Monday, April 18, 2022

DALL-E

Ben Thompson (Hacker News):

Last week OpenAI released DALL-E 2, which produces (or edits) images based on textual prompts; this Twitter thread from @BecomingCritter has a whole host of example output[…]

[…]

[C]reating games, particularly their art, is expensive, and the expense increases the more immersive the experience is. Social media, on the other hand, is cheap because it uses user-generated content, but that content is generally stuck on more basic mediums — text, pictures, and only recently video. Of course that content doesn’t necessarily need to be limited to your network — an algorithm can deliver anything on the network to any user.

What is fascinating about DALL-E is that it points to a future where these three trends can be combined. DALL-E, at the end of the day, is ultimately a product of human-generated content, just like its GPT-3 cousin. The latter, of course, is about text, while DALL-E is about images. Notice, though, that progression from text to images; it follows that machine learning-generated video is next. This will likely take several years, of course; video is a much more difficult problem, and responsive 3D environments more difficult yet, but this is a path the industry has trod before[…]

[…]

Machine learning generated content is just the next step beyond TikTok: instead of pulling content from anywhere on the network, GPT and DALL-E and other similar models generate new content from content, at zero marginal cost. This is how the economics of the metaverse will ultimately make sense: virtual worlds needs virtual content created at virtually zero cost, fully customizable to the individual.

Bram Adams (via Hacker News):

DALL·E is intuivitely understandable on an emotional level while simultaneously being quite unintuitive on a logical level.

[…]

Here are some things I’ve picked up so far that I think can help push the dialogue around DALL·E forward.

Previously:

Update (2022-04-21): Mike Rundle:

When humanity has access to a machine that can render anything you can possibly think of, the creative side of art will be how to describe such an image in a way that connects to the soul of the AI and how it understands the world.

Update (2022-10-07): Matt Sephton:

Each have their pros/cons. In my usage DALL-E could render a Moai 🗿 exactly but had no idea any the style I was asking for, Midjourney gave a rough Moai but had exact style, Stable Diffusion gets both. Personally, I prefer the output of Midjourney

9 Comments RSS · Twitter

Old Unix Geek

I find this conversion of talent to capital concerning.

These algorithms take the output of those who have worked to excel at their art, and replace the need to train future people to have the same level of excellence, by vast arrays of computers learning from those old masters.

This argument applies to artistic endeavors but also to coding, science, engineering, etc.

Why would people continue bothering to make the effort to excel? As it is, our time is marked by many people claiming merit should not be used to determine employment. Instead, they claim, extrinsic characteristics should be considered, that jobs have not intrinsic worth but are simply expressions of status that can be done by anyone.

The problem is that civilizations depend on a small percentage of very bright and usually driven people to advance. But if the majority is not striving to better themselves, because merit no longer matters, the opportunity for these people to gain training commensurate to their ability will fall.

Once meritocracy dies, upward mobility becomes a lottery ticket. And at that point, the rich will work hard to ensure their spawn wins that lottery ticket, despite the guaranteed reversal to the mean... as it is, they are the main impediment to true meritocracy, improving the chances of their genetic line to the detriment of the civilization in which they live.

There is a good chance that if this type of development becomes widespread, it will further collapse our civilization. Introducing this now might not be wise. I have some doubts that we've yet adapted to the widespread availability of social media. This penetration of AI into creative pursuits seems to be an even bigger deal to me.

What is the point of life, if you can't challenge yourself to succeed at something? Do we truly want our descendants to become serfs without options as most of our ancestors were?

I take issue with the implication that this “creates art”. At best, it remixes existing art, but even that is dubious, as even a remix usually has a message. What does DALL-E have to say about society? Nothing, of course; it’s a machine.

It creates images, and those images may still have sufficient usefulness for bland, generic wall decor that says everything yet simultaneously doesn’t say anything at all. It’s perfect for corporate environments.

That’s not art. If I generate a thousand pieces of music with an algorithm and then focus-test which one generates the most likes, what does the “art” I’ve created even mean? (See also: NFTs.)

There are edge cases, of course. Is No Man’s Sky art? Its planets are largely procedurally generated. But it still has enough manual human involvement, and a plot (I can’t comment on how deep it is).

@Sören:

What does a paintbrush have to say about society? Nothing, of course; it's just hairs on a stick. But the artist uses it to create an image that conveys a message. Similarly, the phrase that a person gives to DALL-E might have been chosen to create an image that conveys a message. In other words, DALL-E is just a tool, but so is everything an artist uses, and so I think it can be used to create art, given creative inputs.

In the Twitter thread, many of the examples look like art to me. The elephant tea party is just delightful, and the teddy bears using computers on the moon is such an odd juxtaposition of concepts that you can't claim it's not creative. These do not strike me as "bland generic wall decor for corporate environments". Also notice that many of the input phrases specify the style of the resulting output, such as photograph, oil painting, etc., giving another dimension of control over the tone of the work.

So, I think that a creative person with something to say could use DALL-E to bring their vision to life if they otherwise lack the necessary drawing/painting skills to do it themselves. Of course it can also be used to churn out bland uninspired images; it's up to the person as to how it will be used.

@ Doodpants I basically share OUG's concern that the intent (and/or unintended possible outcome) here is to substitute a creative worker's job with a machine, not to empower creatives with additional tools.

"The problem is that civilizations depend on a small percentage of very bright and usually driven people to advance. But if the majority is not striving to better themselves, because merit no longer matters"

This is not how human psychology works for the vast majority of people. I think Star Trek is essentially correct: if you take away people's *need* to work, but still create a society that gives them the opportunity to do so, we will see people succeed and excel who are now stuck working in an Amazon warehouse.

People *want* to better themselves, but our current society prevents many from doing so.

"Once meritocracy dies, upward mobility becomes a lottery ticket"

Can't tell if sarcasm.

I don't personally think any of the examples qualify as art. But I can definitely see a plethora of use cases for this. Illustrations for blogposts, newsletters, farmers markets, local theatre troupe posters etc etc.

Or imagine this in the hands of a person who wants to make a graphic novel but can't draw. Or sell t-shirts and dorm room posters.

Yes, it will take work from illustrators for sure. But it will also create new pieces of culture, and sometimes even art. I mean, now that Moxie Malrinspike has created an interesting art piece using nft technology, then the sky is the limit.

But first we need to put the means of production in the hands of the people.

Old Unix Geek

@Plume: not sarcasm, but meritocracy is weaker than it was. That's not saying it was uniformly distributed before, but within a certain sphere merit did matter. Today, not so much.

As to your Star Trek comment, that's true for some of us. I used to believe it was true for all of us. I'm a lot less certain of that these days. I also don't think there is any evidence that we are moving towards that future. Unfortunately, the society described by "The Expanse" seems much a more probable successor of our economic system.

"meritocracy is weaker than it was"

I genuinely do not believe this to be true. It's different from the past, but weaker? That's probably confirmation bias, but I don't see any evidence of that.

(I don't count it as evidence that white men are whining about the fact that non-white non-men are now succeeding more than they did in the past, because that argument is premised on the idea that non-white non-men succeeded less in the past because they were incompetent, and are succeeding more now simply because companies hire incompetent people, which is an incredibly implausible assertion.)

"I also don't think there is any evidence that we are moving towards that future."

Correct me if I'm wrong, but I thought the premise of your argument was that there was no reason for humans to apply themselves to anything, since AI algorithms would just take their work as training data and then do an adequate job at whatever the human was doing. This is pretty much a description of a post-scarcity society, which isn't what The Expanse depicts. It is, however, what Star Trek depicts.

Star Trek does have one thing that we might not get to, which is a society that values people who achieve unnecessary, but interesting things, such as exploring space. Perhaps that's not going to happen, and people will, once they have achieved universal wealth, just kill each other out of sheer boredom. That's probably an equally plausible outcome.

The third option is that we will create artificial scarcity (such as NFTs) and just, as a people, create pretend-work for ourselves, and needlessly let people die out of apathy.

Actually, now that I think about it, that's the system we already live in.

Old Unix Geek

@Plume:

Just two examples.

The rate of discovery has collapsed since the 1960s. There's lots of PR, but very few earth shattering developments. The only real innovation in technological growth has been lithography. Everything else, including the current buzzwords, was described in the 60s. They just didn't have the computational power to test out their ideas. Similarly Physics has essentially stalled since the golden age of the early 20th century.

Second example: school exams. The current STEM exams are a joke. I'd expect the level to go up with technological sophistication, not down. Few youngsters would be able to maintain low level software for instance, but that stuff runs the world. At the rate we're going, the West will not be able to maintain its technological superiority much longer, but the places that actually still train their students such as China have cultural impediments to crazy new ideas.

I expect AI to replace a lot of workers. Most people will live on Basic Income. It's "post scarcity" in the sense people will be fed. It's "scarcity" in the sense that there's a scarcity of jobs, and basic income is bad. As to the "lucky few", many of them will be the spawn of rich parents. Others might be chosen by lottery to get an opportunity as in the Expanse, or more likely, will be skilled to some degree. But these people will be not trying to rock the boat, not trying to innovate, not trying to lose their position and fall back to basic. The overall IQ of society will fall. Personally, I think education should be free to all, but the level required to enter into a university should be extremely challenging. Give everyone a chance, but ensure that those who graduate actually are experts.

In the Expanse, those who are skilled and want out, ended up in Space/on Mars. That's because most livable places on our planet are now inhabited. They had to pull the trick a second time with the rings. It's an allegory for the history of Europe. Those who were suffocated by the status quo moved to their Empire's colonies to make their fortune, which is why there are light skinned people all over the planet, even though light skin is a mutation to cope with a diet high in vegetables while living far north.

Leave a Comment