have from right now form the
moment I say and that the the the the the the the the the the the the the the the kind of thing and so on and so on and I can also have an old and that don't know and was
next that hackers got so here this is the only hacker eigen right user might just under high on the loser I want the founders of the Afghanis which is the company that bought by Spotify environments Spotify and so the idea is to make your favorite song swaying so swinging a song is you know where to go and don't don't don't don't bother about about so you just need to stretch the 1st part of the beating shrink the life that you get a nice long song so policy with this sounds like here's unswung
play a right here this long version
the mind and so on and so on and a lot of time and so on it's awesome all right so this next 1
is called the boxes maybe my most well known happy will hold whichever of the infant about and by the so the idea is of the infinite boxes for when your favorite song is not long enough so you'll have a assigned which would go on forever and that's with infinite to box doesn't it does this we know where all the b terms on we know with the beat sound like so we can find actually sound very very similar so the claim that the song insists that instead of just playing back in sequential order we can jump to us similar sounding we do this a little bit randomly so we end up with a song a place for ever but it's always changing so it's not just a symbol look so here are on the outside of the circus circle are all the beaches colored by the timbres and then the it's it's sound most similar connected up and so as I played through the song you'll see some and sometimes it will flashing green on those are 2 that's when we're we're jumping to a different part of the song so here we go but you know so we now have to you have the means of a sense of the noun and a kind of hand in a way that you and I might have a hard time with that of the and I found that the median in the trip so he so we just get 2 more to go and so this next 1 and I am particularly proud of this happen because and it makes me actually feel like a created something beautiful which is something
very rarely get to do as a as a programmer and this is called the auto Canonizer the idea the idle canonises to turn in any song while maybe in on way a lot of songs into a can do not a again uses a song like Row row row your boat where you can play it at the sun against an offset copy of itself and it sounds pleasant so the other Canonizer does the plays a song state through but then find other parts of the song that would mesh well with the with the 1st voice and tries to generate a pleasing version of the song so in this example would only playing the song all the rainbow by the wine singers is an and and so this is not a duet but you'll hear him singing it as a duet so we can make this deceased singers singing a duet with himself so here's what sounds like and the the the the the the world of all of all of the people of the the of all of the of a and engage a child and and all that you do all of this thing and so on and so on and play the issues and and all the more so than in all and then to this is the rule of law in relation to the the the and so you can so that brings us to the final
experiment is called a girl talk in a box and the idea here is to turn a song into a musical instrument girl is a remix artist and and so on the ideas we can take a song we can break it down into its beats and then you can interact with it all runs in the about browser now I'm going to attempt a live performance for you of all looking and that was just sitting in the back ground measurements advices written some
tests that I right and note that so there's uh some fancy featuring or whatever so all these are the Beatles song and I can interact with this in all sorts of ways of 1st like sort of play the song and dance band Tennessee of the colors in the that the song then backwards don't about avoid every other beat there is a lack of whole world meaning under agenda that are only guardians of anything anything anything anything the and on the right in dealing with people in different environments that the like and time information that in and and so on the 1 that is being done in the beginning of the so that's words the another try to play as a song there we go this list of practice quite a bit and I'm doing is really because I've never performed in front a thousand people before even when I was playing trombone at my name quite nervous more nervous about this than any other part of the day you know screwed up right stereo and to do don't don't don't there are things that are in the the now in the in the in the data and danger gain and even in the in learning and maintained by the International Union and is engaged in the the the in the in the
in the in the in the in the in the in the in in the in the in the
in the in the state it's that it's that it's that it's that it's that it's that it's that the of the of the of the of the of the visible and the model to do and I think that the sum of the of the of the of the the the the the the the the way in which the set of DDT DDD DDT and it's that it's that they don't like the the the the the thing I you uh
so uh so we all got 30 million Sergeant pocket now what right so all this data and that we've been collecting and using it we're really gonna use that to improve the listening experience but it just it generalizes for just a moment and so I'm really blessed to be every work with the things that I love most about music code but of your working in some domain and I really strongly encourage you to have on your data to get to know your data love your data play with it do experiments with it Bill crazy stuff on some other stuff will just be crazy but some of it actually becomes really really useful so beloved data have to be so thank you very much all of these experiments are online and you can go and try them out so you really great audience and my Special applies for the for you know the and I were you know if
Beobachtungsstudie
Videospiel
Datensatz
Reelle Zahl
Vorlesung/Konferenz
Turbo-Code
Code
Metropolitan area network
Softwaretest
Maschinenschreiben
Bit
Subtraktion
Programmiergerät
Gemeinsamer Speicher
Code
Quick-Sort
Streaming <Kommunikationstechnik>
Dienst <Informatik>
Lesezeichen <Internet>
Menge
Software
Gruppe <Mathematik>
Vorlesung/Konferenz
Bit
Subtraktion
Sichtenkonzept
Total <Mathematik>
Güte der Anpassung
n-Tupel
Zahlenbereich
Biprodukt
Quick-Sort
Code
Data Mining
Zeitreihenanalyse
Prozess <Informatik>
Reelle Zahl
Benutzerschnittstellenverwaltungssystem
Gruppe <Mathematik>
Vorlesung/Konferenz
Hacker
Schreib-Lese-Kopf
Maschinenschreiben
Lineares Funktional
Bit
Systemaufruf
Abgeschlossene Menge
Hasard <Digitaltechnik>
Plot <Graphische Darstellung>
Mailing-Liste
Frequenz
Quick-Sort
Roboter
Dreizehn
Streaming <Kommunikationstechnik>
Datensatz
Rechter Winkel
Gruppe <Mathematik>
Rotationsfläche
Arbeitsplatzcomputer
Mustersprache
Vorlesung/Konferenz
Spezifisches Volumen
Speicher <Informatik>
Figurierte Zahl
Unterring
Physikalischer Effekt
Gruppe <Mathematik>
Vorlesung/Konferenz
Korrelationsfunktion
Quick-Sort
Schreib-Lese-Kopf
Sichtbarkeitsverfahren
Subtraktion
Wasserdampftafel
Physikalismus
Code
Raum-Zeit
Metadaten
Dienst <Informatik>
Weg <Topologie>
Rechter Winkel
Gruppe <Mathematik>
Vorlesung/Konferenz
Äquivariante Abbildung
Hilfesystem
Fitnessfunktion
Mathematisierung
Gebäude <Mathematik>
Abfrage
Systemaufruf
Sampler <Musikinstrument>
Ausdruck <Logik>
Informationsmodellierung
Weg <Topologie>
Dienst <Informatik>
Fächer <Mathematik>
Rechter Winkel
Gruppe <Mathematik>
Vorlesung/Konferenz
Wort <Informatik>
Metadaten
Bit
Gruppentheorie
Wasserdampftafel
Gruppe <Mathematik>
Vorlesung/Konferenz
Statistik
Subtraktion
Web Site
Prozess <Physik>
Web log
Spider <Programm>
Betafunktion
Natürliche Sprache
Quick-Sort
Whiteboard
Code
Eins
Arithmetisches Mittel
Metadaten
Deskriptive Statistik
Benutzerbeteiligung
Menge
Gruppe <Mathematik>
Server
Vorlesung/Konferenz
Wort <Informatik>
Bit
Minimalgrad
Rechter Winkel
Gruppe <Mathematik>
Vorlesung/Konferenz
Mailing-Liste
Ähnlichkeitsgeometrie
Wort <Informatik>
Frequenz
Term
App <Programm>
Wasserdampftafel
Hochdruck
Ähnlichkeitsgeometrie
Frequenz
Raum-Zeit
Quick-Sort
Wechselsprung
Arithmetische Folge
Gruppe <Mathematik>
Mereologie
Visualisierung
Vorlesung/Konferenz
Wort <Informatik>
E-Mail
Rechter Winkel
Gruppe <Mathematik>
Vorlesung/Konferenz
Quick-Sort
Message-Passing
Hilfesystem
Subtraktion
Mittelwert
Rechter Winkel
Fächer <Mathematik>
Gruppe <Mathematik>
Datentyp
Zahlenbereich
Vorlesung/Konferenz
Bridge <Kommunikationstechnik>
Gerade
Klasse <Mathematik>
Zahlenbereich
Plot <Graphische Darstellung>
Kartesische Koordinaten
Ereignishorizont
Loop
Mittelwert
Fächer <Mathematik>
Automatische Indexierung
Rechter Winkel
Gruppe <Mathematik>
Binärdaten
Vorlesung/Konferenz
Auswahlaxiom
Metropolitan area network
Schreib-Lese-Kopf
Subtraktion
Sichtenkonzept
Punkt
Wellenpaket
Zahlenbereich
Schlussregel
Binder <Informatik>
Kontextbezogenes System
Kommutator <Quantentheorie>
Fokalpunkt
Entscheidungstheorie
Datensatz
Weg <Topologie>
Menge
Rechter Winkel
Gruppe <Mathematik>
Datentyp
Vorlesung/Konferenz
Garbentheorie
Wort <Informatik>
Urbild <Mathematik>
Resultante
Demoszene <Programmierung>
Mapping <Computergraphik>
Gravitation
Bit
Mereologie
Vorlesung/Konferenz
Kurvenanpassung
Teilbarkeit
Rechter Winkel
Vorlesung/Konferenz
Beobachtungsstudie
Bit
Shape <Informatik>
Überschall
Datenmanagement
Freeware
Rechter Winkel
Mereologie
Familie <Mathematik>
Vorlesung/Konferenz
Softwareentwickler
Schnittstelle
Momentenproblem
Gruppe <Mathematik>
Basisvektor
Vorlesung/Konferenz
Tropfen
Momentenproblem
Mereologie
Vorlesung/Konferenz
Gleitendes Mittel
Datensicherung
Mapping <Computergraphik>
ATM
Virtuelle Maschine
Bit
Analoge Signalverarbeitung
Schwebung
Gruppe <Mathematik>
Mereologie
Vorlesung/Konferenz
Schlüsselverwaltung
Quick-Sort
Virtuelle Maschine
Bit
Weg <Topologie>
Rechter Winkel
Schwebung
Gruppe <Mathematik>
Vorlesung/Konferenz
App <Programm>
TVD-Verfahren
Weg <Topologie>
Gruppe <Mathematik>
Natürliche Zahl
Mathematisierung
Vorlesung/Konferenz
Figurierte Zahl
Quick-Sort
App <Programm>
Virtuelle Maschine
TVD-Verfahren
Syntaktische Analyse
Momentenproblem
Rechter Winkel
Natürliche Zahl
Gruppe <Mathematik>
Mereologie
Zwei
Snake <Bildverarbeitung>
Programmbibliothek
Vorlesung/Konferenz
Varianz
Spieltheorie
Browser
Gruppe <Mathematik>
Codierung
Visualisierung
Meter
Vorlesung/Konferenz
Quick-Sort
Computeranimation
Instantiierung
Domain-Name
Vorlesung/Konferenz
Wort <Informatik>
Programmiersprache
Schnelltaste
Web Site
Zellularer Automat
Mailing-Liste
Binder <Informatik>
Code
Physikalische Theorie
Computeranimation
Eins
Virtuelle Maschine
Benutzerbeteiligung
Schwebung
Reverse Engineering
Rechter Winkel
Prozess <Informatik>
Gruppe <Mathematik>
Leitungscodierung
Mereologie
Programmbibliothek
Vorlesung/Konferenz
Optimierung
Ordnung <Mathematik>
Gerade
Metropolitan area network
Videospiel
Momentenproblem
Schwebung
Rechter Winkel
Mereologie
Vorlesung/Konferenz
Hacker
Programmierumgebung
Bit
Kreisfläche
Quader
Versionsverwaltung
Symboltabelle
Medianwert
Term
Unendlichkeit
Arithmetisches Mittel
Schwebung
Rechter Winkel
Mereologie
Vorlesung/Konferenz
Ordnung <Mathematik>
Programmiergerät
Quader
Browser
Relativitätstheorie
Versionsverwaltung
Schlussregel
Gibbs-Verteilung
Gesetz <Physik>
Datensatz
Schwebung
Gruppe <Mathematik>
Mereologie
Vorlesung/Konferenz
Einflussgröße
Aggregatzustand
Softwaretest
Bit
Subtraktion
Mailing-Liste
Quick-Sort
Schwebung
Rechter Winkel
Gruppe <Mathematik>
Mereologie
Vorlesung/Konferenz
Wort <Informatik>
Kantenfärbung
Information
Programmierumgebung
Informationsmodellierung
Gewichtete Summe
Menge
Vorlesung/Konferenz
Aggregatzustand
Domain-Name
Momentenproblem
Gruppe <Mathematik>
Vorlesung/Konferenz
Code

Metadaten

Formale Metadaten

Titel Closing Keynote by Lamere
Serientitel RailsConf 2016
Teil 88
Anzahl der Teile 89
Autor Lamere, Paul
Lizenz CC-Namensnennung - Weitergabe unter gleichen Bedingungen 3.0 Unported:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen und nicht-kommerziellen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen und das Werk bzw. diesen Inhalt auch in veränderter Form nur unter den Bedingungen dieser Lizenz weitergeben.
DOI 10.5446/31503
Herausgeber Confreaks, LLC
Erscheinungsjahr 2016
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract Paul Lamere works at Spotify, where he spends all his time building machines to figure out what song you really want to listen to next. When not at work, Paul spends much of his spare time hacking on music apps. Paul's work and hacks are world-renowned including the 'The Infinite Jukebox', and 'Girl Talk in a Box'. Paul is a four-time nominee and twice winner of the MTV O Music Award for Best Music Hack (Winning hacks: 'Bohemian Rhapsichord' and 'The Bonhamizer') and is the first inductee into the Music Hacker's Hall of Fame. Paul also authors a popular blog about music technology called 'Music Machinery'.

Ähnliche Filme

Loading...