How imaginative is AI

I have gotten some remarkable stuff out of AI. But it's hard to 'art direct'. Even when you feed it an image to inspire from. 

I thought I'd give it a test. I have this old newspaper insert cover. A humorous look at how the Boston Symphony instruments find their way to Tanglewood for summer season. Nutty for sure. 

I tried 3 AI apps and they all failed  big time. They didn't understand the word humor in the prompt or the concept that clarinets can fly let alone a grand piano. I was quite disappointed but realized that human are very abstract thinkers. 

I'm going to keep trying to make this work on AI.


  • Mid Journey. Not super funny but has a nice charm going on

  • There is no imagination and no intelligence in AI whatsoever. You told here about your career path but you did not say if you touched the early photorealistic computer graphics of the late 80s early 90s. There were AI things too, but all that is pure math and statistical models looking (not learning) from what has been done earlier. Been there done that, as they say. 3D stuff or drawings coming out of the computers were really impressive, some would be even today (Jurassic park). The only difference today is the computing power per unit of money and  marketing. 

    It is nice to play with for sure, but remember the limits and above all, don't be fooled by that.
  • i am very inspired after seeing this beautiful art
  • I this AI reminded me of a sketch I did…
  • I think of AI art as the product of a distillation of human visual and aesthetic knowledge and its embodiment in an artificial substrate. In that sense, AI art is only human art. And AI is now capable of learning. We have programmed that ability into it. There is nothing inherently sinister and malevolent or unnatural and fake about AI art. However, as @KingstonFineArt points out, it still has a way to go. I think it will be difficult (but not impossible) to program AI to understand and generate humor in art.
  • @whunt
    Here's a couple of images based on your drawing. I fed your drawing to Midjourney. The left one is straight in. The right has a weak modifier of flying instruments. I kept the organic nature and technique, pencil, of the original in both. A great tool.

    All the AI apps are interfaced through language. In the sense that we feed then our 'sentence' structure and syntax with modifiers, --ar --s, even images to influence. Again based on a modifer --iw. Depending on the access to the language interface. Dalle has none. MJ only through modifiers and seeds. StableD on a PC full install has the most control. 
    The engine in the back is the AI. Held in check by the input or allowed to create at will. That engines has been trained on up to billions of images which are tagged and valued to put it simply. The values are compared and recompaireded in the visualization process. In MJ you can see it happening. Areas of tonality and noise are evaluated, compared and matched. Pretty fascinating. Sort of like the human mind.
Sign In or Register to comment.