If we strip the buzz, the latest hubbub and media attention on generative AI are large scale data models used to generate youtube video scripts and images that are approximately human-made in quality. But this is not really new or novel. Anyone who have seen computer graphics evolve the last 20-30 years would know we have came a long way from 8-bit Mario, PS2 Cloud Strife, to whatever the hell Final Fantasy movies they stopped making because people are no longer interested in feature films with hyper-realistic computer visuals.
The really interesting or attention-grabbing aspect about ChatGPT or whatever is that now a larger group of people have easy access to these tools in order to further achieve our self-actualization of desires, to put it crudely. It doesn’t take millions of dollars, just tens, to thousands now. The average above-average gaming rig can make more ero content faster than one whole Comiket in the same half-year span, to use a random, cherry-picked benchmark, in voice, words and images. Not that it would be as good, although it wouldn’t be something we can completely ignore.
And I think this is where it stands. Proliferation of computer 3D modeling tools gave us dancing Miku, but also a lot of 3D ero content. Those are just some examples, and it’s really everything in between and more. But as they say this is old hat. It’s not that we can’t create life-like visuals, enough to fool people. Unreal Engine 5 is a compelling example, even if it leaves much to be desired still. But these gaps will be there for the foreseeable future–it’s a feature of the uncanny valley. And arguably we are only at the early stages of discovering these types of features in these other forms and modes of expression.
I asked Google Bard about this topic in general, and it seems that there are companies, productions and animation shops that use generative AI in their work, of course besides the examples I listed of the older technology that effectively does the same thing. It is actually not a huge surprise that Bing and Google Search are in the mix of this, because a search engine, or Google Assistant, is effectively the same sort of mode in which a chatbot that gives you some answers largely based in search results can provide.
Just like this person who made a Miki Hoshii bot to send him emails (AI mimicking games mimicking life), I expect these creative uses to actually be in the products I consume. It’s all there already, just missing the LLM part.
Bard also gave me some pitfalls or dangers, mostly having to do with creating unwanted content in more authentic forms than ever–fake, spoofed stuff mostly. So, sure, authenticity is the real issue at core, but the expression is likely slightly different. It’s not deepfake trying to manipulate humans, it’s deepfake targeted at bots. The bigger risk is just what’s problematic today: plain, garden variety spam. It would be easy (as it already is done) to create a whole web of sites for SEO purposes. You can GPT4 your way into this automatically, and it wouldn’t cost a lot. Does Google Search know how authentic the content is? Because if we rely on tools today, built yesterday, with yesterday’s assumptions engineered into the way it works, will it still work tomorrow when these generative systems are pervasive and better than ever?
We need the strongest antispam features these large data models can provide, right?
There are also the legal aspects of it, which are mirky at best. I think it’s a fair callout for all the artists worried about their prospective customers can GAN their way out of commissioning jobs. But that is not reason enough to throw shade at the tech and its advancement. This is very much a horse buggy moment for those folks, so they can use our support in ways to move out of that job or work in a way where it’s more resistant of being automated out of their jobs. Writers and other creatives already had to live with other tech eating their lunch for decades as more of the world become code. Not that it’s “okay” but I guess in 2023 we should not be surprised about any of this, as there are more and more people who are organized to deal with the fallout and changes these disruptions can bring.
Frankly, the less we can say about copyright the better. Yes it’s broken. People like myself have been saying this for decades. But there is no will to go fix this once and for all, just small amount of people all over being aggrieved about some aspect of it, even if in collective, that’s a lot of people. It is not really the right mechanism to regulate the way internet has transformed the lives of individuals, for starters. [Nobody cares about Mickey Mouse, really, so what is Disney worried about?] It’s hard to understand, hard to rally around, it has too many stake holders, and it’s really important–too big to change without having some fallout possibly. It’s not that we should talk loudly about copyright, but it’s not in this space I think. Marching up to Congress with a million protestors is the right thing to do perhaps, to use an extreme, but it’s gonna happen because of furry artists are getting paid less.
Anyways, it’s a great time to actually learn more about the nuts and bolts of what these systems really are, to cut through the hype marketing that surrounds any next-gen tech deals with when it finally goes from the lab and ivory towers of academia to the hands of everyday folk. It can give amazing results, but it’s also not that crazy! As they say, something is indistinguishable from magic is only because people are incapable to learn how to, but this is not the case with this tech. Plain old human beings came up with these, it really is not rocket science.