Scroll Over Beethoven: Machines Making Music

Scientists have reported seeing some unusual giant panda behaviour: some of the creatures appear to have a penchant for rolling in fresh horse manure in winter. The theory is that the muck contains chemicals that deaden the panda’s ability to feel the cold. Katie Haylor explains...

Katie - Using infrared cameras to observe a year of giant pandas in the wild, scientists in China have documented around 40 incidences of these bears sniffing, wallowing in and rubbing horse manure - yep, horse manure -  all over themselves, particularly in the colder months. So to reflect on this rather - at least to my ears - unusual finding, I spoke to Anglia Ruskin behavioural biologist Claudia Wascher.

Claudia - It is quite unusual to actually see a species rolling in the faeces of another animal. Individuals avoid going close to faeces because potentially these materials can have parasites, which animals can get affected with, or bacteria. So yeah, it is quite unusual.

Katie - Now I'm no panda expert, but I wouldn't have thought they spend that much time around horses. But it seems the ancient trading routes that cross the bears' habitats have brought these animals in close proximity. And timing is important here because it seems the fresher the faeces, the better.

Claudia - So they have compared how often does this behaviour occur when there is fresh manure compared to all older ones, and it is most frequent in the first couple of days.

Katie - Lovely. So why on earth are pandas giving themselves a horse poo spa treatment?

Claudia - They think that there is a specific chemical component in the horse faeces which is an adaptation to cold. So basically rubbing in the horse faeces would then make the pandas feel less cold.

Katie - In addition to that beautiful monochromatic fur the theory is that pandas are cleverly seeking out additional ways of staving off the chill at colder times of the year, mostly in November through to April.

Claudia - Don't think about it like having another coat on or having a pullover on. But it is more of a reaction. Similar to if you would eat a hot pepper, you would feel warm because there are certain chemicals in the food, which cause a chemical reaction, which causes the sensation in your body. And this is a similar case.

Katie - Claudia is talking about beta caryophyllene, or caryophyllene oxide. Chemical analysis showed that they're particularly prevalent in fresh horse manure. And tests with giant pandas at Beijing zoo suggested that they seem to prefer hay with these chemicals on, compared to controls. And because this poo rolling seemed to be temperature dependent, the scientists wondered if these chemicals could have a role in how animals sense temperature.

Claudia - They have chemically extracted this component from the horse faeces and have then injected mice either with this chemical component, or with a saline solution. And then they have looked in the mice, what kind of behaviours they display in response to cold. And the mice showed more behaviours which are expected as an adaptation to cold; so more shaking behaviours with the controlled substance. And this hints towards that the mice injected with this chemical component would feel less cold.

Katie - So what are these chemicals getting up to? The authors did some laboratory work to delve into this question. And I asked Claudia to summarise this.

Claudia - This chemical component actually responds to a specific channel within the nervous system. What basically happens is that the sensation of cold is a specific sensation, which means that certain parts of the body respond to a certain environmental condition. So nerve cells in the skin, activation in these nerve cells would cause us feeling cold. These nerve cells are inhibited by this chemical component.

Katie - So the authors reckon that on a molecular level, these chemicals are blocking receptors involved in cold sensation. Fair play to giant pandas! It seems rather clever to make the most of the surrounding environment like this. Others are less convinced. In the Science journal magazine this week, the University of Glasgow's Malcolm Kennedy points out that perhaps the pandas are just curious about this horse manure. And that if this lack of feeling chilly were to stop pandas seeking shelter against the cold, this could be highly problematic. Turns out that rubbing stuff all over oneself is not without precedent and the animal kingdom.

Claudia - There's a whole group of self anointing behaviours where vertebrates or different groups of animals rub themselves in some sort of biological substance. Faeces of other species, soil materials, plant materials. They are usually very poorly investigated. This is one of the first studies like this, which I'm also seeing opens up a lot of avenues for future research.

How much carbon is in the humble crisp? If you take into account the carbon footprint for growing the actual potatoes, it’s a reasonable amount. Nevertheless, it could soon be 70% lower! That’s because crisp-makers “Walkers” have adopted a new chemical process that recycles waste carbon dioxide and potato skins and turns them into fertiliser. Developers CCm are intending - initially - to capture carbon dioxide from a nearby brewery while they set everything up at the crisp factory, and then embrace other sectors of the food and drink supply chain later on. Phil Sansom heard how it works from CEO Peter Hammond...

Peter - Walkers have agreed to be the first users of our technology in the food sector, transforming waste from the crisp factory in Leicester into a fertiliser material, which can then be used to grow more potatoes to supply next year's crisps.

Phil - What's the point? Don't we already have fertiliser?

Peter - You do. But unfortunately, fertiliser is a very energy intensive material to produce, and doesn't necessarily reach plants in the way that you might hope. So one way and another, it has a very high carbon footprint.

Phil - What's this technology then? How does it work?

Peter - What we do is use carbon dioxide to stabilise nutrients that are contained in waste materials from the Walkers factory in Leicester. Those waste materials predominantly come from potatoes - it's the skins - but it's also carbon dioxide that's produced in the plant.

Phil - Can you give me any insight into how that works? Because obviously carbon dioxide: that's something that everyone's trying to get rid of, but it's not an easy, easy task.

Peter - No, that's true. Carbon dioxide is quite unreactive, and so that makes it a difficult customer to get rid of. However, we're very lucky in waste materials that one of the chemicals that's actually relatively freely available in them is ammonia. And that ammonia very much likes to react with carbon dioxide. The first step then is this capture of the CO2 by the ammonia that's held in the liquid. We have to add some additional materials, particularly ones that are high in calcium, and also some nitrate material, so that the ammonia that's held there is transformed into ammonium nitrate, which is a really good plant nutrient. And the whole system is stabilised by the creation of calcium carbonate; chalk to most of us. And that helps literally glue the whole system together.

Phil - So I've got my waste and carbon dioxide going in, and then I've got my ammonium nitrate and calcium carbonate coming out the other end?

Peter - Absolutely. All stuck on to the organic substrate underneath, which comes from the potato skins.

Phil - Does it work well enough to make a good amount of fertiliser? And then does that fertiliser itself work well enough?

Peter - It certainly does. We've actually been testing it for the last six years now. And we found that across the board, we can attain the same yield and growth patterns that you would see with conventional fertilisers, so that's been very encouraging from the start. But also we started to see additional benefits in the soil, which are due to our adding these additional materials that wouldn't be delivered with normal fertiliser materials.

Phil - Now, Peter, you also said that sort of the point of all this is dropping carbon emissions. What sort of carbon emissions come out of your process? Obviously you said carbon goes in, but does carbon go out, and how does that compare to fertiliser at the moment?

Peter - Typically a ton of nitrogen based fertiliser will generate between three and a half and six and a half tons of CO2 for every ton of material that's produced.

Phil - That's a lot.

Peter - Yeah, it is. That figure is now dropping down to less than half a ton of outputs from our process. And the reason we can do that is we're simply drawing those materials that normally have very energy intensive production routes. We're recovering those from what are currently regarded as waste streams, and it actually stops them being waste materials.

Phil - Is that just a drop in the ocean, or is that something to write home about?

Peter - No. It's that use of fertiliser that makes agriculture have such a huge greenhouse gas impact. There's a lot spoken about the methane from cows and that sort of thing, which is obviously significant, but the largest input is coming from the use and misuse of fertilisers.

Phil - If I just take a sample out... I've got my lunch here. Right, this ready salted crisp: how much savings is there on it compared to how bad it was for the environment before?

Peter - At least 70%.

Phil - At least 70%?

Peter - In two years time, it will be a hundred percent better than it was before.

Phil - I'm going to tuck in with pride.

It's Beethoven's 250th birthday, and so we’re looking at how maths and computing can analyse his music. There’s no doubt that he has inspired generations of musicians; despite Chuck Berry telling him “to roll over”, you can hear his flourishes in the samples of the rapper Nas, and subjectively, his style in the instrumentals of bands like Yes and Pink Floyd. But objective statistical analyses can tell us even more, and suggest that he may - for a century or more - have been the most influential composer around. Phil Sansom heard from statistical physicist Juyong Park, who’s been mapping out the similarities and differences between the notes of different pieces of music, to try and pin down an abstract concept into maths…

Juyong - We're trying to find out who's the most creative composer. And of course in order to be creative, you actually have to have new elements that you introduce that have never been used before. But one of the problems with using only novelty as a measure of creativity is: it's actually pretty easy to try something new, just find something that hasn't been used before. So we actually have two measures, or two elements, of creativity. One is novelty, and the other is influence: how much you inspired others who have come after you.

Phil - This is sort of your model of creativity, then, is: how much new stuff you tried, and how much people liked it?

Juyong - Yes, you can say that. Of course, this doesn't really cover everything about creativity, but I think these two are very essential ones when we talk about how to model or how to measure creativity. We compared compositions by nineteen composers, all the way from the earlier ones from the Baroque era, like Haydn and Bach, to the late romantic ones like Rachmaninoff. And what we discovered was: Beethoven was the most influential composer on the evolution of music after his time.

Phil - How can you tell?

Juyong - We're actually looking at not just one note at a time, but one note and the one that follows it. Let's say you just play do re mi fa, then we're looking at "do-re", so one transition; "re-mi", one transition; and "mi-fa", one transition. And those transitions actually are like a fingerprint of that composer, and depending on how common those elements are, we can actually determine how similar the composers sound.

Phil - Were the same people both really novel and really influential, or who was what and who was the other?

Juyong - Novelty and influence don't necessarily correlate perfectly. Beethoven is actually not the most novel, but the new things that he actually tried were actually widely accepted by the others have followed him. And his influence remains number one for many decades after his death.

Phil - How much of creativity do you think that you've covered in this analysis? Because obviously it feels instinctively like it's quite a human concept that you can't translate into a computer.

Juyong - Yeah, words like 'creativity' actually bring up a lot of thoughts in people's minds and a lot of ideas, because it means a lot of things. Some people say, "okay, that is what makes our civilisation evolve". A good analogy would be energy. To me when I hear the word energy, I immediately think about, you know, half mass times velocity squared, for instance; but to many people energy actually does still mean 'life force', like, "hey, he's got a lot of energy". Because we were able to come up with a physical theory of energy, we were able to build those great things that actually help our civilisation. And I think creativity is in a very similar stage, where we're not trying to restrict the meaning of creativity, but actually we're trying to come up with a self-consistent useful framework for measuring creativity. And hopefully that will help us come up with new science and new technologies that actually help us be more creative.

So far we’ve heard how computers can both analyse and inspire creativity; but are they starting to turn the tables on us and become creative themselves? The field of 'creative computing' has progressed far in the last few years alone - at least according to creative computing expert Rebecca Fiebrink. She told Phil Sansom that the scale of musical possibilities would have made poor Beethoven's head spin...

Rebecca - I think he'd be astonished. Obviously computers don't just allow us to replay music; they allow us to record it, they allow us to compose it, and to use increasingly interesting algorithms to process and generate music within the computer.

Phil - Is that true? An algorithm can actually generate music?

Rebecca - Well, it depends on your definition of music, but I think we're getting very close - if not already successful - in having computer algorithms that generate patterns of sounds that people would identify as music, and even very enjoyable music in some cases.

Phil - You're talking here about machine learning, right?

Rebecca - That's right. Machine learning is a set of computational techniques for finding patterns in data, and then generating new data that includes similar patterns.

Phil - And what are they looking for in the music? Is it which notes come after which notes, or is it other stuff?

Rebecca - It depends on what kind of machine learning you use and what kind of musical representation you use. One common way of using machine learning in music is to think about music as a sequence of notes over time. And this doesn't work equally well for all music, but if you want to get a computer listening to the melody of a pop song or a folk tune, then actually it's not too bad. Some of the current machine learning techniques that have just come out in the last year or so have been successful using a different representation of music, more like the representation of music you would use if you stored a recording on your phone in order to listen to it. That representation of music is much more complicated, right? You're not just capturing the notes that are playing at which times they're playing, but you're capturing information about the instrumentation, the volume, the texture; like the way that a piano sounds, or a drum machine sounds, or a singer's voice sounds.

Phil - That last one you mentioned boggled my mind. Can it really capture how a singer sounds and can it recreate a singer as if it was that actual singer?

Rebecca - Absolutely. And I would not necessarily have believed this even five years ago, but machine learning technology has moved so quickly. In April this year a company called OpenAI came out with... it's called Jukebox. And one of the first examples I listened to there was a song; and I use that word loosely, but it's a new, fake song sung in the style of Frank Sinatra, called... I think it's "Hot Tub Christmas".

Phil - Can we take a listen now?

Rebecca - Absolutely.

Phil - Wow.

Rebecca - It's hilarious! It's timely, it's actually a pretty fun song; I'm putting it on my Christmas playlist right now. And it sounds like Sinatra.

Phil - You're saying those aren't samples of words taken out of different songs of Sinatra; those are actual bits of sound that it has generated to mimic Frank Sinatra.

Rebecca - That's right.

Phil - That's crazy to me.

Rebecca - It is crazy. To be fair, this algorithm did not also do the job of coming up with the lyrics. The researchers along with their lyrics generation system decided, "oh, this would be fun as lyrics for a new Frank Sinatra song".

Phil - I mean, elephant in the room: the backing track sounds like a haunted funfair.

Rebecca - Yes, yes. That is a limitation of this type of music generation algorithm. The OpenAI folks who made this system did some clever tricks to figure out how we could even do this at all. And one of those tricks involves some noise in the generated piece of music; it's going to sound a little bit like background noise, like you might hear from an old record; some of it is going to have some weird stuff happening; and the pitches that you hear... things might sound not totally in tune.

Phil - Given unlimited computing resources, do you think that they could basically recreate modern music as we know and love it? Or are there more fundamental limitations to an algorithm trying to be creative and make new music?

Rebecca - I absolutely believe that these techniques are going to get better and better in the next few years, approaching the kind of quality we would expect from a recording of a real musician. There are a couple of things that I still see as big barriers. One of the barriers is that it's really, really difficult to generate music that sounds believable in terms of the longer term structure. A Sinatra song is likely to have verses and choruses; when you look at classical music you're going to have much more complicated structures, things like symphonies, where one theme from the first movement might reappear in a changed way in the end of the piece. And these systems really don't have the ability to represent or generate structure at that scale right now. Another big barrier is these systems tend to be very hard for people to control. You can hit redo and have it do the whole thing again, and come up with something different, because there is randomness in these systems; but if you say, "you know what, I'd really like this to be a little bit peppier," or maybe, "I want this to be in a different key," or, "I want there to be a trumpet in this song," right? There's all sorts of things that humans would want to do, if you start looking at this as a tool for making new kinds of music.

We’ve seen how computers have musical strengths and weaknesses that are very different from those of humans. And Phil Sansom has been to visit a musical project that no person today could ever play; or even fully listen to. Longplayer is a thousand-year-long piece of music - yes, you heard that right, one thousand years - inside a former lighthouse in London. Phil met its creator, formerly of the Pogues, Jem Finer…

Phil - Good Lord. That's incredible.

Jem - Well, this is the sort of mezzanine level. There's a large circular system of shelves upon which are over 200 singing bowls and they are the actual physical instruments from which Longplayer's composed.

Phil - I'll be honest. I feel like I'm about to be reborn into some sort of ancient religion.

Jem - Right, okay. Well, what I was interested in was time as a much longer slower process than a human lifetime and human experience. So how do you keep something going for a thousand years? If it's a piece of music, how do you keep it playing? What technologies do you use?

Phil - How does it work?

Jem - How does it work? There's source music and an algorithm which operates upon that. So the source music is made up of singing bowls, sort of brass handheld bells that are common in many Eastern cultures. And you can play them in two ways. You can either strike them - but more interestingly, you can take a kind of wooden beater and run it around the rim, rather like you could rub your finger around a wine glass. There's six pieces of very short source music - twenty minutes twenty seconds long. And each is composed of singing bowls and silence.

Phil - My maths isn't amazing, but that sounds like about two hours, not a thousand years. So how do you get to a thousand year long piece of music?

Jem - It's not like they're sort of all added together. What happens is that every two minutes, the algorithm chooses a start point in each of those six. So what you're hearing is a superimposition of six different sections. Each plays for two minutes, and then at the end of the two minutes the start points move on. And the amount each start point moves on is different for each of the six pieces. The easiest way to visualise it is to think of a diagram of the solar system. So you see the sun in the middle and the planets arrayed out, and they're all going around the sun at different rates. So once in a few billion years, they might all be in a straight line, but it will take a very long time until they're all in a straight line again.

Phil - So you're saying, if I listen in a hundred years, I might hear five of the pieces that are in exactly the same spot that I heard last hundred years, but the sixth piece is going to be slightly different, and you'll never get them all in the same place for a thousand years time.

Jem - That's what I'm saying. Yeah.

Phil - To be honest, if it was repeating, I'm not sure I could tell.

Jem - I mean a lot of jokers say, Oh yeah, I've heard this bit before. And in a way they probably almost have. It's a bit like weather - you know you've seen skies like that, but you'd never seen the clouds in exactly those shapes in those places.

Phil - Is there a sense that this is more algorithm than composition?

Jem - Well it's certainly algorithm. I think one of the great things that digital technologies afforded is this idea that you can iterate through huge spaces of possibilities in ways that you really couldn't necessarily as a human. So I see it more as a sort of prosthetic aid to the composer to move into realms of music, which are sort of beyond human endurance to actually perform them in some way.