> increasingly indistinguishable from that produced by real humans operating in the real world.
Hell, I imagine it is going to surpass even the most creative, talented humans "pretty soon"(5-10 years), to the point where people will actively search out AI generated content.
My concern is whether this will trigger the end of human creativity, or if humans will use it to inspire themselves and still go on to continue creating art.
Over the last 20 or so years musicians have gotten used to using a lot of prerecorded riffs and samples. It's now possible for someone that can't play any instruments to create very goods songs. Unfortunately, the use of these same samples has given us a situation where lots of modern music sounds the same[1].
I suspect that's where we are headed with all the coming automation. A lot of human creativity is going to slow. I don't think that's a good thing. But unfortunately, we have no choice. It's coming at a furious pace.
Musk has been saying for years that AI will damage/destroy society. We've all thought that for that to happen AI would some how need to be sentient. But given what we've seen lately, all that's needed is a tool like GPT-3 to be used as a weapon against each other. Good luck to us.
Good point, the tech will destroy us trope has been around for a long time and has been better articulated by others. I picked him because he came to mind. But you're right. He's just repeating what others have said.
> Over the last 20 or so years musicians have gotten used to using a lot of prerecorded riffs and samples. It's now possible for someone that can't play any instruments to create very goods songs. Unfortunately, the use of these same samples has given us a situation where lots of modern music sounds the same[1].
Coming form various EDM genres and spending time online and in the clubs with many of the producers growing up and now also studying AI and ML I think this is the closest analogue: but it fails short to encompass the entire scope of the situation; I think for now it's apt, but in time it will be able to do much more and that and that is what I think is what we are in for a hard landing.
I don't think it will be 'the bots took our jobs' but more like moating will be ever more precocious and wide-spread as a means to maintain relevance, 'if you can't moat then you can't float' will be a common maxim which was always implied in startup land if you wanted to go the VC route, but is difficult for 99% of projects for who that is impossible. This has dire implications that I'm not entirely sure we can really be aware of until it happens.
To follow the earlier mentioned music analogy: why learn to learn 10 different instruments to compose when you can jump on youtube and learn how to sample and add the effects you're looking for and then be on your way?
I think this will also apply to various things in the Arts, and possibly in the Sciences (like co-pilot is for programming) which in theory should lower the barrier of entry to produce: people keep going on about Elon, but Grimes has been talking about this for sometime, too.
She attributes her entire music career to being able to get on Ableton and sampling and harmonizing ad-hoc to create her compositions, I don't follow her but I've met people who have and her live stuff is more performance art than it is a typical musical concert and she has managed a way to stay relevant even in a sea of similar sounding artists--I've since seen similar acts, many of whom I think are far more talented than Grimes (Sierra or Meg Myers) who simply don't get the recognition they deserve because of other factors like PR.
But in the end: you can auto-tune a IG influencer all you want, but if they aren't able to make a connection using the medium it doesn't matter how polished the product is if it just plain sucks. They could probably be used for commercial teen idol pop stuff to sell stuff, but that isn't a very high barrier of entry, either. The 90s boy band craze proved just how low that bar is, while still being a commercial success.
My point is, that while the barrier of entry may be lowered if the art itself doesn't have appeal it won't suddenly be comparable to Led Zeppelin or Mozart.
I remember in the early days of Bitcoin, like 2010, their were a ton of programmers (mainly who wrote in C+) who wanted to contribute but had no direction or idea of where to begin, they had all the skills to be able to jump on a project but because they lacked any vision they didn't know where to begin.
I think the advent GPT and Dall-E is comparable, though it must be noted that very few have actual access to it right now: the access seems to have some non-chronological factor, and apparently you have to provide social media and linkedin accounts, which makes it all the more creepy.
Personally speaking, I really wish Altman had just focused on how to best deploy this instead coming up with World Coin, he could have raised so much more awareness and funding maybe even launched his own token within this ecosystem and maybe even created a real usecase for 'web3,' instead he will always be seen as the guy trying to scam you for your biometrics. And as cool as this is, I doubt it has that potential anymore because of that.
I think it will be a huge boon for human creativity. Imagine if anyone could create a high budget movie with less effort than it takes to write a novel today. Might take a while, but that's the direction we're trending in.
About two decades back, I knew a B-list Hollywood director who wanted that. He'd found out that Reboot, the first all CG cartoon show, took a staff of about 30 people to produce an episode each week. . He wanted to be able to make a movie with about $10 million and maybe 50 people. Because then he could direct. Making a $100 million movie with a thousand people on staff is project management. He was spending far more time in meetings than on-set.
That didn't happen, though. Productivity did not go up for Hollywood. Movies became all big scenes, all the time. Production costs went through the roof. TV shows had to upgrade to production values previously seen only in film. Movies are now made twice, once as "pre-visualization" to get approval to spend the money for the full version. Look at the end credits scroll by on a effects movie. There's no longer a cast of thousands, there's a staff of thousands. Since nobody can afford a failure, everything is a predictable sequel.
Yet, at the same time, there are people on YouTube making broadcast-TV-quality content with nothing more than a camera and a laptop.
Modern movies are expensive to make because they can afford to be expensive to make, not because they need to be. As technology improves, the gap between what independent creators can do and what high budget studios can do will continue to narrow until, for certain types of content, there'll be no discernible difference.
AI is very bad when it comes to making a linear narrative due to it's memory limitations. I doubt we will be seeing long form content that is made 100% by AI even in 10 years.
I can see a sub genre being born where authors let AI auto complete every few sentences though.
I am willing to bet up to $1 USD that AI will be able to generate a 5,000 word essay on an arbitrary but common topic which is indistinguishable from human writing to a panel of 5 normal humans, all by Jun 28, 2032.
What are ‘normal humans’? If I take 5 humans at random from a non university town, I am willing to bet up to $1 they cannot distinguish this now when generated with gpt3.
I struggled with that phrasing because it wasn't immediately obvious to me how to describe a class of people capable of judging the problem but who wouldn't be biased in some manner. I'd readily submit that no human could determine the difference, but I'm not ready to pay the testing costs...
I wonder if we'll get an AI D&D dungeon master anytime soon. It could be maybe given an over-arching plot, and then just let it riff off the player actions (so, hopefully the players act in a way that imposes the linear narrative and common sense, by staying in character).
Maybe even filter these, edit a bit, and sell 'em as stories.
Huh, I tried it out. It seems like a neat proof of concept/prototype, but not much of a game of course. I'm sure someone is working on pairing it with more of a proper game system.
Hell, I imagine it is going to surpass even the most creative, talented humans "pretty soon"(5-10 years), to the point where people will actively search out AI generated content.
My concern is whether this will trigger the end of human creativity, or if humans will use it to inspire themselves and still go on to continue creating art.