Before we begin, and in case some of you were wondering:
Early last year, San Francisco-based artificial intelligence company OpenAI launched an AI system that could generate a realistic image from the description of the scene or object and called it DALL.E. The text-to-image generator’s name was a portmanteau coined after combining the artist Salvador Dali and the robot WALL.E from the Pixar film of the same name.
https://analyticsindiamag.com/whats-the-big-deal-about-dall-e-2/
Dall-E 2 is amazing. There are ethical issues and considerations, sure, but the output form this AI system is stunning:
And just in case it isn’t clear yet, no such painting/drawing/art existed until this very sentence, the one that is the caption, was fed to the AI. And it is the AI that “created’ this image. Go through the entire thread.
This has led, as might be expected, to a lot of wondering about whether artists are going to be out of a job, and the threats of AI to humanity at large. I do not know enough to be able to offer an opinion one way or the other where the latter is concerned, but I do, as an economist, have some points to make about the former.
These thoughts were inspired by reading Ben Thompson’s latest (freely available) essay on Dall-E 2, titled “DALL-E, the Metaverse, and Zero Marginal Content“. He excerpts from the OpenAI website in his essay, and this sentence stood out:
DALL-E is an example of how imaginative humans and clever systems can work together to make new things, amplifying our creative potential.
https://openai.com/dall-e-2/
And that begs an age-old question where economists are concerned: is technology a complement to human effort, or a substitute for it? The creators of Dall-E 2 seem to agree with Steve Jobs, and think that the AI is very much a complement to human ingenuity, and not a substitute for it.
I’m not so sure myself. For example: is Coursera for Campus a complement to my teaching or a substitute for it? There are many factors that will decide the answer to this question, including quality, price and convenience among others, and complementarity today may well end up being substitutability tomorrow. If this isn’t clear, think about it this way: cars and drivers were complementary goods for decades, but today, is a self-driving car a complement or a substitute where a driver is concerned?
But for the moment, I agree: this is an exciting new way to generate content, and is likely to work best when used as a complement by artists. Note that this is based on what I’ve seen and read – I have not myself had a chance to use or play around with Dall-E 2.
The title of today’s blog post is about substitutes and complements, which we just finished talking about in the previous section, but it also includes references to demand and supply. What about demand and supply?
Well, Ben Thompson talks about ways to think about social media firms today. He asks us to think about Facebook for example, and asks us to reflect upon where the demand and the supply for Facebook as a service comes from.
Here’s my understanding, for having read Ben Thompson’s essay: Facebook’s demand comes from folks like you and I wanting to find out what, well, folks like you and I are up to. What are our friends, our neighbors, our colleagues and our acquaintances up to? What are their friends, neighbors, colleagues and acquaintances up to? That’s the demand.
What about the supply? Well, that’s what makes Facebook such a revolutionary company – or at least, made it revolutionary back then. The supply, as it turns out, also came from folks like you and I. We were (and are) each others friends, neighbors, colleagues and acquaintances. Our News Feed was mostly driven by us in terms of demand, and driven by us in terms of supply. Augmented by related stuff, and by our likes and dislikes, and news sources we follow and all that, but demand and supply comes from our own networks.
TikTok, Thompson says, is also a social network, and supply and demand is also user driven, but it’s not people like us that create supply. It is just, well, people. TikTok “learns” what kind of videos we like to see, and the algorithm is optimized for what we like to see, regardless of who has created it.
But neither Facebook nor TikTok are in the business of generating content for us to see. The former, to reiterate, shows us stuff that our network has created or liked, while the latter shows us stuff that it thinks we will like, regardless of who has created it.
But how long, Ben Thompson’s essay asks, before AI figures out how to create not just pictures, but entire videos. And when I say videos, not just deep fakes, which already exist, but eerily accurate videos with depth, walkthroughs, nuance, shifting timelines and all the rest of it.
Sounds far-fetched?
Well, I remember taking an hour to download just one song twenty years ago, and I can now stream any song in the world on demand. And soon (already?) I will be able to “create” any song that I like, by specifying mood, genre, and the kind of lyrics I want.
How long before I can ask AI to create a movie just for me? Or just me and my wife? Or a cartoon flick involving me and my daughter? How long, in other words, before my family’s demand for entertainment is created by an AI, and the supply comes from that AI being able to tap into our personal photo/video collection and make up a movie involving us as cartoon characters?
Millions of households, cosily ensconced in our homes on Saturday night, watching movies involving us in whatever scenario we like. For homework, read The Secret Life of Walter Mitty by Thurber (the short story, please, not the movie!), Snowcrash by Neal Stephenson, and The Seven Basic Plots by Baker.
There are many tantalizing questions that arise from thinking about this, and I’m sure some have struck you too. But I don’t want to get into any of them right now.
Today’s blog post has a very specific point: it doesn’t matter how complicated the issue at hand is. Simple concepts and principles can go a very long way in helping you frame the relevant questions required for analysis. Answering them won’t be easy, as in this case, but hey, asking (some of) the right questions is a great place to start.