AI Art


Hey guys,

Been fucking about with some AI sites and AI upscaling.

You can create some really fucking weird abstract art with these things, feeding in creative prompts.

I’ve been using Craiyon to generate the source images, but they’re really low res, so I then pass them through the Zyro AI Upscaling tool to increase their resolution. If Zyro gives you an image larger than 750x750, but still not high res enough, downscale it back to 750x750 and upload it to Zyro again, it’ll probs give you something nearly 2000x2000 after that.

Have a play, be cool to see what stuff people come up with.

Here’s some of the pics I’ve made with this process. I’ll probably use some of these as album art.


Here’s some others I was experimenting with but never upscaled.

These are just super weird! But I love it!

IMG_0017 IMG_0018 IMG_0019 IMG_0020 IMG_0022 IMG_0023 IMG_0024 IMG_0025 IMG_0026 IMG_0027 IMG_0028 IMG_0029 IMG_0030 IMG_0031 IMG_0032



All good examples of how AI is more A than I … at least for now. Super fascinating where all of this computer science is headed.

I’m looking forward to a world where imagination generates reality automatically. BUT!!! Remember the KRELL!!!

Thanks for sharing @psyber


Haha, I think we’re already at the point where you can’t really tell the difference between AI art and human made art. Particularly with DALL-E, a recent innovation in that space, which is currently only available for use if you pay for it. The one I linked is based off an old, less capable version of DALL-E Mini.

Now, without cheating, tell me which one of these images are generated by AI and which are made by a human…


OK… I’ll give this variation on the Turing Test a crack.

No cheating… (google etc.) just my eye and my thoughts about what I see and what I think.

In the order they appear:

This is tricky in that all of these could have been made by a human. In fact, given that a human created the AI and provided the input for the subject, technically, they are all ultimately generated by humans.

So… how to discern the difference? I noted a difference in level of storytelling: Static vs Dynamic.

Best example is the boy and dog… they seem to have personality, specifically, a sense of joy and wonder and also companionship… sharing the joy and wonder. Seems more likely that a human could add these layers in such a subtle but compelling manner.

Next best example: the tower of buildings: they are all compositionally related but each building has little quirky individual characteristics.

The difficult ones were the first two. On the first one, I can imagine a deeper story (mythology or history of some sort) but I’m projecting it rather than learning it from the piece. So… Computer.

The second one I was stuck on until I realized that the Great Red Spot of Jupiter was located precisely to best effect to elicit a classic SciFi theme. In fact, this piece looks like the cover to a paperback from the 80’s.

How did I do?

By the way… even if I failed 100% I stick by my opening statement…. And further… the whole exercise begs the question “how much help did the AI have with tools it was given by humans?”

Is spellcheck or even grammar check AI? Certainly not… but often smarter than me.

AI seems to always be the things we say computers can’t do… until they can…(chess anyone?) and the it turns out computers still do what they are assigned to do… like any good tool in the hands of a creative human.

I want AI to be real…:star_struck: not convinced yet. :thinking:



Well, the actual results are:

So you got 2/6 right, which is about the same as chance – you’re 50/50 if you select one at random.

In all the discussions I’ve had or read about this, in person and on other platforms, I see a lot of confirmation bias – people know ahead of time which ones are AI and which are made by humans, and then they retrospectively make observations that fit with the preconception. Conversely, though I haven’t seen this, but I want to try the experiment, is pick some good AI images and say that they’re actually made by humans, and then pick some human art and say they are actually AI – I bet people will just as well make similar observations.

In blind tests, people can’t tell. You found story that seemed to fit the situation, but you have no way of verifying that. This is an example of a cognitive bias discussed in Thinking Fast and Slow by Daniel Kahneman.

It’s fine though, I don’t think I could tell, either. :slight_smile:

But is not everything a human does influenced by another human? We’re just copying parts of ideas and composing new big ideas our of those smaller abstract pieces.

What really is the difference between that an an AI learning from what humans have done?

These machines aren’t told what art is “good” or what art is “bad”, the fitness function is basically just “how much have I convince the second AI that these images are real?”. The second AI is the adversary (from adversarial deep learning), and the second AI’s fitness function is “how well have I predicted whether this image is real or not?” The AIs iteratively compete and evolve – it’s literally a survival of the fittest, and so evolution is an applicable term here (though biological evolution is actually not about survival of the fittest, it’s “how many copies of my genes can propagate into the next generation”).

No true scotsman…