ARTICLE

DATE
September 15th,
2016

Creativity Machine Intelligence

SPECIAL

From The Seminar

Creativity Machine Intelligence

Brian Eno / ブライアン・イーノ

Creativity Through Machine Intelligence: A Conversation with Brian Eno

(Moderator, Yamamoto Koichi ; Yamamoto;Dentsu : Brian Eno; Artist : Togo Kida ; Dentsu Lab Tokyo)

Yamamoto  Hello. So today we’re very, very tremendously honored to have Mr. Brian Eno here today. We’ll have a very casual conversation about creativity and technology.Brian, as you know, he’s been a pioneering artist in all forms of art.
You all know what he’s done musically in the way of basically inventing or conceptualizing, and developing the genre of “ambient music”.

Of course, he’s been a producer and really creating new directions for bands that were fantastic to begin with. He’s totally turned bands like U2, Coldplay, into new directions,and then, something we don’t know that much (about) is your visual art work. He’s been a really pioneering, developing, very unique, I think, visual installations like 77 million paintings, which, they have been exhibited around the world.

So, today, we’ll have a very casual conversation about creativity and technology.
The reason why we’re doing this is actually the “Dentsu Lab Tokyo” is currently working on – or has been working on – a project for Mr. Eno’s title track from his latest album “The Ship”. It’s an interesting thing; I’m not sure what to call it. It’s a visual experience that accompanies the title track.

Let’s first talk a bit about your art and technology, and then move on to the more specific subject we’re talking about, the project.

the word “technology”

Yamamoto  So, first, I’d like to ask a very general question of your relationship with technology.Technology is very important for you. How do you see the role of technology in your work?If that’s changed over the years, how has that changed?

BRIAN  Okay. So first of all, I want to talk about the word “technology”. I have a friend who’s actually a scientist and a technologist, who says: “technology is the name we give to the things that don’t work properly yet”.

[laughter]

Because actually, lots of the things we use are technology. Like, a violin is technology, the grand piano is extraordinarily complex technology. But because they work, we forget that they’re technologies. We’ve just taken them into our lives. This is technology, you know. Everything, there’s a lot of technology. Humans are actually very small creatures without all the technologies that they surround themselves with.

So I started really in music because of a technology. I can’t play any instruments even to this day – I still can’t play any instruments very well. [laughter] But towards the end of the sixties, when I was in my late teens and early twenties, a technology was coming into being that really opened up a whole lot of possibilities.

I was studying painting – I had been at art school for five years – and I suddenly saw that music, with the recording studio, was really a new way of painting. You know, the way you make a painting is by putting one color down and then adding another, and then scraping that one off, and so on and so on.The painting is made in time like that.

Whereas traditionally, music was recorded instantaneously; the band stood together and played – and that was the recording. What multi-track recording did (was) it suddenly allowed you to become a painter in sound. You could build up a recording like you would build up a painting – and this is actually how everybody records now. Very few people record in that old fashioned way of standing in front of a microphone and playing.

So, multi-track recording, which was the technology that released me, was indeed a huge step forward, because it suddenly made music a medium that a lot of other people – who hadn’t grown up within a musical setting or a musical education – could use. So, it suddenly sucked in a lot of new talents from different areas.

Just to give you an idea of what musical technology does, if you think of a very simple device, like a microphone. So the microphone was invented to make voices louder, essentially. The effect of the microphone was to enable people to sing more quietly. So, instead of having to sing [sings] like that, to project your voice out to a roomful of people, [sings] you could sing very quietly.So, in fact, nearly all of the modern styles of singing that we recognize really came out of the fact that the microphone existed; it allowed the other 90% of your vocal possibilities to be usable.

You couldn’t do that in a concert hall before unless there was nobody else playing. But as soon as there was an orchestra, you were forced into that operatic way of singing to get your voice across. So, technology is always invented for a historical reason. And it always, very quickly, has a whole lot of new uses.So as soon as the technology exists, suddenly, a whole lot of new possibilities that nobody even thought of before come into existence. The history of music is full of that.

s_CADENTLIONS16_399

Construct methodology

Yamamoto  So, has your view on technology in your arts – how has that changed your views? Has it been the same since you started fiddling with tape recorders and synthesizers way back when?

BRIAN  Um, no. I think what’s…well, I’ve always really worked in close partnerships with technology because I like it, and because I like the fact that if you’re working with new technologies, you’re very often stepping into territory that has only just opened up – nobody else is there. It’s a very nice feeling to be, to not have any competition. [laughter]

And so, that’s something I’ve carried on doing for a long time. One of the interesting things about technologies is that they have become more and more intelligent. actually more than that. Yeah, maybe 35 years ago. And that was an idea that instead of writing a piece of music down like a symphony, where you write down all the notes and all the things that are going to happen in the music, instead of that, you invent a kind of technical system of some kind, that produces music for you.

So, what I was doing, essentially, was setting up systems that generated sounds. I could control the systems, I could set up the rules for them.
I could give the system certain inputs, but then I’d let them run, and they produced a music that I had never heard.

So it’s different from the classical view of the composer, someone who has a conception of the music in their head, and then they realize it in some way.What I was doing was having a conception of a way of making music, and then building that, and letting it happen.So I called this “generative music”, and the term has actually found most of its application in games.
A lot of the scores for games now are generative, in the sense that it doesn’t rely, as it used to in the past, on a subset of little pre-recorded loops; you go into this scene and that loop plays, or you go into that one and another loop plays – it isn’t like that.

I did the music a few years ago for a game called “Spore”, which is a William Wright game. Instead of doing pre-recorded pieces, I set up a little synthesizer within the game which creates music depending on certain conditions. When certain things happen in the game, certain rules come into play.But it means that when you go back to that same place in the game next time, you won’t get identical music.You’ll get the same kind of music, but not an identical piece.

Yamamoto  It’s an automatic soundtrack generator of sorts.

Brian  Yes, it’s an automatic soundtrack generator, yeah.It saved me no end of time. [laughter]

Concept of The Ship

Yamamoto  Shall we move on to “The Ship”?

Brian  Yeah, sure.

Yamamoto  Yes, so could you tell us a bit about how “The Ship” came to be? And also a bit about why you asked the Dentsu Lab Tokyo to work on this project, maybe?

Brian  OK, so, I was invited two and a half years ago by Fylkingen, which is the oldest electronic music studio in the world, in Stockholm, to make use of the fact that they had a multi-channel reproduction system there.They had lots of amps and lots of speakers, and it was a room, probably about the size of this and about as high as this.

So I started working on a piece of instrumental music with the view that that music would be installed in three dimensions, so it would come from around many places around you and so on.
Whilst working on it, I noticed that I could sing the lowest note in the piece. [laughter]
This is one of the only advantages of getting older [laughter] that your voice gets lower and lower.

And I was fascinated a) by the fact that I could sing this note, which I have never been able to do before, it was a low C, and b) by the idea of using a voice in a context like this. I’d been thinking very much in terms of an abstract, sonic experience, and I started singing this low note and the word that I kept singing was ”roll, roll” I didn’t really know why.And the piece of music sounded to me like a sort of ocean of some kind – there was a lot of that kind of movement in it. It didn’t have a rhythm exactly, but it had a sort of wave form to it. Roll, roll.

I was thinking, “I’m on a ship here, on some kind of a ship”. 「Roll、Roll」。So gradually, the words of this song came to me. In reverse, I was drawing the words out of a description of what the sound was doing, really. Anyway, it turned into this piece called “The Ship”.

Now, I had been reading, for many years, about the First World War, which particularly interests me, and about the Titanic, the sinking of the Titanic.As you know, the Titanic sunk in April of 1912, which was two years before the First World War started, And to me, those two events were very much connected, in the sense that the Titanic showed the sort of hubris of the Victorian era; the feeling of “this is unsinkable, we know how it all works, nothing can stop us now”, and then it hit an iceberg and sank.And the First World War started in exactly the same spirit; it was the same spirit of “we’ve solved all the major problems”.

Actually, that was very much a feeling that was around at the beginning of the last century, that “we’ve solved all the major problems”. Ernest Rutherford, the physicist, said in 1905; “as far as I can see, there are no major problems left in physics”, and that was the same year that Einstein published the “Theory of Relativity” [laughter] and three other ground-breaking papers.
So this is a sort of human tendency to think of, to sort of sit back in your seat, rub your hands and say, “okay, we’ve got it all fixed now, we understand how it works” and then you can be sure that that’s the day before everything crashes. So, I saw that comparison.

The other comparison I saw, of course, was with the beginning of this century where in the year 2000, the American government published a document which was called “The Strategic Overview for the United States”, and it begun with the words: “At the end of the twentieth century, it is evident that the only viable form of government…” and then it went on to describe American democratic capitalism as though, again, all the problems had been solved.

And this was in the wake of the Francis Fukuyama book, “The End of History”, which was the idea that really, we didn’t have to think about it anymore; the future would be democrats, capitalists, and gradually the rest of the world would realize that and everything would be fine.
And then, two years later, well, one year later, there was 9/11, and then, two years after that, there was the Iraq War, and then the whole thing hit the fan.
So it’s that same profile of events again, the hubris that — we’ve done it; we’ve solved all the problems.

And then of course, the crash, and then the paranoia that always follows the hubris, because the biggest characteristic of huge power is huge paranoia. And suddenly, America found itself incredibly vulnerable; saw terrorists in every crack in the wall, and gave rise to the biggest industry of the 21st century, which is security, [laughter] and the biggest obstacle to the rest of our lives. So, partly, the record was coming out of me thinking about this echo effect of this thing that seems to happen about every hundred years that we solve all the problems and find we haven’t.
s_CADENTLIONS16_419 (1)

Yamamoto  So, why did you come to us? We were very honored, but frankly we were very surprised.

Brian Well, thank you. I think, as I said, I’m interested in finding out what new technologies can do, because I always know that they can do something that nobody ever thought they could do; they were invented to do one thing, but you can be sure they do something else much better.

So, artificial intelligence is something that I’ve been interested in for quite a long time. I have several friends who are working in that area.And I’m not frightened of it.I’m frightened of the people who currently control it, [laughter] like the NSA and so on. Artificial intelligence isn’t the enemy – the enemy is the people who currently have control of it, but I think it’s a tremendously fruitful and interesting area.

And the second thing is, just as I am excited about the possibilities of artificial intelligence and new technologies, I’m so, incredibly, numbingly, bored with conventional music videos, that I just couldn’t imagine wanting to do that.So, really, it was an attempt to say “is there some other way we could think about doing this thing?”

Yamamoto  So, just to give you a brief description of the lab, the lab we created to experiment in new ways of creativity, so how can we find new creative processes leveraging technology, This is really a project that really is symbolic of what we’re trying to do at the lab.
So we are so, so excited that we’re able to work on this project with probably the best creative director out there for, like the lab, Mr. Eno.
To talk about the project a bit more, let me bring out Togo Kida from the Lab.

[applause]

プロジェクトの構想と機械知能の社会的位置

Yamamoto So Togo, run us through how you interpreted the concept of “The Ship” that Brian just talked about how you interpreted that from a technological point of view.

Kida Yes, of course. When we were approached by Brian Eno and we learned about this project, and we heard about his concept about “The Ship,” about this contrast between hubris and paranoia, when we thought about what this topic would be about in the current society, we kind of thought about how it resembles how artificial intelligence, or, we like to call it the machine intelligence, is situated in the current society.

So, we decided to not rather deliver a music video – he just mentioned that he hates the idea of a music video – but we decided rather to develop a software – a machine intelligence – that constantly generates this music video on top of the video itself.
So, with that fixated, we were thinking about how this machine intelligence could make an impact and what kind of meaning it would have on this music video.

We were thinking about how we can achieve creativity through machine intelligence and make this question, about is intelligence approached by a machine intelligence, is it something creative, or is it not really creative? We wanted to ask this question, so we decided to set using machine intelligence as the core of the project and decided to further explore the idea.

Brian Can I just add one thing to what you said?

Kida Yes.

Brian Part of the objective is to make not “a video” but to make something that is constantly different each time you look at it. So, the point is not to make something that just reproduces exactly each time, but to create a kind of machine that generates something new in relation to the project each time.Sorry, I just thought I’d make that clear because it’s not an idea that many people are familiar with….

Kida  So we have a video that kind of explains this, an introduction to the project, so if you could go on to view that.

****************************************

****************************************

Markov-Chain Generator

Yamamoto Rather, Brian, let me first ask you, you’ve used, actually, in the piece itself, in The Ship, you’ve used a Markov-Chain Generator, a random number generators to generate lyrics and stuff. So what did you think of this kind of approach?

Brian  Well, I love it. [laughter]
What I always like is to create things that take me somewhere that I haven’t been before.
That’s always the thrill to me. It’s like coming over the brow of a hill; suddenly there’s a new landscape there, and it’s full of things to discover.

So I like tools that do that – as you said, I used, on this record, a thing called a Markov Chain Generator, which is a sort of statistical randomizer, as a way of writing some lyrics.
It produced incredibly interesting results, but the important caveat is that what really matters if you’re using machines and systems and algorithms to generate things, what really matters is what you put in at the beginning and how you select what comes out at the end.

So, typically, with the Markov Chain Generator, for example, I would say that about 90-odd percent, actually more than that, 95%, probably, of its output was of no use at all.
So I’d read through these reams of stuff coming out of it, and with a highlighter, I’d occasionally find, “oh, that’s an amazing sentence”, or “that’s an amazing combination of words”.

So there’s still very much a human being involved in this. [laughter] What the system is doing is putting things together that a human – at least this human – would never think of putting together. What the human is then doing is saying “does any of this make any sense to me?” So, you do need to be at both ends of the process.
So in a sense, although we are talking about artificial intelligence, there’s a lot of human intelligence still mixed in with it.

Yamamoto  So in that context, Togo, what did you input into the machine intelligence, and what’s the, kind of, generating process?

Kida  Yes, So, what we did with this machine intelligence is we had this machine intelligence study various kinds of images available online – some in public domain – but what I just wanted to mention is that what we did was, we collected various images on public domains and we made this machine intelligence learn about it.
What the machine intelligence does is, after its learning, it would start to understand things that are available in the images.

We then, after, we collected over, images from over a century or even more. And what we did was, in a way, kind of outsourced the human memory – collective human memory – on the internet or on the public domain.
After that, we had this machine intelligence look at now in the present society, the images available online, and had this machine intelligence making associations based on the learning from the past. So as Brian mentioned, it’s about what kind of input you put into it.
In that sense, the images from the past were the input, and the outputs were determined by the images being observed from the current society.

There’s a video of this so please have a look.

********************************************

********************************************

Yamamoto   So in the way of visualizing this, tweaking the machine intelligence, could you talk a bit more about what you did?

Kida  Sure. It is a fairly complex topic. We’re talking about human intelligence and trying to relate that with artificial intelligence, and trying to replicate this process of associating things from the past and the present.
When you say something like this, it’s very easy.
It’s something that we do every day. We see something, and we imagine things, and we kind of associate with the past memories.

But in terms of visualizing, this is something fairly, very complex. You can explain about the process, but you can’t really tell how it’s done.Nobody ever knows what’s going on inside your head, so that’s one of the biggest struggles we had.

There’s a video that kind of explains about how we approached (this) from our other project members, so please have a look.

********************************************

********************************************

Yamamoto  So, this is still very much a work in progress, I believe, Togo, but could you explain a bit about what it is at this stage?

Kida  Sure. So, as you can see, there are three sets of structures that’s being built within the software here. The image you see on the bottom – I think you briefly saw what it was, but – it kind of shows, it takes in some of the latest images from Twitter.

This juxtaposition of images you’re seeing here right now are based on the machine intelligence software. It’s picking up relevant images studied from the past.
So right now you see: Islamic state claims responsibility for a bombing near Damascus. So it’s looking at this image and based on the learnings from the past it’s recurring some of the old images.

Right now, you see some catastrophic images from the past – some could be from the tsunami or could be more of a war situation.
It’s picking up some of these images based on relativity.
And at the top of the image here, it’s combining these images as more of a collage figure and kind of undulating it and shifting towards different images, kind of mimicking the situation where the machine intelligence remembering things from the past based on the current events.

Yamamoto  Fascinating. So we’ve mentioned this is a work in progress. Actually, Brian and us got together about an hour and a half ago, and had a discussion about where we should be taking the project.
It turned into a real work session and rapid prototyping session, and this is actually something that we just churned out an hour ago, really, and something we might evolve this towards.
Can you talk a bit about this approach, Brian, what we’re trying to do here?

Brian  I’d been looking at the output from the original system – the one that we were using until about three hours ago – and it was a single screen piece. It lacked dramatic tension, I felt.
It was very interesting: what it was doing was putting images together and melding them into one another.
Then, I remembered the painter David Salle, who you may know, who always makes his pictures, or used to make his pictures, by putting two often quite contrary images together.
So it occurred to me…what would happen if we started doing exactly the same processes as we’ve been doing, but we did two of them in parallel, so the same question is being put to the software; basically. Here’s a piece of contemporary news footage – we’re in a particularly abstract moment here, I have to say – then, it’s searching through its memory files and through the net, to find the things that it judges as being similar to that.

We just did this over the weird melon soup that I had in the hotel. [laughter]
It’s quite abstract here, but what we’re trying to do is to tune this a little bit more so that the two panels of images start to rub against each other a little bit more.
So we’re going to do that, we hope, by both the use of different visual treatments on the images, so that they separate out a bit more, and by tuning the algorithm so that we don’t get quite so long periods of gray. [laughter]

These images – well, of course, they’re supposed to be with a piece of music, which we’re not hearing at the moment – I think they’re very striking with the music. 

Artificial stupidity of artificial intelligence

Yamamoto  Yeah, he told us to be prepared to do a lot of changes going forward. Could you give us a heads-up in like longer range, what kind of new uses of artificial intelligence and machine intelligence, where we might be able to take this beyond maybe even the end of this project?

Brian  So, artificial intelligence, is, at the moment, proceeding so quickly, there’s so much going on in that field, and yet at the same time, it’s so far away from where it thought it would be now. I’ve been following AI for a long, long time now, and at this point, we all expected to be much further ahead.

What that is, is a tribute to how complex the human brain is, actually. – our intelligence is so complex that some very simple tasks, like understanding context, which we do very easily, are very difficult. So, one of the things we can do is to capitalize on what the computer does have, which is artificial stupidity. [laughter]

Computers make some very weird mistakes and a lot of those mistakes are very interesting, actually. So, one of the things I should have said about technology earlier, is that one of the things artists are interested in technologies for is the things that they do that they weren’t supposed to do.

In fact, the dominant texture of any era is really captured in the shortcomings of those technologies.For example, if you think of the early 20th century, you’ll think of grainy, black and white or sepia photographs. Those were limitations – people weren’t excited by the fact that they could only do black and white pictures. They wanted to do color but they couldn’t.
And similarly, if you think back to the recordings of the thirties, [mimes music] they all sounded like that. And that now becomes the sound of the 1920s and 1930s.So, the characteristics of an era are actually to do with the shortcomings of the media of that time.
Shortcomings that artists like, for instance, technicolor, an as-yet-unsurpassed color system, was a mistake – it was a way of trying to record color on three different reels of film at the same time, and it always came out wrong, and we loved the way it came out wrong.

The films of the early fifties and in to the early ‘60s, actually, before the deadly hand of Eastman Color appeared and everything went brown.
Jesus, this getting a bit abstract, isn’t it? It’s like I’m going into Mark Rothko territory [laughter].
So what I think is what we’ll keep doing is capitalizing on the things that don’t work, as well as things that do work – so the artificial stupidity, as well as the artificial intelligence.

Kida  On top of what Brian just said, I just wanted to (add that) since this is a piece of software that can evolve in different forms, not just something that’s visible on screen, but it could be transforming to installations, or more of an interactive experience, that you can have it more in a physical way.
So I’m kind of excited to…how far we can explore this project now, and of course, this artificial stupidity there’s more possibilities that we can explore.

Brian  You can use organic stupidity [laughter] We could get Donald Trump if you like. [laughter, applause]

Yamamoto  We can use Trump’s often reversing quotes and opinions and kind of feed it into the machine intelligence, see what that turns out.

Brian  But anyway, this is very much the beginning of something, I think. It’s very much the beginning of using a technology that is around and is generally being used to do things like track your phone calls and who your friends are and finding ways of selling new things to you. We’re trying to see what else it can do, I think.

Yamamoto  So I guess I’ll put the URL up there. Please visit www.theship.ai. June 28th should be the date, but I believe we will be updating and I believe evolving the experience on and on and on and try new approaches as we go along.
So if you find the experience captivating, please come back repeatedly. You probably will find a totally different experience hopefully as the project progresses….

Q&A

AUDIENCE MEMBER Do you think this technology could be used by young people, like in the next week or so?

Brian  Not in the next week. [laughter]

AUDIENCE MEMBER  Like young musicians looking to make some new video or something like that?

Brian  Well, yes, I mean… it would be lovely if we came up with a sort of system that was usable for people, yeah. That would be great. And I’m sure that if this works and becomes popular, or even notable, a lot of people will be doing it too.

AUDIENCE MEMBER I have another quick question as well, more of a serious one. Would you say this kind of technology, especially with the new stuff, do you think it could be used for educational purposes?

Brian Do you mean education in relation to news? For instance, like studying history, for example? 

AUDIENCE MEMBER Yeah, exactly, because that’s what came to my mind straightaway with all those historical images.

Brian  Yes, I think that would be wonderful. Education is such a disaster zone at the moment that anything would help, I think. [laughter] I always say that children are born with amazing imaginations, and it takes about 11 years of education to really get rid of them. [laughter]

AUDIENCE MEMBER 2 Don’t you think that what defines your work more than technology is the ability to create emotion and narrative, and universal things that resonate with the rest of us?

Brian  Okay, let me repeat that, because if it gets too long, I can’t remember the beginning.
He said that don’t I think that what defines a work of art is not technology so much as the ability to create emotion.
Yes, I do believe that, but I think that’s something we’ve always done with technology.
Think of the first cave painters, you know, those pictures of hands sprayed on the wall. Well, what were they sprayed with? With the technology, you know. [laughter] They had a straw and they blew their silhouette.
And as I said earlier, the grand piano is an amazing and very complex piece of technology.
So I don’t think there’s an opposition between the artist and the technology.
Technologies are what we live in. We all got here, in every sense, not only were we born by, but transported to this room by technologies.
So, as I said, the only technologies you notice are the ones that are new, or the ones that don’t work very well. The other ones we just accept as part of our lives.
But you’re quite right; that has to be the objective – if we don’t achieve that, then it’s just cleverness.

MODERATOR Thank you. I think that…big hand. Thank you so much.

Brian  Thank you.

NEW POSTS

Dentsu Lab Tokyo