My thoughts on AI tools, AI generated art, and why AI will *probably* not take your job.
“Will AI take my job?”, is a question a lot of people have been asking recently and a question that is too complex and nuanced to be answered in this blog post. Instead, I will be discussing my teams approach to a more reasonable question – “How can AI make my job easier?”
To answer this question, my team set out to note every relevant AI tool we could find that would be relevant to taking a programming project from start to finish. This was no trivial task, as with the recent uptick of AI popularity there has been an equivalent surge of new AI tools ranging from shockingly good to, well, just as shockingly bad. The project we chose to design was a web-based 3D Escape Room, which was chosen partially because it will test a wide variety of AI tools and partially because it’s just really cool. In order to complete this project, we will need to utilize AI tools to create the story/theme, generate art, code the front-end/back-end, and just give us some guidance along the way.
The biggest hurdle we faced from the get-go was cost. Most of the AI tools worth their salt had some sort of pricing plan. Midjourney, Github Copilot, and ChatGPT Pro are some of the most notable AI tools that require a subscription. In my opinion, those subscriptions are worth the cost though. Those products are already in a state where they can provide significant value from the start; however, a lot of tools seem to be lacking that. This is especially true for a lot of the AI art generating tools. They all require some sort of credit system for generating pictures, which is totally understandable from a business perspective, but annoying for a student trying to conduct research. For example, when generating the picture for this blog it wasn’t until about the 20th attempt I got something I liked. Luckily, my generations were free, but coming from someone who absolutely never uses limited resource items in video games – I don’t like this system.
Another hurdle we faced was that generative 3D art is still very much in its infancy. When we were first compiling all the AI programs, we had multiple 3D generating programs and it looked like we had plenty of options to make it work. Upon closer inspection we realized that many of these tools were too specialized to work for any of our purposes. One example is GANverse3D by Nvidia is only trained to produce 3D models of cars – specifically sedans. To say I was disappointed is an understatement. Although using AI tools to generate 3D models may be out of the question, tools like ChatGPT can still greatly assist the task of creating and implementing these models in tools like Blender.
In summary, generative AI art is extremely impressive but not in any state to replace human artists at the moment. It does an amazing job at providing quick inspiration or suggestions, but most of the time it can’t really give you a finished product. This can also be said about the code and text generating tools. What the AI can give you could be exactly what you want, or more often leave you to fill in the gaps with human insight. I think we live in exciting times and as AI continues to improve, I think humans will grow with it. After spending hours researching and using dozens of AI tools, it is my opinion that the best way to use them is to supplement your own work – not replace it.
Leave a Reply