Say what you see!
Say what you see! is an AI prompting art game created by Jack Wild on Google Arts & Culture. The game follows a simple structure with easy-to-follow gameplay and several levels of play that get progressively harder as you go on. The game begins at level 1 with the goal that you need to reach a 50% match to pass the level, and it provides an example image with keywords and a helpful hint of things to remember when making a prompt. When you start the level, you are given an AI-generated image with a blurred-out description on the left; on the right is an answer box where you are prompted to write a description of the image in 120 characters or less, which will prompt Google AI to recreate the image. The goal is to hit the correct percentage for that level. If you do not pass, you have two additional tries to modify your prompt to get to a closer match percentage. The little AI brain character will give you some feedback as you go. If you do not pass in three tries, it will result in a loss for that artwork and kick you to the next artwork. If/When you pass, the unblurred original AI description will go live. You will not be able to see the original AI description if you do not pass that particular artwork. Each level has three artworks you need to describe and try to match. At the end of the level, it will show you a review of the level, and then you can move on to the next. Whether or not you pass or fail the whole level, you can go on to the next level. As you progress, the percentage goals of each level get higher. Level 1 = 50%, Level 2 = 60%, Level 3 = 70%, Level 4 = 75%, Level 5 = 80%, Level 6 = 85%, Level 7 = 90%, etc.** The game also lets you open an explore a page where you can browse real artworks from Google Arts & Culture's digital museum collection that either visually or thematically relate to the AI artwork. (Wild, n.d.)
**The game levels go much higher than this, but 90% appears to be the highest percentage it goes up to. I clicked through and got to Level 15 before I stopped. The only thing I noticed with the game after so many levels is that it started to repeat the same artworks. I think there is a limited data source within the game, meaning the original generated AI artworks are static, not continuously generated into new sources as you play. However, while the repetition means that you may have a repeated source, the percentage match for that level may be higher and thus the level still has a challenge to it, as you more than likely will not be able to re-use the same answer depending on what match percentage you achieved at the lower level.
The following screenshots walk through Level 1 of the game, examples of both incorrect and correct AI prompts from the same image, and a screenshot of the explore page:
There are a variety of ways that I can see this game being implemented into my classroom. Many of my students are interested in AI and are very interested in using the many AI-based tools in the Adobe programs that we use in class. My personal philosophy when it comes to creating digital art is that you can use whatever tools are available to you, but that should be after you have a clear and in-depth understanding of how you achieve that particular function by your own hand. As with many other digital tools, I think users can get dependent on fancy features when they do not really understand what they are actually doing. You may be able to create a cool filter in an app where a bulk of the programming is built into the data, so you click a button and voila, but then in a program where you control everything, you may have no clue where or how to start. With this understanding of students and my philosophy in mind, I believe this game can be incredibly useful.
According to Gee and Shaffer good games can be incredibly powerful learning tools as they provide a mix of practice and guidance, and they use appropriate language and can introduce complex concepts (2010). However, Bradley and Kendall, also mention that participating in games and simulations alone are not enough to guarantee that meaningful learning will take place (2014). Rather, there should be a balance of both, presentation of content and learning goals that are reinforced by the gameplay play which is then reinforced with discussion, interactions, and assessment.
For use in my classroom, the goal would be to use this game as a tool to understand the nuances and complexities of AI in relation to art, and why it is important to understand how the basic tools of a program work, so you can more effectively utilize AI. As an additional learning goal, this game will also help students more effectively look at, assess, and practice visually describing art.
I would integrate the game most naturally into my digital art class, but I could incorporate it into any of my classes. I would introduce the game as a two-day AI play day after we have pre-studied and practiced describing artworks and discussed AI digital manipulation versus "traditional" digital manipulation in Photoshop. Students would work in groups of two with the game and a slideshow that they both need to fill out as they play. While the game does not work exactly sequentially with the same prompts at every level, the following activity could still run smoothly. Students would open both the game and their individual slideshow, then work individually through levels 1-5 of the game. Each slideshow will be personally shared with each student as a prefilled template. There will be areas to write in their prompt guesses if they got prompts "incorrect" or too low a percentage match, and a slide for a sub-level screenshot of their game screen of the "correct" or at/above percentage match. They will fill out the prompt boxes in the slideshow as they play the game and add the screenshots at the end of play if they wish, as long as they were taken while playing. Each screenshot should include the original AI-generated image, the written prompt, the new AI-generated image, and their percentage correct, then put the photo into their slideshow. They will repeat this for all levels. When this slideshow is completed between both students, they will compare their answers for common artworks and assess the percentage they received, and if one of them received a higher percentage, assess the written language to determine why they may have received a more accurate AI generation. The students will create a shared Google Doc of their determinations and include a photo of the artwork they are referencing. I will compile all of the students' findings about the same artworks into a slideshow to be discussed in the following classes.
Talking about AI and prompting language in the way I would approach this lesson/unit would scaffold to the concept that you need to have a full understanding of what your goal is. Whether that is an AI prompt, using a digital art tool, or trying to create a specific edit or effect, and you cannot rely on AI to complete it perfectly, you need the background knowledge first. This game, the process of this lesson, and the learning goals of this lesson can be assessed formatively through self-assessment and my assessment. While playing the game, students will be working through their own version of formative self-assessment, as 'there is an iteration involved in videogame learning that includes new, revised, or repeated moves based on various forms of assessment that enable players to evaluate and reconsider positioning and progress' (Lynch, 2021). Students will be constantly assessing answers and re-evaluating their decisions to try and improve their previous goals. I will be reviewing personal slideshows and group Google Docs to check for understanding that students were able to meet the concept goals of the lesson and the game.
As a continuation of this lesson, I would use the compiled Google Docs to make a slideshow for a discussion with the students over the various ways they created prompts for the same artworks and their effectiveness. This would eventually culminate in a project where students would create an Adobe Photoshop or Adobe Illustrator artwork by hand, then they would have to use their new AI prompting language skills along with an AI program to visually describe the artwork they created and try to get AI to replicate the artwork they made. This project would reinforce the concepts that AI is only as strong as the prompt given to it, and the prompt is only as strong as the understanding the individual writing it has of the techniques, tools, processes, and goals they are writing about.
More:
Click on the link below if you would like to check out this game:
https://artsandculture.google.com/experiment/say-what-you-see/jwG3m7wQShZngw
OR
Use this link if you want to check out some of the other game options on Google Arts & Culture:
https://artsandculture.google.com/play
Highly Recommend:
- ARTETIK (emotions, themes, connections to artwork, meaning making)
- GeoArtwork (geogrpahy, art & art history)
- Play a Kandinsky (music & art)
- Learning Light (photography/lighting)
- Sparky (art, engineering, play & research)
References:
Bradley, E. G., & Kendall, B. (2014). A review of computer simulations in teacher education. Journal of Educational Technology Systems, 43(1), 3–12. https://doi.org/10.2190/et.43.1.b
Gee, J.P. and D.W. Shaffer (2010). Looking where the light is bad: Video games and the
future of assessment (Epistemic Games Group Working Paper No. 2010-02).
Madison:University of Wisconsin-Madison.
Wild, J. (n.d.). Say what you see - google arts & culture. Google. https://artsandculture.google.com/experiment/say-what-you-see/jwG3m7wQShZngw
Lynch, T. L. (Ed.). (2021). Special issues, volume 1 : Critical media literacy : bringing
lives to texts. National Council of Teachers of English.
No comments:
Post a Comment