AI: A Year Later
In late 2022, ChatGPT was released and the expectations were sky high. The technology was amazing. You could ask it to write a paper on Shakespeare in the language of a pirate and ten seconds later it pops out. You could ask it to program a Crestron control system with an Epson projector, and again, 10 seconds later you have a program. I was very bullish on what AI would provide us, and while I still am, I have a more nuanced understanding of the technology after a year of using it and learning about it.
Another higher ed technology manager, James King, likes to say that the AI we see right now is nothing more than really sophisticated programming. I have mixed reactions to this stance, but one can not deny it is thought-provoking. My first reaction was that this statement diminishes, or causes us not to fully imagine what AI may be doing in the near future. The more I think about the statement though, the more it resonates with me. After all there are some glaring problems with AI that show us it can not truly think for itself. For example, when creating images AI is notorious for not understanding that the average person has two arms and five fingers on each hand. It has gotten better with faces, but they tend to be blurry and awkward as well. It simply can not make good digital representations of a human. Perhaps the most consistent example of what AI currently does is asking it to do a math problem. As easy as it should be for AI to compute math, it gets it wrong. Why? Because AI is not actually doing the math to solve the problem. It is a sophisticated algorithm (sophisticated programming) that searches the internet for a common solution and tries to apply that. In this case it seems to be a beefed up search engine. Rather than give you a solution that may solve the problem, it determines the solution that may solve the problem and applies that solution without actually knowing if it is correct or not. When we ask AI to write a paper on Shakespeare in the language of a pirate, again, it is not creating any new information. It is not reading Shakespeare and making genuine commentary on it. Rather, it is indexing millions of other people’s thoughts on Shakespeare, combining them and spitting them out, Matey!
In an AV realm, you would say that auto-focusing or auto-tracking cameras are examples of AI. It could also be argued that really the cameras are just doing what they were taught. The software was taught what a human face looks like and to track that face. Automatically adjusting audio is similar, it knows the limits and levels, maybe even uses sensors to judge levels in a space, but it is only doing so inside of the limits and boundaries that someone has taught it.
On the other hand, what AI is doing and the speed at which it is doing it is still amazing. It’s powerful and it will change the way we work, live and learn. This makes me wonder about making a statement like, it is only really good programming. This seems to have the possibility of stifling creativity and imagination in what we can do with AI.
The concept of AI as “sophisticated programming” will continue to further drive the integration of modern programming languages, data and AV. AI will not replace our jobs in the AV industry, but it will revolutionize them. We will all become much more sophisticated with traditional IT knowledge, and we will be integrating systems and databases on a more regular basis than we will be programming interfaces. Perhaps, after my first full year of exposure to AI, I am prepared to take the very safe middle ground. I agree that AI is sophisticated programming. Rather than diminish the power of AI, I now believe that statement empowers us to better understand it, and use it in the work we do.