While waiting for Worlds and new game to be revealed, I stumble across the trend of AI generated action figure look like?
I played with ChatGPT with this prompt:
“Draw a picture of me as a real-life action figure. Be specific as possible based on what you know about me. And put multiple interesting things that represent who I am inside the packaging.”
ChatGPT did have some idea about VEX Robotics and WalshBots as my interests. I did give it a picture of me (one that was one internet for decades. It generated the following image:
A little AI confused, but interesting! I am sure with more refinement to prompt you could get some good renders.
It would be fun to see other teams AI generated action figures - collect all 32,000 teams!
8 Likes
I did note that my picture has been around for a zillion years - ok 30+ years - I am also part of the original face recognition database developed at MIT Media Lab. Fun having my friends finding them while they were doing their own facial recognition.
Privacy is indeed a global issue - I do believe that ship has sailed long ago.
You can feed some random picture that kind of resembles you a little that is not you.
5 Likes
I revisted ChatGPT to give it a descriptive prompt vs using images of a real person:
“Draw a picture of of a blond hair teenager is part of a VEX Robotics Competition team as a real-life action figure. The teenager is part of the WalshBots team number 9791C known as Roboty McRobotface. The team played VEX Robotics Competition Turning Point in which robots shot yellow balls at blue and red targets arranged in a 3x3 matrix along the far side wall of the 12’x12’ competition field. The teenager designed, built, coded, and drove the competition robot with the team. put multiple interesting things that represent team member inside the packaging. Include the team’s logo in the packaging design.”
and additional prompt:
“The teenage male is around 5’ 8”, has medium length blond hair, and likes to wear shorts, hoodie with the team logo, crocs, and mismatched socks."
First pass it came up with this:
Nicely mucking up the logo, ignoring some details, so I scolded it:
“Please use the actual logo uploaded and do not modify it in any way. Also it is missing the other details of the additional prompt: “The teenage male is around 5’ 8”, has medium length blond hair, and likes to wear shorts, hoodie with the team logo, crocs, and mismatched socks.” missing are shorts, mismatched socks, and hoodie."
and it came back with this:
Which is much better! Not sure why it hates our wildcat logo 
So, I asked to fix two elements:
“The team logo is not a bird, but a wildcat using the progress flag for a background, can you fix it? Also, under the logo on the hoodie include “WalshBots”.”
This is all to show, you don’t have to use pictures of yourself, you can give descriptions and it will do it’s best.
2 Likes
This whole scenario relies on people rationalizing toward it being fine. We’re engineers. As well as creating behaviours, we’re here to find and either eliminate or redirect misbehaviours.
5 Likes
See, my perspective on this issue is that there isn’t much of a problem in using AI to make action figures of yourself. But, I think it’s sort of lame. “Look at this souless image the computer made of me” really doesn’t feel like that fun of a statement.
I also feel this way about teams that use AI for their logos. I feel as if it doesn’t allow teams to create an identity of their own. I also think that these images end up looking really low quality.
5 Likes
Here is last post on topic of 3D action figures. It turns out MakerWorld in the Maker Labs section has an Image to 3D model converter which is surprisingly effective. We cropped the action figure and fed it to the converter and it generated a nice 3D model that printed well. Here are the results.
3 Likes
Agreed about seemingly “soulless” feel, but I think that may be how new this approach to design is. Same could be said about early 3D animation that took ages to generate and felt really far from reality (I have prints from early 80s work that was high tech in its time), yet here we are today with 3D that is used extensively in film production.
So back to fun - right now noodling with the technology it is interesting to play with and see how each iteration can improve the rendered results. I am sure that with time we will see a lot of good output for teams as a means of expression.
I have also noodled with using AI to evaluate notebooks to give teams feedback, it catches details that I would have missed or not amplified that were based on the rubric. It was good for early season, but did not really provide useful feedback to teams that had multiple iterations or highly growth over season. Will these systems get better over time and provide utility, absolutely!
Concerns about privacy are warranted as it is unclear who else gets access to information you provide to the model.
As an arm chair designer 
I still prefer the hand rendered version of our club’s mascot. It has been used in variety of ways and ages well.
1 Like
