But is it Art?

shutterstock_2199097431
shutterstock_2199097431

At the Colorado State Fair, artist Jason Allen won first prize in the “Digital Art/Digitally-Manipulated Photography” category by using an artificial intelligence art creation program called Midjourney (midjourney.com). Alas, the term “artificial intelligence” has become so broad as to be meaningless these days. Programs like Midjourney (DALL-E is another which has gained recognition) take a text prompt (“the moon over a quiet lake,” for example) and based on the images, which have been used to train them, generate a brand-new original image.

The prize-winning image, entitled “Théâtre D’opéra Spatial,” is quite something. (See it here: tinyurl.com/2jqcb8w2). And of course, it sent shock waves through the art community, just as photography—the chemical kind—caused an uproar when it was first introduced.

The difference here is these programs are trained by looking at countless images available on the internet that were originally created by human artists. It’s the digital version of Steve Jobs’ line: “Good artists copy, great artists steal.” (He attributed this line to Picasso, but that attribution is unclear). Given the text prompt “a starry night,” would Midjourney turn out something better than Van Gogh’s masterwork?

It’s interesting to note that the two judges in the contest didn’t know the winning entry was created with the assistance of software, though they commented that it wouldn’t have changed their decision when told of the fact. And Allen did have a role in the work’s creation: first and foremost, he created the text prompt, which generated the final work. Asked about it by online publication Motherboard, “Allen alluded to a “special prompt” that he has used to create hundreds of images, which he ultimately refined to the one that was submitted to the state fair.”

You can give Midjourney a try on their Discord channel (discord.com/midjourney)l, which allows you to send a limited number of prompts to the program, which is still in beta.

While no one is ascribing sentience to Midjourney, this story falls under the category of “software which appears to have human abilities.” Like Google’s LaMD (Language Model for Dialogue Applications) which a former Google engineer claimed was sentient, and GPT-3 (from the makers of DALL-E) which produces human-like text, we’ve gone past Turing’s “imitation game”—it’s no longer possible to distinguish between the creative output of a human being and that of a machine that has been trained on such human outputs.

In the New York Times, Allen empathized with artists who were scared that A.I. tools would put them out of work. But he said their anger should be directed not at individuals who use DALL-E 2 or Midjourney to make art, but at companies that choose to replace human artists with A.I. tools.” Should I be worried that Northbay biz will replace me with GPT-3? (Perhaps trained on my 20-plus years of writing columns, so it can mimic my idiosyncrasies?) Having seen the output of GPT-3 (here’s an example: tinyurl.com/y7m4nf33), I’m not worried just yet. Although I may try to slide a column written by GPT-3 past my unsuspecting editor one of these days…

The bigger concern is that most people aren’t equipped to deal with this technology. Hell, people can’t even reliably discern the truth in text written by human beings: a large number of Americans appear to believe that Trump won the 2020 election (according to a Monmouth University poll in 2021, 32% of respondents said that President Joe Biden only won the 2020 election “due to voter fraud”). On Facebook, someone I know just posted a video of the moon with a semi-plausible description of how it “only happened once a year in the Arctic.” Fifteen seconds of research showed it to be a made-up description of some computer-generated imagery from a Ukrainian artist. But that video has been shared widely, racking up millions of views on Facebook and Twitter. Even worse, politicians with a bare understanding of the technology (and technology in general) are already legislating about the use of “artificial intelligence,” in particular its use in hiring decisions.

This sophisticated mimicry of human creativity is impressive as hell, but it’s a parlor trick of sorts. Human intelligence is based on pattern matching, and computers with their speed and massive storage have replicated some of that. But artificial general intelligence (AGI) remains out of reach, unless you believe Elon Musk’s twitter claim that he’ll be surprised if we don’t have it by 2029.

I encourage you to become more knowledgeable about programs such as GPT-3, Midjourney and their cousins. Although your business may never use them directly, soon you may find them being packaged up in programs your business relies on. While “magic” is a handy explanation of how it works, these programs have strengths and weaknesses you’d be well-advised to understand. You may find this Scientific American article (tinyurl.com/282k38n2) a good starting place.

 

Related Posts

Loading...

Sections