To start things off, I took on the advice that my supervisor gave us in an earlier consultation the previous semester, and looked back on my progress since the start of this project. I divided my reflections into two columns, what I did well, and what I thought didn’t go so well. Most of these points were points I was already quite aware of and included several limitations I had mentioned in my last CPJ entry in the semester before.

Nonetheless, I noted down everything I could think off and used them to work out some goals for myself. I have to say though, while I still personally feel like I didn’t do that much last semester, having the achievements noted down in this format does spark some confidence at this point in time.

With my session of self-introspection over, I returned to preparing my AI. In semester 1, I decided to draw the line early at just having an experimental prototype for the self-censoring AI due to the time constraints. Hence, I needed to get started with preparing my dataset for building an actual model and pipeline for the final prototype I am to have on display at the live installation.

As part of my preparation between last semester and this semester, I also completed my fieldwork where I spoke to several creative industry veterans to evaluate and hopefully gain some validity for some of the concerns that I had surrounding the entire idea of AI and its potential for disruption. I made sure to analyse these thematically and these contributed to my dissertation that ended up being the foundation for the direction of this project.

Next, I moved onto preparing the visual dataset. A visual dataset of quality was necessary for this project as I needed to be able to control exactly what kinds of scenes were likely to come out of the AI when my audience interacts with the prototype. One of my major findings from my research and the driving force behind this project was the potential for the commodification and corporatisation of Gen AI to amplify the templated effect that prevail in many present day organisations.

Hence I opted for a similar approach by tapping on undraw.co’s public library of corporate illustrations. Except that I had to also censor it in every shape or form as part of the project’s intentions was to also showcase the potential future where creativity is further crippled by fears of being offensive. I had initially wanted to create an illustration style from scratch, but I was reminded by a peer that since Undraw.co existed, I had might as well just rework those instead.

Referencing back to the text-based dataset of censored prompts I had created, I began drawing out illustrations based on the prompts. It was also during this time that I finally began looking at the dataset in-depth for the first time and realised that many of the censored prompts also had to be rewritten due to many being quite honestly rubbish generated by ChatGPT.

My approach here was quite simple. I looked for illustrations on Undraw.co, loaded them into illustrator, and mixed and matched them as necessary to create new scenes based on the prompts. I eventually ended up with a fairly substantial amount of illustrations before I decided to take a break to blueprint my idea for the actual installation.

Just for fun, I also went to check out the Canva plugin for ChatGPT Plus as I had heard good things about it, but after some testing, I realised that it was just simply a glorified search engine that searched for existing templates within Canva itself.

I more or less had this idea in mind since putting together my submissions during the previous semester, but this time round, I wanted to really lay everything out so that I could see how all components of the project would come to work together in a single view. Looking at this view now, I feel this project is already quite strong in some regard, however there were several flaws that I just can’t seem to figure out right now.

First of all, beyond this little self-censoring AI I was building, there was no real value being created for the end user. It was a cool gimmick for sure, but it didn’t seem to me like it was going to adequately communicate what I wanted to communicate, that being that a potential future caused by the rise of Gen AI is the potential for creativity to be commodified and for generations to just be dull and soulless. I was also still leaning towards and struggling with wanting to provide with some form of call to action for creatives to take to combat this potential future. Lastly, I had initially planned part of the setup around the potential of having my supervisor’s pen plotter available to be for an additional layer of delight and interactivity, but it turned out to be unavailable, hence I needed to also rethink a new replacement for this lost layer.

I ran my concerns and blueprint through my supervisor who acknowledged the concerns while also mapping out the entire project. We were particularly focused on discussing my negative outlook on the future of the creative industries, but he did give me quite a bit to think about regarding what I wanted to communicate with this project.