Leonardo.Ai reflection

During Wednesday’s class, the studio Augmenting Creativity really harnessed the studio’s name as we utilised Leonardo.Ai by using our own original ideas through narrowing down our own prompts to generate AI images. While being provided with guidelines for what our outcome should resemble, I found that Leonardo.Ai was particularly weak when provided with too many details. It was almost like the model became confused when being fed too much information, and would entangle or twist certain words, linking them with other words in the prompt that were meant to affect different areas of the desired image. I encountered this and thought maybe my punctuation was not accurate in listing the attributes I desired, however even with edited punctuation, the model still seemed confused. I had a lot of difficulties when generating my self-portrait and found that to get closer to what I wanted, I needed to be less wordy, and try to describe less of my features! I am also half Chinese which I think made it very difficult for the model to accurately depict what I look like as putting in words such as “mixed race” or “half Chinese” or “wasian” is a very broad scale, and one that the model could not really get perfect for me! But it tried!

I thought I would also mention that Leonardo.Ai does attempt to attach all aspects of the prompt to the best of it’s ability, even when the given prompt might not make sense or be an incorrect description of what it wants. I tested this by giving it a prompt that I knew would be difficult to generate as showing it in one image wouldn’t work easily. I gave it the prompt ‘A gorgeous beach with coral and fish and crystal water, and soft white sand’. I have linked the image that it generated which was an obvious attempt of incorporating a beach AND an underwater scene with the coral and fish. Though the image did not make sense, it was interesting how the model tried to find a way around this inconvenience, while still making it an attractive image. This is where I think the model definitely takes creative liberties as it works around inconveniences and thinks outside the box to provide an image as close to the given prompt as possible.

While completing the prompts for the class activities, I found that Leonardo.Ai was very successful in creating different scenarios for my prompt for the studio poster. I told it something along the lines of ‘students in classroom discovering and collaborating with AI’, which gave me university students with a type of robot that resembled the artificial intelligence as it’s own physical human-like being, in a futuristic robotic way. I think it made a very clear image for the poster that portrayed the message that in this studio students will be working with AI to further their learning… We will be collaborating ideas with AI to expand the barriers of our creativity.

I am very excited to continue utilising new tools in the coming weeks such as Leonardo.ai, to further my understanding on communication between human and artificial intelligence.

I