Becoming Artificially Intelligent
Hi Friends!
I’ll be the first to admit, I’m not a particularly big fan of AI. In fact, I find most current uses of it to be incredibly obnoxious and things like AI phone systems almost instantly send me into a blind rage. In my opinion, there are virtually no applications in which AI is currently being used that is making the world a better place. I also feel like AI sets all kinds of potentially dangerous precedents and is a software upgrade that humans are just plain not built to handle.
Despite this, it’s clearly not going anywhere, so to a certain degree, I am begrudgingly starting to force myself to make peace with it. I figure there’s really only two options here: I can either learn it, understand it and stay current or I can fight it, ignore it and get left behind. This is where my relationship with AI basically starts. I’ve decided that I’m not going to let AI train me, but I’m going to learn how to train AI instead. To accomplish this, I’ve started at the bottom, with the basics. Primarily, ChatGPT and Perplexity.
After about a month or so of goofing around with both of these, I’ve pretty much nailed down that they are basically not much more than glorified search engines. They scan the internet looking for data related to whatever you’re asking it look into, combine sources and spit out an abbreviated version in plain language which you can use to solve whatever problem it is you’re trying to solve at any given time. Then it learns based on what you’re asking it, how to serve your needs better. Even if you ask it opinion or personal type questions, that’s really all it’s doing.
This kind of got me thinking…..
As many of you know, I’m really interested in mid-century Italian racing bikes right now. I’ve done things like preservations, but now I’m in the very early stages of performing a restoration on my 1956 Gilera. The fact is, these are obscure motorcycles, there’s not a lot of information out there about them and the information that is available, is often times not in English (usually Italian or Spanish as a huge community for these motorcycles exists in Argentina).
I saw post on Instagram a couple of weeks ago that was basically a tribute to the early 2000s forums. The comments were filled with people rattling off what their favorites were and many more acknowledging that despite the fact that they may not currently be active, they are still used and are still filled with valuable information. Fortunately, a lot of the bigger ones have been preserved (thanks to companies like Internet Brands for example, who continue to own and operate several dozen of the more notable communities), but admittedly, it’s getting harder and harder to sort through posts that are in some situations two decades old or older and on platforms that haven’t had updates available for years as well.
It seemed to me that this could be something that AI could handle really efficiently, so I started testing my hypothesis. My first question was “how many parts on a 1956 Gilera 150 Sport require paint?”. Almost immediately, ChatGPT came back with a number. I’ve been further refined and asked what color each of the parts needed to be painted and what modern paint codes could be used to achieve a factory finish. Within an instant, it spit that info back at me and with pictures too. It even asked me serial number related specifics on or to help refine further.
Needless to say, I was impressed. I don’t think I could have figured any of this stuff out on my own over the course of a month. ChatGPT did it in under 30 seconds. Since then, I’ve been sort of putting together a general layout of how I want to proceed with this restoration and I’ve been able to figure out a huge amount of things, alarmingly quickly and put together what looks like a pretty efficient game plan. I think a little bit of an asterisk needs to be applied here as I’m not sure how much of this information is correct. I have not really gotten going yet and there’s a high likelihood that ChatGPT could be wrong here, but so far, it’s pretty amazing.
For the last couple of years, I’ve been running an ad on Instagram non-stop encouraging people who have knowledge of niche vehicles, makes and models to document their work and to put their knowledge online for future generations to be able to take advantage of. We talk about having a staffing crisis surrounding trades people and restorations and maintenance of classic vehicles. Of course, we always want to encourage real people to go out of the world, get involved in things like this and become those resources themselves, but we may not get to a point where replacement is realistic. I think this little discovery underscores even further the importance of people putting knowledge together and sharing it openly. If anything, it’s proven that platforms like ChatGPT are at least in the process of learning to sort through the noise and quite likely could provide the guidance needed to keep this stuff on the road for more generations to enjoy.
I’ll keep you guys posted on this……
Side note: This week, I appeared on the Kidd and Friends podcast. If you want to check that out, link is here
That’s it for this week……
Darin Roberge
