“What was Saachshaini Sadchatheeswaran’s research about?”
ChatGPT4o responds with three paragraphs, each one with a specific source link. The first two are generic summaries of my PhD research from the University of Cape Town’s web pages. But the third one makes me pause:
“After completing her PhD, she transitioned into a role as a business intelligence reporting analyst in Toronto, Canada, applying her analytical skills to the financial sector. As of May 1, 2024, she embarked on a year-long sabbatical to reconnect with her scientific and storytelling roots, focusing on the sustainability of urban waterfronts, particularly concerning food systems in cities like Toronto."
source: Saachi Sadcha
I click the source link provided and find myself on the first page of the website on which you’re reading this blog post.
Image credit: Saachi Sadcha
“Huh…I didn’t realize it had access to everything on the internet.” I shut down the chatbot and decide this month’s blog post will be about working with artificial intelligence while I puzzle through CSS this month.
My name is Saachi and I’m on a sabbatical. For those who are new, I would normally give the project’s elevator pitch here, but ChatGPT4o beat me to it. So…moving swiftly on.
In last month’s post, I stated I wanted to build three apps in November to represent Fall 2029 in my project. Attempts were made, but I think I spent this month more wisely by figuring out why I wasn’t happy with what I published last month.
When I published the Summer 2024 apps last month https://saachisadcha.ca/october-2024, I wanted to be proud of myself. I had sculpted, painted, typed, and coded my little heart out. The result was not one, but three working apps. Subsequent testing showed that the apps had issues but worked mostly as intended.
And yet, I didn’t like the physical look of anything I had published. I also didn’t like that the apps worked best on desktops in landscape mode. Ideally, I wanted to use them on my phone in portrait. But even with AI, the CSS - cascading style sheet - that governs the UX/UI (user experience, user interface) of my website, was beyond me.
So I did what I normally do when I’m frustrated - I pulled out my iPad, opened Nomad, and started sculpting and painting quite randomly.
Sometimes the outputs from such sessions are terrible. But sometimes an answer to an unspoken internal problem takes shape. And in this particular session, the ultimate look I wanted for my website came into focus:
Image credit: Saachi Sadcha
This painting of oranges gives me the feeling I want when interacting with this project. Quiet, intimate, cozy, with a streak of brightness. If this was the first thing I saw of a project, I would think the project would be high quality and worth exploring.
Image credit: CSS Design Awards Website
I’m comfortable making individual elements when making art in general but I find composition and layout difficult.
So I put myself through a crash course via this channel:
https://www.youtube.com/@juxtopposed
In particular, her videos on UX/UI and Figma were exactly what I needed.
Created by: Juxtopposed
Created by: Juxtopposed
This developer also created amazing tools, such as Stack Sorted and realtimecolours, which I highly recommend using in the design stage.
Image Credit: Saachi Sadcha, Figma
With this new knowledge now crammed into my noggin, plus my new principles, I opened Figma and refined my website flow
Created by: Saachi Sadcha, Figma
and then created this:
There are many skills required for this project: 3D modelling, sculpting, painting, research, writing, and UX/UI. All of these I do myself. The only skill I outsource to AI is coding. And, in the last seven months, my experience coding with AI has been…enlightening.
Since I’m only using AI for coding, I can test the results instantly, which is great. However, I expect instant gratification, and in reality, I need to do a lot of fixing/debugging. I’m pretty sure an experienced coder would take an hour or less to create what I plus a chatbot can accomplish together in a few hours or even a day.
For example, I wanted to create a minimal but interactive 3D data viz that will encapsulate an oversimplification of this project’s overall message, which is:
Minimizing food waste and turning a concrete harbour into a spongy wetland will help increase food security.
Why? Lots of reasons. But I’ll keep them to myself for now. If this project is done well, those reasons will become clear when I publish this project in full.
Image credit: Saachi Sadcha
The user is presented with a pie on a persuasion map and prompted to use their finger or mouse to drag the pie from the bottom left quadrant to the top right. As they do, the pie grows. A bigger pie means a better food oasis.
This will be the very first 3D interaction that the user will encounter on the site, so it needs to be simple to use and understand.
Created by: Saachi Sadcha
It took three hours for ChatGPT4o (which I pay CAD$32/month currently) and I to build the app, which I have condensed into two minutes:
I think AI is an incredible, albeit ethically questionable, tool even when restricted to coding. It allows a non-coder like me to create a playground to expand the use of the 3D models I spend so much time making.
But it does, like all tools, have its limitations. Most of the time I can’t just use the code it gives me. I have to spend time debugging and breaking the code into little bits to reuse it as needed.
I probably wouldn’t learn so much if I didn’t have to.
That being said, I’ve noticed a clear difference in this chatbot’s abilities in different languages. When it generated React code, the code would work with some adjustments most of the time. However CSS-generated code sometimes took too much effort to fix.
Image credit: Ramzi Bach
So I’m glad I decided to spend 30 minutes every day this month working through this great CSS course by Ramzi Bach:
https://www.slayingthedragon.io/dashboard/css/introduction
I’m about 70% complete and I’ve already learned enough to make up for the holes in AI-generated CSS code.
Now that I have most of the final framework and code complete, I’d like to rework the three “Summer 2024” apps I made in October so that they’ll follow the rules I’ve set above. I also want to complete the three “Fall 2029” apps I started in November. It would be nice to publish the website with these apps embedded and send it to testers. I also want to complete the CSS course by Ramzi Bach, which includes creating a portfolio website (but only with CSS and HTML).
However, December is usually when the least amount of work can be done due to the upcoming winter celebrations so this is a “we’ll see what we can get done” situation.
Thank you kindly,
Saachi
Two days ago, ASAP Science dropped a video about AI's impact on water scarcity: (How much water does ChatGPT actually use?). At about 2 mins they note that based on Microsoft's own 2022 annual report, 3 chats with ChatGPT (about 10-55 questions) could use up 500 ml of water. I'm pretty sure I've had about one chat a day with ChatGPT for coding help, so since May 2024 that's 200 chats. If this relationship is correct, I could roughly estimate that my conversations with ChatGPT4o have used 200/3*0.5L or 33L of water in total. About 10 days' worth of water someone my size should be drinking.
Water consumption to cool down data centers in our digital age, in general, is already very high (link). I was aware that the data centers that support AI require more energy than other data centers, but the actual water cost is new to me. I'm not going to stop using AI any more than I'll stop using my computer or phone, but I want to learn more. Especially because water scarcity is very obviously tied to food insecurity.
Thank you to Erik Chan for reading this blog post and suggesting edits before publishing. And thank you to Sophia Li, Cindy Ngoc Han Doan, Gilbert Pierre for user testing the ‘Summer 2024’ apps and providing much needed and kind feedback
We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.