Music Playlist: Spotify, Apple Music

Transcript

PART 4: ARTIFICIAL CREATIVITY

In 2015 artificial intelligences started making images based on nothing but text input. This was basically like reverse engineering photo captions. The results were very low quality, but that it worked at all was stunning.

By 2021 AI image generation was doing things like this.

It's better art but still mediocre at best. What was historic was that the AI combined ideas together in a variety of ways. The AI seemed to exhibit... *creativity.*

2022 was one of the most whip-lash transitions of the modern technology era. There were now several image generators and they were making images like these. No human hand drew a stroke here. All these were created with nothing but text prompts.

Creating a sophisticated illustration was suddenly as easy as typing a word or two.

Enter JOY and you get this.

Enter CAT and you get this.

Enter BRAINY EYEBALLS and you get this.

Then you can combine words together for infinite variation.

AI-created art is no longer cute and clumsy. It still has weaknesses, like human anatomy, especially hands, and it mostly lacks the expressiveness and the storytelling of real artists, but AIs are *creating* *art*. And they are doing it with beauty, with stunning versatility, and even with subtlety.

AI has had similar breakthroughs in text generation and coding, but it's AI art that sparked the fiercest debate and generated anger and fear in the art community.

Fran Blanche: I can't imagine that there's any writer or artist on the planet right now that isn't really thinking about this and wondering where they're going to be in five years.

This anxiety was triggered by a profound development in human history. Machines have breached a sacred realm we thought was solely the domain of people.

The first battleground of The Age of AI is art.

Will AI replace human artists? Is AI image generation ethical? Will the future of creativity be ruled by AIs?

In this final episode of Everything is a Remix, we venture into the newly emerging field of Artificial Creativity.

Title: Part 4: Artificial Creativity

Let's begin by addressing the most common emotional reaction to artificial intelligence: fear.

Storytellers have long warned us about the seduction and the danger of technology. The Greek titan Prometheus stole fire from the gods and was brutally punished by Zeus.

In Mary Shelley's Frankenstein, which was subtitled The Modern Prometheus, Dr. Frankenstein is obsessed with uncovering the secret to life.

Frankenstein creates a man, but is horrified by his creation, who then seeks violent revenge.

Stories like these are a warning about meddling with the sacred and unknowable. They're a warning about arrogance.

"Frankenstein" clip: It's alive, it's alive! In the name of God! Now I know what it feels like to be God.

In recent decades, the subject of these tales has taken on a particular form: the computer.

HAL 9000, a prescient imagining of a computer assistant, was one of the first popular fictional computers.

"2001" Clip: I know that you and Frank were planning to disconnect me, and I'm afraid that's something I cannot allow to happen.

HAL ultimately decides to sacrifice its crew for the sake of its mission.

"The Terminator" films feature a powerful defense network AI called SkyNet.

"The Terminator" Clip: They say it got smart. A new order of intelligence sighted after eight microsecond extermination.

Our dream of technological progress has reached a nightmare conclusion.

"Avengers: Age of Ultron" Clip: Everyone creates the thing they dread.

We are now imagining the day when we are supplanted by our creations.

"Ex Machina" Clip: One day the AIs will look back on us that same way we look at fossil skeletons in the plains of Africa. An upright ape, living in dust, with crude language and tools, all set for extinction.

The topic of human extinction by AI is no longer limited to science fiction. It's popularly discussed by intellectuals.

Yuval Noah Harari: We are probably one of the last generations of Homo Sapiens. In a century or two at most, I guess, that humans like you and me will disappear and Earth will be dominated by very different kind of beings or entities.

Many of the leaders of the field of artificial intelligence claim the time when our creations will match us is rapidly approaching. Some think human level intelligence, known as Artificial General Intelligence, or AGI, will be reached within a couple decades.

Demis Hassabis: I think that it's coming relatively soon in the next, I wouldn't be surprised by the next decade or two.

After AGI comes an "intelligence explosion," with AI rapidly improving itself and spawning superintelligence. Humanity will then be the parents of... gods.

The belief that AI will soon surpass us and take our place is widely held among many brilliant people. So, why not believe them?

Because similarly brilliant people have been making similar predictions for as long as there has been artificial intelligence and they have all been... wrong.

Many people in AI fall into the same old trap that true believers always fall into. They think the great whatever is almost here, I swear it's just about to happen.

Let's take a brief tour of AI's many failed prophecies.

Many of the pioneers of artificial intelligence predicted that machines would attain human-level intelligence by about the 1980s.

More recent predictions have been just as wrong.

Shane Legg, cofounder of Google DeepMind, said in in 2008: "Human level AI will be passed in the mid-2020s.""

It's 2023 right now and I think we can safely say... no.

In 2015 Facebook CEO Mark Zuckerberg said their goal by 2025 was to "get better than human level at all of the primary human senses: vision, hearing, language, general cognition.”

This one is looking like no, no, no and no.

An engineer even claimed a Google chatbot was sentient in 2022.

Blake Lemoine: In order to be capable of convincingly arguing that you are sentient requires sentience.

I have no idea why this is supposed to make sense.

One of the most prolific and optimistic forecasters is Ray Kurzweill. He's spent decades predicting the arrival of "the singularity," which entails AGI among other things. And his date for arrival of AGI is rapidly approaching.

Ray Kurzweill: I've set the date 2029. A machine, an AI, will be able to match human intelligence and go beyond it.

I'd like to get in on the prediction fun too, so I'll say AGI in 2029 is exponentially wrong.

Of course, there are plenty of people in AI who believe AGI is nowhere in sight.

Erik J. Larson, author the "The Myth of Artificial Intelligence, "argues that current AI technologies are not going to lead to AGI.

Eric J Larson: Any foreseeable extension of the capabilities that we currently have do not result in general intelligence. Just point blank. They just don't.

Oren Etzioni, an esteemed figure in the field of AI, flatly states that we have no idea when AGI is coming.

Oren Etzioni: My answer to when is, take your estimate, double it, triple it, quadruple it — that's when.

Matter of fact, expert projections on the arrival of AGI range from now... to never. Translation: they don't know.

And here's an unpopular opinion we might want to ponder: maybe human level artificial intelligence is impossible, maybe human level artificial intelligence. Maybe the universe can do things we can't.

We don't know when – or even if – AIs will match human intelligence. It's unlikely they'll murder us anytime soon.

But there is something they want to murder now: your job. And they don't need anywhere near human-level intelligence to do it.

This is why illustrators are so upset. They are the first to suffer what's called "creative destruction". Old jobs are eliminated by new technologies and ideas, resulting in lost livelihoods and real pain.

"Upgrade" Clip: When you look at that widget, you see the future. I look at that thing, I see ten guys on an unemployment line.

However, this also leads to increased productivity and fresh growth.

Automation is now expanding beyond the domain of muscles and entering the domain of the mind. It has crossed over into arts and expression.

But actually, this isn't quite new either. Specialists have been getting replaced for decades... without AI. Let's go back to hip hop.

With the birth of rap music, suddenly you didn't need to play an instrument, didn't need to know anything about music, didn't even need to sing. If you had a turntable, a drum machine and a mic, you could make the most exciting music around.

And this trend has only accelerated. Anybody with a laptop and a bit of music software has tools that would have seemed like science fiction to early DJs like Grandmaster Flash.

And this is more than just music. Anyone can now easily build a website or build an app or launch a shop or shoot gorgeous photos or shoot gorgeous videos.

Art has been getting cheaper, faster and easier since the printing press, which creatively destroyed an entire class of monks who painstakingly hand-copied books with quill and parchment.

If machines can make images as well as we can, then why shouldn't they? What's the issue?

The issue is *how* the machines learned to create images. Let's put image generation on trial and determine if it's guilty or not guilty of crimes against creativity.

Here's the evidence.

The simple version of what the AIs did is this: it studied countless images, without permission, then it emulated them and created its own versions.

So yes, this is just like you. The entirety of this series demonstrates that this is how we all create.

But... it's more complicated than this. Let's zoom in.

Image generation has three steps. I'll explain each, and all of these need to be ethical.

Step 1. Tons and tons of images were scraped from the internet. These images are called a training set.

It looks like a mountain of junk. If you found a folder of this stuff on your hard drive, you would immediately throw it out.

Step 1 is just obtaining a zillion images from the internet. Step 1 is ethical. Search engines do the same thing. And you can go download as many images as you want right now.

Step 2, the AI processes the images and creates a model. This is their version of studying the images and learning from them. Y'know what, this ain't simple. I'll come back to this.

Step 3 is open and shut: the AI processes requests from users, which are written prompts, and creates images. If someone just wrote a program that can draw, that would be fine. Step 3 is indisputably ethical.

It all comes down to Step 2. This is the tricky bit. What the AIs did with copyrighted images is called diffusion. Noise was added to the images over many steps until they're just noise. Then it runs this process in reverse, with the goal of creating a new image with the same meaning. The cat should be a cat, not an identical cat, but a cat.

I have no idea why this works either. But somehow it does.

If diffusion is copying then AI image generation is copyright infringement. Is diffusion copying?

On the one hand, it's kinda like copying because it reproduces the watermarks from stock photos. On the other hand, it's pretty bad at it... so it sorta made something new?

The clear conclusion... is that is unclear. That's why this topic is so controversial. It is truly ambiguous. This is like the dress controversy all over again, except furious.

My guess is that this ambiguity will result in diffusion being considered fair use. It will be hard to definitively prove that it's copying because... this stuff is complicated.

This is going to multiple courts, we'll learn a lot, and we will get an answer.

Let's just assume for now that diffusion is kinda-like copying but not totally copying. And let's take a swing at the most important question of all. Is it ethically right – or at least acceptable – that artists' images were used without consent?

Megan Rose Ruiz: It seems like it's a pretty general consensus in our community that we do not want our work to be used to train A.I. models.

I am sympathetic to how artists are feeling, but it does seem acceptable to me.

For starters, most of the training images are pretty generic and in this context, they seem public domain. Sure, this might be your photo of a pretty girl or a dog or a quesadilla but it's very similar to thousands of others. Nobody owns the idea of these images and that's really what's getting emulated.

The biggest controversy is over a small minority of the images. These are artwork by professional artists and serious amateurs.

Let's get this clear up front. No artist owns their art entirely. If you don't believe me, here's the artist Scott Christian Sava saying the same thing.

Scott Christian Sava: My art is a mosaic, an amalgamation of the art and artists that inspired me on my journey to become the artist I am today.

The collective achievements of art belong to everyone. They are as free as the air.

Too many artists are getting overly possessive about what they believe is theirs. This artist went viral claiming their art was used to train an AI model.

Deb JJ Lee: AI art is theft. It is an awful, awful way to just, like, steal from artists. It's evil. And if you use AI art, you are dead to me.

They based this on images like these, but the only similarities are the color palette and the basic composition. They're otherwise very different, like for instance, this is trash and this is good.

Yes, there is some piracy going on in AI image generation. There's some piracy going on everywhere. I'm doing piracy right now and you're watching me.

There are plenty of caveats. Training AIs on individuals artists' work does seem wrong. Everyone should be able to opt out of all training sets. And maybe AIs should simply not train on images from active art communities.

Also, some company should make an image generator trained on public domain and licensed images, which would avoid this hornets' nest entirely. Somebody please do this.

But for me, I don't see deep injustice here. In sum, AI image generation seems not guilty.

How disruptive AI art will actually be is not yet clear, but it will definitely have some sort of role. Artists are going to have to adapt.

And the rest of us should take note. If you think what's happening to a bunch of illustrators doesn't concern you, think again. The fear and anxiety the art community feels is going to spread. Many of us will have to adapt. Any mind work that can get automated will get automated.

Blue collar workers have been living this for decades. Now it's white collar workers turn.

Of all humanity's technological advances, artificial intelligence is the most morally ambiguous from inception. It has the potential to create either a utopia or a dystopia. Which reality will we get?

Just like everybody else, I do not know what's coming. But it seems likely that these visions of our imminent demise will someday seem campy and naïve – because our imaginings of the future always become campy and naïve.

AIs will not be dominating creativity because AIs do not innovate. They synthesize what we already know. AI is derivative by design and it is inventive by chance.

Computers can now create but *they are not creative.* To be creative you need to have some awareness, some understanding of what you're doing. AIs know nothing whatsoever about the images and words they generate.

Most crucially, AIs have no comprehension of the essence of art: living. AIs don't know what it's like to be a child. To grow-up. To fall in love. To fall in lust. To be angry. To fight. To forgive. To be a parent. To age. To lose your parents. To get sick. To face death.

This is what human expression is about. Art and creativity are bound to living, to feeling.

Art is the voice of a person. And whenever AI art is anything more than aesthetically pleasing, it's not because of what the AI did. It's because of what a person did.

Art is by humans, for humans.

In some videos about AI, the big reveal is that *this video* was actually made by AI. But this video and this series is the opposite: nothing has been AI. Except where I cited AI art, this is entirely human made.

The words are all my mine, but they're merged from the thoughts of countless people. Everything you've seen and heard is from real filmmakers and musicians and game developers and other artists.

All these thoughts and all this media were remixed by me into something new. And yes, I did it it all without permission.

“Everything is a Remix” is a testament to the brilliance and beauty of human creativity. In particular, it's a testament to collective creativity. Human genius is not individual. It is shared.

You, my dear viewer, are a human – the best technology there is, with no successor in sight. The future is ours. And it will be won or lost by us brilliant, stupid, horrible, beautiful humans.

But there is only one complete certainty about what's coming: AI will not stop. And we need the help that artificial intelligence can potentially bring to the complex problems of the 21st century.

We are saying goodbye to the old world and entering a new one. But we are not obligated to accept this new world as is. Our duty is to make it the best it can be, to make this revolution better than the last one.

We are launching into the unimaginable. As we always are. We are always hurtling into some inconceivable future. There is no other way to move forward. So... here we go.