Concrete vs. Abstract?

Concrete vs. Abstract – can AI represent more?

What can AI imagine?

As I was playing around with Craiyon, I experimented with taking an abstract concept and putting it next to a concrete noun. In every instance I tested, Craiyon ended up anchoring the image to the concrete noun. The tool is generating for what is “easiest” for itself to understand and what it has experience and data to generate images of. I wonder how useful tools that leverage AI to help us humans will be at making us think outside the box – outside the tactile world. Will AI help us understand and push us to consider concepts?

For example, I searched “cowardice toothbrush”, “motivated toothbrush”, and “religious toothbrush” and got results that were identical.

For fun, I searched “Christian cult Harvard” because late this Saturday night, I walked into some sort of initiation ceremony which consisted of well dressed boys linking arms (and bodies) singing songs while parading a large cross around in a circle with presumable “leaders” instructing them on what to do. I wondered how Craiyon would represent what I saw and the tool could not incorporate the non-tactile parts of my text. The AI tool needs practice in representing a wider range of concepts – to “think”!




Leaked: This is the winner of the Harvard Yale Game, according to Joe Biden


AI Limitations: “There’s No Place Like Home”

Student comments on Concrete vs. Abstract?

  1. These are hilarious prompts to feed the AI. I wish it had been able to tell the difference between the types of toothbrushes! I would have loved to see that interpretation.

  2. Very creative inputs, Riya. Craiyon would love short nouns but you gave it very difficult ones. I guess since the “cowardice toothbrush”, “motivated toothbrush”, or “religious toothbrush” is not a common or specifically determinable noun, Craiyon had a hard time without much creativity

  3. Hi Riya, thanks for sharing these generated images. I like that you are trying to engage the AI with abstract and concrete words to test its understanding. I think more training and iterations are needed to get the AI to understand abstract words. However, since abstract words may also be interpreted differently between people, the AI may never properly pick it up.

  4. That was great. When you ask questions like that you are pushing forward what the tool can do. I would struggle to draw a motivated toothbrush, but if AI is going to help it should be suggesting things we can’t do for ourselves. Maybe it needs more training. I hope you escaped the cult.

Leave a comment