My Thoughts on AI
I'm trying to blog more since firing this thing back up a few months ago but I've hit my first bit of writer's block. So instead of banging my head against a wall and reaching for content I took to Mastodon and asked what my next blog post should be about. Well @JEBofChristTheLord@deacon.social suggested I do a write up on ChatGPT. So taking that germ of an idea and tweaking it just a little bit, here are my thoughts on AI in general.
First, AI is pretty neat. ChatGPT and others like it are cool to play with and honestly, pretty convincing in their responses. But the real nuts and bolts of things comes down to the Turing Test, proposed by Alan Turing in the 1950s. That "Imitation Game" test proposed was that if a human were communicating with another human and a computer could they tell which is which?
ChatGPT and the other AIs seem to be good at some things but here's the deal: they don't "know" anything. It's just really good at recognizing patterns in gigantic collections of text cultivated from all over the internet. Then it passes that information off to you as intelligence. The answers you get to your questions might sound correct and maybe even sound like someone speaking with authority on a subject. But they also can be completely wrong.
Having said this I could see this a tool that adds to the increasing "information gluttony" of the age. This is the idea what you don't need to know who won the Oscar for Best Picture in 1994 (it was Forrest Gump by the way), or Newton's Third Law of Motion (every action creates an equal and opposite reaction) because you can easily look it up. This has lead us to not really learn information and retain it but instead only how to find that information. I could see a world where AI leads more kids to cheat in school because ChatGPT and others can write a convincing essay on a topic even better than they can. Now, I yield this might be a pessimistic view. I mean, the calculator was likely viewed in the same way. But is this more like the calculator saving me from having to do long division (even though I know how to do it) or like a player piano robbing us of things that can only be communicated through human emotion?
On the other hand, AI can be a wonderful learning tool. If I'm learning how to code I could ask an AI to write me a code that performs the action I'm looking for then use its response as a starting point - properly vetting it to better understand the actual code and changing it to ensure its accurate and efficient for what I'm trying to do.
Either way, the AI doesn't know anything. Since it's just a collection and regurgitation of various sources on the internet at this point it might best be viewed as the next generation of search engine. Even though the response might look genuine, it still needs to be vetted by a human with the ability to understand if the AI is actually correct or not. This kind of "chatbot" AI also isn't really able to step farther in terms of advancing the collective knowledge of humans since it has to stand on existing material in order to compute answers.
The other thing that concerns me about this AI is that when you're asking it questions you're not just playing with it. You're teaching it. Feeding it data. Correcting its responses. Helping it to hone itself for a purpose we don't yet know. That combined with the requirement for an account containing and email address and a phone number makes me give this thing a little side-eye.
Now, do I think AI is going to decide to purge all humans from the Earth? No. Could I see an AI rearranging proteins in a way humans have never considered and end up ending a major disease? Possibly so.
I guess in the end I'm still on the fence about AI in terms of whether it's good or bad. But I know that the "smarter" it gets the more concerned about it I'll become.