Reflections about AI and genAI.

People seem to think I have some abject disdain for Artifical Intelligence (AI) and it's specialised version Generative Artifical Intelligence (genAI). I don't dislike the concept of both of those things but there's a certain "weirdness" to them that I feel needs to be brought up.

To clarify my views, I don't think AI and genAI are inherently bad things. I just dislike the seemingly corporate need to shove it into every single possible context possible, that just detracts away from doing actual work.

Before I do that, I want to talk about the class we had as part of ITLI102 regarding this topic. It was an offline class, and even though it was at 8 in the morning, I was somewhat looking forward to it.

The guest professor for the day was really excited to talk about this technology with us. Of course he uses chatGPT as his reference point, which while I get the reasoning behind, felt really reductive as a choice of model.

He firstly gave us a brief historical overview of the term AI itself and how it's development shaped up from the mid 20th century to the early 2000s to what it is today. Everything from who coined the term, the hardware capabilities it needed. The hoops the term itself and the development of it had to jump through to get the recognition it has today. It was really interesting to see it's original algorithmic usage within the medical field.

I'm glancing over the history part of this class mostly because it was all stuff I knew. I've worked with modern reimplementations of ELIZA and I know what a Lisp machine is. That should tell you enough ;).

Further ahead, we discussed the advatanges and disadvantages of using AI tools like chatGPT. He asked us about how we used it as students in an academic setting such as the assignments we had to do. He talked about how he also used it now in his life and how it's acted as a convenience tool to help him with menial administrative work.

"Academic Settings"? Are you telling me you openly talked about how you use AI to get your assignments done?

Hey don't look at me, I don't even use AI for anything. Everything I do and write (including the content here) are all certified free range thoughts.

Towards the end of the class, we actually got to use some of the in house tools developed by the professor and his colleagues. The first one was a game of using genAI to create an essay about Taylor Swift and her "Eras" tour. The professor's objective was to guess whether an essay we prompted to generate was AI or human. I would've probably done better if I had upped my general awareness about Taylor Swift.

We also got to engage with an LLM designed towards language learning. I chose to interact with the French language bot since I've been learning the French language here at uni and earlier at school as well. I gotta say it was pretty impressive in its ability to correct me about the things I was doing wrong. I purposefully threw some incorrect sentences at it and it managed to corrrect them all. Though I do have to say that beyond holding a conversation it's not very helpful, it couldn't answer a question posed to it asking how to say a particular English word in French.

Going back to my views on AI, I still stand by what I said. Higher ups at various companies seem to think that AI is the only future, and so they feel the need to insert it everywhere, automate everything. It simply just doesn't work out and then they choose to blame developers for it. I'm not even concerned about the idea of it taking jobs, I'm moreso just concerned about how we can't even train models on data that are ethically sourced. I know artists who've had to "poison" their artwork to prevent their art style being integrated into art generating AI models. This is just one example of many cases.

I do think that there are places where AI is being used as it should, as a tool to assist not create. Zoom recently rolled out a new feature that lets an AI model generate 'minutes of the meeting' for conferences, and it's actually pretty accurate to a degree. Taskade has an AI model that translates natural language into their in-house query language. These are all good uses of AI.

I still think it has a ways to go, but we're getting there and I stake no claims on what the future holds for us.