• last year
In the last video, we explored the idea of text-to-image processors and the blurring of lines between fiction and reality. But can AI images be referred to as art? What is art? Why does AI make people afraid of losing their jobs? Rukmini Ravishankar explores.
Transcript
00:00 Artificial intelligence.
00:03 In September 2022, at the Colorado State Fair in the USA,
00:07 Jason Allen won the first prize in the digital category
00:11 for a painting called 'Theatre for the spatial'.
00:14 This sparked an outburst.
00:15 This award, which is relatively unknown outside Colorado,
00:22 would have gone unnoticed, but for Jason Allen.
00:25 The cause of the outrage?
00:27 Allen used Midjourney, an AI program to create the painting.
00:31 Artists called Jason a cheater,
00:32 because most of them used other digital illustrators for their entries.
00:36 So the question is, can an AI-generated image be considered art?
00:40 But before we answer that question, we need to ask, what is art?
00:44 A simple Google search would give you this definition,
00:52 the expression or application of human creative skill and imagination.
00:57 Right there, in the first line, human creative skill and imagination.
01:01 But humans built AI, and AI models learn from art that is created by humans.
01:06 So, do AI-generated images qualify as art?
01:10 In order to get a perspective on the matter,
01:12 I decided to speak to Ravi Kumar Kashi,
01:15 a Bengaluru-based visual artist with over 20 years' experience
01:18 in the fields of painting, sculpting and photography.
01:22 Till the time when AI can go bonkers and start creating on its own,
01:28 there is still a human agency that is involved.
01:32 There are two parts, right?
01:34 You have a human mind, human imagination,
01:37 then you have a tool which will make it happen.
01:40 So here, instead of the conventional tools,
01:44 we're using an AI tool.
01:47 Of course, that is bringing in its own intelligence,
01:51 if you want to call it in a very limited way.
01:54 But still, it's a tool.
01:56 So, I would accept it.
01:58 So, art that is generated by artificial intelligence
02:01 is art in its own right.
02:02 It cannot be compared to traditional art forms as we know them today,
02:06 or art that is created by humans.
02:08 We cannot apply the same principles that we use to judge a work of art
02:11 that is created by humans to AI-generated art.
02:18 But art is more than the medium it employs,
02:21 whether it is a painting or AI-generated art.
02:24 I do feel that there are certain things that need to happen
02:33 to make something a work of art.
02:34 And I think a large part of that is to sort of, in some ways,
02:41 push boundaries of what has existed till now.
02:45 Pushing boundaries to make people have an experience, an encounter,
02:49 that they probably haven't had earlier.
02:51 So, that makes them sort of think and reflect on the artwork
02:56 and what it's trying to communicate.
02:57 So, of course, every image generated by AI cannot be called art,
03:09 just as every photograph ever taken cannot be called a lens-based artwork.
03:13 I think it's more of these social sharing channels
03:18 which have started calling everything that is made
03:22 with a prompt-to-image-based generation system, like #AIart,
03:28 which I think is a bit problematic
03:29 because it then loses the nuance of what a work of art means.
03:39 It's been made quite clear to me that whether something is art or not
03:43 simply doesn't depend on the medium used to generate it.
03:46 So, why did Jason Allen's Colorado State Fair prize cause such an outburst?
03:51 Because there's a key difference.
03:53 Learning an art form is an emotional experience.
03:55 It takes years of passionate hard work for an artist to develop their skill.
04:02 AI is a machine.
04:07 It doesn't possess the ability for great passion or emotion.
04:10 Moreover, it produces images within seconds,
04:13 while a human would take years to develop skill and refine it
04:16 and produce something they find acceptable.
04:18 What generative AI has done is it has made visual expression easily accessible.
04:26 In the previous video, I showed you what I drew based on a text prompt
04:30 versus what a few minutes on an AI tool could do.
04:36 This process, which continues to amaze all of us, also produces fear.
04:40 One worldwide concern is that AI could soon replace all of us by taking our jobs.
04:45 The other is that the cornerstone of human expression, art, literature,
04:50 could soon be taken over by AI as well.
04:52 Art is sort of that deeper engagement with a certain subject
04:56 that really lets you create a work of art.
04:59 I think that is still very much relevant.
05:02 That aside, there are still a couple of very real concerns
05:05 that the emergence of AI image generation raises.
05:08 In the last video, we spoke about datasets.
05:11 A dataset contains millions of images that AI models refer to and learn.
05:16 The problem is that a lot of AI models learn from images
05:19 that might be sourced from the internet,
05:21 or they could learn from famous pieces of art.
05:24 AI tools do not discriminate when reproducing an image
05:27 when you or I give it a prompt.
05:28 So, potentially, I could ask an AI model to generate an image
05:32 that's in the style of Van Gogh.
05:33 But Van Gogh spent several years developing his style,
05:37 and now I can produce it simply by using his name in my prompt.
05:40 Or even more worryingly,
05:42 I could ask for an image with Starry Night as the prompt,
05:45 and the result could very well end up with parts from the Dutchman's famous art.
05:48 There are many tools which can make your image look like Van Gogh's image.
05:55 He used to paint in a certain very staccato kind of colour application.
05:59 So, they're bringing that surficial level of treatment.
06:05 So, it cannot really replicate what the kind of response that Van Gogh had
06:14 looking at a flower or a chair or a landscape.
06:17 That is not possible.
06:19 In the last video, I tried to compare how a human learns an art
06:22 as opposed to how a machine does.
06:24 And that comparison made it very clear to me that the two processes are similar.
06:28 We learn by observing examples,
06:30 and our style is an amalgam of everything we have observed.
06:33 I'm doing a certain kind of work because I'm living in this moment.
06:42 Because I've seen what has happened before me.
06:44 I may not copy-copy them, but I have kind of digested the earlier artists.
06:56 I respond in a sense. Of course, I transform, I change and all that.
07:00 But still, I'm sitting on their shoulder to look at the world in a different way.
07:06 So, if I cannot be accused of creating an artwork that is inspired by everything
07:10 I have seen and digested over the years,
07:12 how can an AI tool be accused of this?
07:15 Human artists have data sets too.
07:17 They're just not as defined.
07:19 Regardless, contemporary artists are affected by the potency of AI art
07:23 to create something they could potentially lose money to.
07:25 Copyright laws are still to catch up with AI.
07:28 To the artists, I would say that if it's possible,
07:31 just avoid using names specially of your contemporary peers.
07:37 Because they haven't really given consent to being used in the training of AI algorithms.
07:44 Let's say the tech platforms were actually putting these systems out.
07:48 I think one way for them to look at this is to avoid, as much as possible,
07:54 using artworks of artists, specially artists who are currently continuing to practice.
08:01 The other problem with AI data sets is that they contain biases.
08:06 That if I were to search for a soft, kind human being, the images are of women.
08:15 Or if I were to search for a photo of a married couple on their wedding day,
08:23 they're all heterosexual, white couples.
08:27 So can we use this in some ways as an opportunity to also,
08:32 you know, more meaningfully and more consciously create ecosystems
08:39 that allow for creation of data sets and trained AI algorithms
08:43 that have more representation of different populations,
08:47 different beliefs, etc. included in them?
08:52 We don't live in a world where artificial intelligence has taken over
08:55 and is working on its own agency.
08:57 We're still the ones using AI as a tool to bring our vision to reality.
09:01 And so it is upon us to engage with AI in a manner that furthers this evolution
09:06 and helps us do things that would have proven impossible a few years ago.
09:10 In this video, I tried to explore the connection between human creativity and AI art.
09:15 I tried to address some of the ethical and moral concerns that the emergence of AI raises.
09:19 And we heard both sides, traditional art forms as well as AI art.
09:24 In the next video, we'll be exploring another form of AI art generation called Deepfakes.
09:28 This is a technology through which anyone, anywhere can make anyone say anything.
09:34 No, I buy you out, you don't buy me out.
09:36 And it has the potential to be incredibly dangerous.
09:39 Subscribe to make sure you don't miss it.
09:42 (upbeat music)
09:44 (music fades)
09:46 (music fades)
09:48 (music fades)
09:50 [BLANK_AUDIO]

Recommended