I know there’s a fair amount of controversy about these tools too - potentially replacing jobs, and potential for how they use source data in a non copyright complaint way - but clearly these are here and staying around.
“Well, if the bar for being an artist on Fiverr is just realizing that the images are good, then you might as well go for it! Who knows, you could be the next Monet… or at least the next Fiverr sensation.”
“Wow, it sounds like ChatGPT is an incredibly powerful tool! I’m glad to hear that you’re finding it useful and that it’s able to assist you with such a wide range of tasks. It’s always exciting to discover new and efficient ways to work, and it sounds like ChatGPT is helping you do just that. Keep up the great work!”
(I asked Chat GPT to reply to you… rates itself highly)
Bear in mind that when it comes to research, it can make up very accurate sounding articles that are actually full of stuff and nonsense about things that do not exist.
I can’t find the Twitter threads now because it was before Christmas and I didn’t save it, but there was a scientist who asked ChatGPT to write articles in his special field, and they were really convincing, looked sound - and he found it really hard to spot the fact that the theories were made up and the citations not real, because it all looked so authentic at first glance.
My opinion right now is, they aren’t going to replace real talented human artists no matter how much they’re crying about it Twitter. These AI still have limitations humans don’t, and they can often have pretty bad artefacts. Too much of it has the uncanny valley effect. They’re faster though.
The writing one. Yep. I’m sold. That’ll replace human writers easy, because I wouldn’t be able to the difference. Or more accurately, they’re going to accelerate the dangers of fake news.
They can code too, so watch out developers!
Not long now until The Machine or Samaritan is real if it’s not already.
I think this is the thing right (taken out of context of course) - it will accellerate a lot of areas - for the art side - will speed up the ideation process for non creatives to get concepts and ideas, or generate quick and dirty story boards, concepts, whatever, to act as inspo.
And on the counter side, will make all the negative bits faster too. Double edged sword.
Hadn’t thought of the art side that way honestly, despite seeing this is a benefit for coding, so it makes sense. Artists aren’t as on board about it as the rest of though, I suspect because they’re worried about what products like Lensa means for their wallet.
If people can get a hundred portraits in an hour for a fiver, why would they pay 20 times that and wait a week for a human to do the same thing? The human does it better still and I’d rather still commission them than settle for the uncanny thing.
Here’s the best Christmas portrait Lensa produced of my dog. It was in the £3.99 pack and took about 20 minutes to get the images. Notice the teeth? It was like that in everyone where it had the mouth open, most worse than this:
I don’t think artists need to be worrying about Lensa. I’d still choose to pay that £45 and wait a month than gamble £5 on something that’s probably going to look uncanny and have weird artefacts.
Haven’t tried running a description through stable diffusion myself yet though, which might yield something better. But Lensa is the one they’re going after and accusing of destroying their livelihoods.
I haven’t tried any of the image creating ones, but I have used ChatGPT to create email introductions, press releases and social media posts for work and nobody has noticed.
As others have pointed out, the main problem with ChatGPT is that because it understands HOW to write, it will sound convincing whether it’s correct or not.
I can’t see how things like Dall-E or Midjourney will take the job of people who photograph/video actual real life events like sports photographers for example. But for digital artists or even people who do things like landscape photography, I’d be feeling very unsure about all this.
I’m not sure how google would spot it honestly. For me, the articles pass the Turing test. And they go beyond. They’re better written so much of the human produced stuff out there. Maybe that’s how you catch them? But then you risk penalising those who take the time, care and effort to produce well written articles.
This for instance is something I was going to test this year, to see if anyone noticed. Can’t work out of this is cool and super awesome. Or downright terrifying.
The important thing, I think, is to not be that guy who pays £0 to £5 for the AI art and then tries to hire an artist for £10 to fix it.
My biggest problem with AI art is the fact it has been trained on real art, meaning it’s effectively a copyright infringement factory. Googling something like “artist whose art was used to train ai” finds plenty of stories like this:
I think that’s fair - and our copyright and licensing laws are fairly antiquated - it’s not like we have the right mechanisms to properly deal with the future of this stuff yet, and how it should be / could be protected.
I don’t personally know where the distinction between “using artists source material to train an AI” vs “using an artists source material for inspiration” lies. Obviously, there is one, and effectively remixing other peoples work to make new works is how all genres of anything creative come about. (looking at you Pachelbels Canon in D).
I think there will be a lot of interesting case law in the future about trying to define that distinction though.