Skip to main content

tv   Coding Art  Deutsche Welle  January 27, 2021 10:15am-11:00am CET

10:15 am
fortunately wants to practically ban the tourism much as as israel has done that's according to members of her church. up next doc film a documentary about the use of artificial intelligence to n.p.r.'s world stage of. life on earth one of a kind and. a joy to coincidence. where the total happened. it's a bit like winning the lottery. for unique starts feb 11th on t.w. .
10:16 am
amsterdam's rights museum is home to the nightwatch rembrandt's most famous painting. researchers are gathering data that will be fed into an ai machine learning algorithm. how did rembrandt create this extraordinary work. could there be a computer system at the heart of such masterpieces. could there be a code for what we call our.
10:17 am
all time to unpack. in his berlin studio the artist foreman lipski shows us how the dialogue between him and his music began 3 years ago. lipski paints and computer answers. in a i muse who knows precisely how he works. and is one. of the dust but this was the starting point i wanted to paint this image i took the motif and painted it 9 types. from the top left to the bottom right in the end. and what i did was digitalize the
10:18 am
images including all the interim phase and. then i sand florrie on a file of about 50 images of the digital photographs i'm sick. them so that he had information on the algorithm or. you saw the algorithm works based on looks these images the air i system has learned to replicate the artist's style. the system can now turn whatever data it is fed with into a lipski painting. it's going to have a thing that i'm looking at what the individual networks have generated. how they alter the original picture. and with that i'm trying to create an optimal. and. an image i can
10:19 am
use as the basis for a painting on a campus. this is. basically it's like an ultra modern sketch. software on these it's in there largely basically our music is a type of software. these so-called neural network we use very specific architecture a machine learning system is capable of analyzing roman lipski paintings distilling facets from them deconstructing them and then constructing something new and generating hundreds of thousands of variations. so again. he does his theme of here you can see how the system interpreted this section on the right has a lake so added blue that it. must have thought for it's the only image generated that has that feature. so that really fires up room or inspires him to
10:20 am
take it to the next level and pick up on certain details of the line it does 60 ai is incredibly fast but roman once said to me that what's exciting is that this allows him to take his time as. never paints exactly with the artificial muse proposes it simply points him in a direction gives him an idea a spark a stimulus. in the last few years all he has to do when he's seeking inspiration is plug in his computer. the arrival of the artificial muse in his life helped him out of a profound creative crisis. this is only for the. work this here is a later work. in
10:21 am
my work used to be representation of the i was trapped in a pattern published at some point i started thinking more about abstract painting the. as a book form and father were shifting when i learned that none of my attempts worked out i just wasn't coming up with anything better than me and i ended up with a creative crisis in season. i was wracked by doubt. unsure if it was what i should be doing been. living this could be i'm still alive you see and i have been plenty of artists who killed themselves this isn't this is seriously these doubts can be overwhelming when you start to question your very existence. it's not cool or existence in fog or stirs this is called. why do human beings create art.
10:22 am
where does the impulse come from. impulse strong enough to dispel overwhelming feelings of despair self-doubt. mozart's requiem is about mortality. shakespeare explores the concepts of love power and jealousy. according to anthropologists the earliest expressions of human creativity date back some 40000 years. modern homo sapiens developed a new kind of self-awareness and began to find ways to project thought and feeling onto the world around them. and i think our artistic and creative works are an attempt to actually sober scientific problem which is the whole problem of consciousness that i can't know what it feels like to be you and you can't know what it feels like to be me but our
10:23 am
creativity is somehow our best way of trying to share that world in our art is almost the best f. m.r.i. scanner to see what it is like to be another human being. hey siri how can i help you. isn't the sunrise magical summarizes the moment when the upper limb of the sun appears on the horizon in the morning the term can also refer to the entire process of the solar disk crossing the horizon. rembrandt was renowned for his ability to portray deep human feelings and challenge the conventions of the self portrait. is it possible to replicate his genius. that's a question explored in a privately financed project called the next rembrandt. here data scientists an artist dorian's set out to create a rembrandt painting using artificial intelligence. the unbeliever team was resupplied the surface dater of rembrandt's work we scanned some of his later works
10:24 am
and were able to provide a dataset of their surface structure. to work as. well as and you can see that was very important in terms of ensuring his brush stroke was convincingly replicated in the new virtual rembrandt dance which tim. henman is against and. when we trained the algorithm with this dataset we were able to calculate chemical resemblances the machine could then identify places in the body of work where the paint had been similarly mixed in the palette of any commission order. computer of wood or stone in beautiful links in the enormous and what we have on the left here is a chemical map of the painting depicting the various palette mixes used in rembrandt's workshop. and then on the right we can see where these colors were used
10:25 am
in the painting for sheen and. in beard. parts of the face and we started to compare them and based on this we're able to create a typical rembrandt i love henri. after generating the features we were focusing on the face proportions muse and algorithm that kind. over 60 points in a painting we were able to align the faces and estimates the distance between the eyes the nose and the mouth of the. painting is not a tutti picture it's treaty you can see the canvas you can see the process and that's what makes the painting come off the high school i was essential to make the painting of paint. we incorporate that the hype into the painting comparing to on
10:26 am
a trainee printer that uses a special paint base. it printed many layers one on top of the other which resulted in the height and texture of the final painting. surely only rembrandt could create a rembrandt. yet this rembrandt was created by a machine. so what is it. imitation. or is it art. skipped too much money to move from people are always asking will artificial intelligence be more intelligent than human intelligence as i always say have you
10:27 am
ever used a pocket calculator even that is way more intelligent than you are so it's nothing new the idea that certain technological systems that simulate intelligence can outperform humans but also for us to conclude from that that a pocket calculator is a being that is aware of what it's doing and that is irrational i think you'll any say really see computers being creative in code being creative not not just being a tool for our own creativity. when they have their own internal worlds now that's probably a long way down the lawyer but i don't think it's impossible that a computer will become so complex that it will actually start to feel things and actually have a sense of self. shunts as by how they should to fairly i'm. trying a computer as if the process i was saying i'm not as rigid as you think.
10:28 am
they are is electronic a center and. also known as the museum of the future. artists and its future lab are experimenting with algorithmic composition an area of research recently revolutionized by a neural network called music. trained with music ranging from beethoven to the gaga it can create compositions without having any understanding of the music. simply by recognizing patterns and predicting the next step in a sequence. here it's creating a composition in the style of mozart with a little bit of. by
10:29 am
anyone. with deep learning models neural networks you don't set any rules and down to the machine learns by itself which data works best and that is that it's fascinating that it doesn't have to think about musical structures such as consonants or dissonance. the machine learns completely independently and. more than any other artistic form musical composition is governed by rules. but throughout history scientists have wondered does all art in fact follow rules. even art that elicits an emotional response. this is or even the moment in the act is it not the end that music the saudi national. health island is the that the.
10:30 am
then the the i could not. get the tourist you could argue all these are not there not this in with all that being the visa there was it. is there good luck is if i get the message is there. september 21900. world premiere at the ars electronica festival. the i'm 20 point not work a stroke is performing mahler unfinished a piece of music made by man and machine. the ai system usenet has completed the unfinished movements of gustav monist 10th symphony. as it's annexed what you saw figure we could get for the last was the mother usually not the thief if you 1st the ai system was founded with the complete works of course stuff mama so the dataset contains every fragment of mama sound as well as all the music from the late romantic era of them before. then it was given a theme for
10:31 am
a 5 minute composition written as does build a famous theme in the 1st movement. yes and this. is why the us then they are not common saying. in this work mahler grieves for his wife. ottumwa spelling her initials within the notes of the fi a and e. flat or in german. and. on up into. the machine up you know this is exactly the theme new snatch was given because then it was able to make its own decisions based on face vast data set it was trained with these 4 and they composed a piece of music something completely new company and its mission to my.
10:32 am
sorrow is this model form usenet. it's music. the music is so there is this good food inside and other music is a really good in my opinion or you can identify the theme and then recurring elements from the theme each with a very skillfully developed by the ai model. and you can tell how the game has been deconstructed movie of the listening to tao certain sequences have been given prominent see and think one is the way and how the piece progresses from a gentle pianissimo to a powerful 4 t.c. in the space of just a few minutes emotional it's enough for now coming up. technically
10:33 am
it sounds like mama. but does the music convey a story and intention. whose voice is it worth are we hearing. i mean it's i mean being for $200.00 us women instead of voting for months on vic if we did a blind tast fest if we knew nothing about how these beasts. being composed of trying to divert it's machine made often lime games then would probably be taken in any fire in a wrong note say that is to the rules of voice leading there are intriguing hominids. in technical terms it's astonishing that almost half of the question remains what does it mean when nixon pulled must be thought to discount. the odd can't answer that as it basically the answer is that it means nothing. as worst
10:34 am
as in a severe what's the point and that's the problem when we hear something the immediately searches for meaning or in it's not a name for emotional context in which where does it get us it gets us nowhere it's a simulations newborn this isn't a simulate song. they are uses mathematics and logic to produce results that are astoundingly similar to art. that suggests that the processes involved in making art are actually quite predictable. but predictability is surely the opposite of creativity. which is about finding novelty and different perspectives exploring uncharted territory. intervening.
10:35 am
so what you see here is the movements i just recorded and now this is a continuation of my movements in the style of for the. company so in this case you can see here and any. just in shock that if you look at the story each time there is a new innovation some artists embrace it and make new art with it and i'm the dry person for that but if you ask the artists around this i was surprised that many. really excited by these technologies and he wanted to understand them wanted to see this kind of technology can improve their own. paris home to the headquarters of google arts and culture the tech giants the
10:36 am
nonprofit creative playground. artists are invited to come and experiment here with new algorithms. for now it's all just a bit of fun hardly world changing. first you have to have an age. and. i will start of the 2nd part of the tool which is about on the ring back to. the butterfly and the background merge into an infinite pattern a synthesis of form shape and color. the computer hasn't been told what to do this is all it's own imagination.
10:37 am
the images that this tool makes can visualize a potential of an object and its correlation with its background so it's not just that background cut and paste image but is just much more recourse if and in it for some like it looks like my work. i always created from an instinct and there when i was sitting at the table at google it's just i felt like everything i ever that brought me to that table and helped me to understand what i have been doing. compile the rhythm has no visual ability. it turns images and impulses and impulses into images. no human being could ever devised these infinite variations. all people are demi
10:38 am
doc has to do is select the images she likes best and build upon them i think there's a very big misunderstanding in the role of why humans are born in a sense i do believe that we are born to transcend we are born to be creative we are born to to dream we are born to seeing where i'm. want to make where we are not born to stamp or we are not born to execute robotic tasks so i do believe that ai machine will take away all the pragmatic repetitive dirty dangerous boring part of our creativity so that we can just like the visionaries. the invention of the telescope changed the way humans saw the world. ai could be a similar game changer. clinger man is an internationally celebrated digital artist
10:39 am
. he's fascinated by the virtual world to the algorithms generate. throws the doors of perception why don't. we and one of our concerns he's an algorithm takes $128.00 dimensions and breaks them down to. you know way it allows us to visualize a world that isn't visible to. an algorithm creates a virtual map of all of the works of art google's data bank. but what about the gray areas between the artworks. uses a generative adversarial network again a powerful class of neural networks to visualize how the computer sees that. gets. what about here. and it looks like there are no
10:40 am
images in the database that. so another model had to be found one that doesn't sort data but it generates data and that model could try to give us an idea of what it things is there. yes that's the one dated stays as that it out debit is something exists within the world of these models and if i use again i could generate this image and possibly it would be something we recognized would view that often but 2nd. while the human brain has limitations the computer brain can turn chaos into order. the way it selects in structures is astonishing inspiring. it goes against all human understanding of rules.
10:41 am
we see here are i suppose the creatures from the in-between areas in solution of all the. different they have elements of plants insects animals. does this for the act of the machine doesn't recognize categories such as human and animal plant which is very interesting i did this mentally does is. it takes 2 and it's exam is it so these are merely textures that have been put together. and we could say this is more insect like except for its eyes. and now it's more like a blowfish. cooler a fish. is a slightly weasely rodent like a. tear or a very fluid in a non-spatial that's what's fascinating us. as the most you know it's alice alice
10:42 am
and humans have a very pronounced spatial awareness and that's fine because we humans like clear definitions but we like to pigeonhole and sort everything into groups belongs here it doesn't belong there. and to do. this isn't even. we don't there are clear definitions to be she's obviously she isn't but in fact there are these gray areas where everything is in flux. in a 502. in the infinite data cosmos that is ai everything is possible just as the telescope did when it was invented over 400 years ago it can change our view of the universe and our view of ourselves.
10:43 am
but where does that leave artist's. model clean a mind can instruct the computer to produce a brand new part of. the algorithms generate results on an endless loop. the memories of passing by is a machine that i created that will keep on. generating portraits all from on existing people. for ever.
10:44 am
good. news. i think one of the reasons that we're creative as humans is because we want to leave something behind because we realize our own mortality and you know if you create a piece of music or a painting or in my case a piece of mathematics you know i do feel that that will outlast me and may help to extend. my own consciousness beyond you know my death consciousness allows you to do mental time travel you can see yourself in the future and that involves seeing yourself and your own.
10:45 am
can you read me a poem. on the. on the run in a tremulous night sustenance and if. this life. keeps me busy dealing. the love keeps me busy you're light hearted longing golden limbs wild blood secret. your heavenly charm your evanescence source inspires me you healing creature chased. by peacey. they have things as i'm giddy then me someone sends you such
10:46 am
a poem and asks you what you think they have us do and what do you do money a stand there you are or did you sleep on it but. one penned by. algorithm fed with good turn and. extraordinarily authentic sounding. project developed by the digital creative agency tunnel 23 in vienna. how did i feel so well it was quite a shock let me tell you that. i'm shock. for going to machine write a poem and get dish doesn't stand a bus home that made it into the brentano company's poetry anthologies indian to look it up and turn over there. we look at each other we exchange dark words. we love each other like poppy and recollection.
10:47 am
lines written by a flesh and blood poet powered taillight. words that evoke even more than the sum of their parts. a i could imitate ceylon recognize patterns predict the next line. but with the effect be the same. divided among us all we had i tell now we knew that art was created by humans to go where they lived 10000 years ago when there were still alive today that incident about now you might read something that moves you that perhaps even brings tears to your eyes and said it isn't the work of the human disease leisel got some violent thing less than one 3rd is this an interest in a bad poem even though it puts you in touch with your sadness it done is done oh i know big no mitzi said. input output. perhaps as many scientist say humans are simply highly
10:48 am
sophisticated machines so sophisticated that they have developed self consciousness and ponder their own existence. yet the self remains one of the greatest mysteries of science i think therefore i am. for centuries deckard statement was the cornerstone of western philosophy. but many brain researchers nowadays argue that the sense of self is merely an illusion. the suspect into a community says it's impressive how much this argument has developed but it doesn't alter the fact that we are beings that know very well what it means to have consciousness of abuse was a perceived color it would still experience a feeling such as sadness to all we don't need science or psychology to tell us what it means to feel sadness to be and was a size 12 to some. sadness
10:49 am
is described in binary code as 30 wants and 26 zeros. to human beings could have a long conversation about what sadness means. but for a machine meaning semantics is not a factor it has no consciousness. semantics is the holy grail of ai research. on. current. theory it's. hungry man reach for the book it is a weapon said bet one question. these days with tens of thousands of new books published every year. a book is also a commercial product. publishing houses need bestsellers bam hey i can help find.
10:50 am
an algorithm can calculate a bestseller. status list anguish midgets down a technique midgets concession to against nuclear stuff i thought it must be possible to use tech and artificial intelligence to identify patterns and looks and bestsellers and apply the findings to new works and it's called that naturalized us to figure out for example what your target audience is and what kind of razor sharp a book will have at what currently there are 9 in houses in germany he was in the program including some of the big ones but we've signed confidentiality agreements because they don't want to be named because in this kind of program doesn't make any judgments that it doesn't distinguish between good or bad literature we don't think in those categories everything has validity but for publishers and writers it's also important that it has commercial potential and we need to see if it will find readers or if we want to deliberately pitch it as a nice idea and uses its image to. the program as put to the test
10:51 am
with a manuscript that talked the bestseller lists. but the manuscript is uploaded once about the software reads it and analyzes it is going to see all. the algorithm analyzes factors such as language plot subject suspects and even emotional contact. with the machine was trained to recognize a positive sentence for example i was walking along the street the sun was shining and i was happy it's an issue that's a positive sensitive also it was also trying to recognize a sense expressing negative content and i was absolutely miserable and everything seemed hopeless. for an orphan. it's interesting give me so here are the results yes it's see in what it tells us is that this is a novel with a stick it has elements of a thriller
10:52 am
a suspense crime thriller so it contains excitement of us which i'm going to punch him in. i. was in school for i don't see. it scores 81 percent that's a very good decent men to watch out if it's because she disliked him in other words readers in this market segment of crime fiction suspense novels will enjoy this book immensely again so i. think that's ok it's kind of all of both plus. 6 a computer can take the whole of all victorian literature and you know assimilate it in and off to new and start to see interesting new connections between those which will then help us to be guided through this you know immense library of literature that we just. hope to to navigate on our own as human.
10:53 am
i. should say 650. strong feelings not just love but sadness relationships be tend to be read more by women. the average reader is a woman over $35.00. the book cover is more important than you might think. men often give up after 20 pages women after 50. these are the sort of things that andrew on back notes. he works for
10:54 am
a publisher is testing ebooks out 130-2700 readers figuring out what and how quickly people read as well as when and for how long. we look at the manuscripts like a black box if we don't know what's in it we're only gauging reader response. we collected data on how readers react to the problem is that it takes time it sometimes takes 2 to 3 weeks to gather enough data whereas an algorithm trained with bestseller lists takes about 10 minutes. but what makes a bestseller is not just the actual book but also the cover the blurb designed many factors. so the algorithm hasn't been trained with all the information it needs and anyway write his card right according to a formula tastes and changing all the time for the advantage when you see how
10:55 am
people are reading as i do is that you can factor all that in and if an algorithm is only trained with data from the last 10 years for then that's the only time frame you're working with coleman is to so you can't predict how a trend might evolve in trend. dob it developed at the technical university and constance is a robot that can paint. a machine that lacks consciousness. it won't paint until it switched on ready. if it did have consciousness it might use art as a means of communication. and there we have it. even an algorithm that can self program itself has no curiosity or interest in learning about the world. 'd looking for meaning in what it generates is futile.
10:56 am
for now and for the forseeable future only human beings look for meaning seek ways to express themselves. in ai will remain only a tool towards this and. i see beauty in an eye i see beauty in machine because i see it as a natural next step in the progress of humanity if it's used well and so my work i want to make people see and feel the beauty that i seen and so it's done wit. radical positivity this mommy it's kind of i'm not worried that software and machine learning and digitalization are going to result in a fully mechanized world making this too should vote for you on the contrary we'll have a greater appreciation for everything that software can't replace is the use of business to move for.
10:57 am
a. long ways the last one. this one's cold. to her is why i always look forward to the next image the next to be generated just. to name. her. man the ratio. of her dream.
10:58 am
to some of those white magic in. the human olivier or. they are really any moves. inland. many places not many miles since apparently hydrogen being sold as a clean all round as the energy source of the future many industries are very interested in the sealed to neutral field. but what potential does it really carry . made in germany. 90 minutes on w. children to come to us
10:59 am
from one giant problem i live in no limit to see the 1st year you. gain by reading a little. the change in the lead if you're gonna get it. how will climate change affect us and our children. come slash water. the fight against the coronavirus pandemic. has the rate of infection been developing. measures are being taken. what does the latest research. information and context. coronavirus up to the code. monday to friday on w. if
11:00 am
the w news live from berlin it is international holocaust remembrance day exactly 76 years after the liberation of the nazi concentration camp in auschwitz the world is called on to commemorate the 6000000 jews and others killed in the genocide and to never forget.

33 Views

info Stream Only

Uploaded by TV Archive on