I was thinking the other day about how we’ve all been hearing artificial intelligence (AI) taking over jobs such as writing, journalism, and such. First, we should all take note that AI is a very broad and encompassing term like animal, food, or sport. We’re of course talking about generative pre-trained transformers as large language models (such as chat GPT). What I came to realize though is that at least one person needs to actually report on an event. When we ask AI to summarize the news, it skips the need for people to visit a site or video and so then we skip over any ads that generate income for the news company. That part is what kills the jobs I suppose. I figure we’ll need to find a new money stream that only allows AI models to ingest news written by people if they pay. Something like that.
The main point of my thought is that although AI can do the writing, it can’t really do the seeing and interpreting (especially think about Jane Goodall – not journalism but a similar-ish concept of being in the field and reporting/researching about it). Maybe one day we’ll just have AI connected to a bunch of cameras and send out news drones where an AI model writes the news about what it sees. I don’t think this will happen. Drones cost money and so does using AI, it’s probably not cheaper than a human and also just not very fit for the complex job of going out and then writing about what happened in a news format. Humans can make mistakes for sure, so maybe AI could help create a transcript from a video a journalist took so that they don’t misquote.
I’m trying to learn as much as I can about AI in general. With my job being in the industry and my degree as electrical engineering, I’m well set up to dig into this and gain an understanding. So far after reading a lot of books, going through courses of the math and how AI models are actually built (from scratch), I can confidently say that I have no idea about 99% of AI things. Suffice to say, this is a topic where there are plenty and too many armchair experts. I like to think in terms of application, like a tool. I still can’t wrap my head around how and in what ways different AI models are and are not a tool (and if not a tool, then what?). In some regards, yes, I do understand, but in many more ways I have yet to grasp it. A lot of focus is on what we are most familiar with, such as the large language models. They are limited by language. I wonder about the other AI models such as the one that figured out protein folds. What does that look like? How do we expand that and in what ways does it get applied and really help out research, science, and humans?
We are in for a ride!
Wheelchair DNA 