I’ve poo-poo’d the very idea of artificial intelligence as just another marketing scheme to get people to buy something. But after using X’s Grok (now behind a paywall) and Chat-GPT (hello, new friend), I’ve finally come around to admitting that it really is some impressive technology. Is it actually “intelligent”? It doesn’t seem that way. They are very good at searching multiple websites to gather data and summarize the findings in a way that looks like human speech.
Using text to “converse” with a computer is nothing new. We did that in the 1980s with software from Infocom, makers of the famous all-text games Zork, Infidel, and Hitchhiker’s Guide to the Galaxy. Those were really cool when I was 10 years old, chatting with my computer and seeing the imagery in my imagination. I still play them from time to time on a virtual machine.
Chat-GPT and it’s competitors have really turned those speech conversation algorithms into something really cool. A new version of Zork could probably access the internet and help me more than it did 40 years ago. These AI tools we have now are basically a mix of Zork technoloy and web search. An AI tool can search and parse multiple web pages much faster than I can, which is invaluable when I use it for my work. Sometimes I come accross an error that don’t make any sense but I can throw it into Chat-GPT and get a breakdown of what IT thinks the problem might be. The problem gets fixed abnd I get the credit. Pretty cool!
So while I’m not quite ready to call them “intelligent”, I do consider these AI tools to be very good at appearing to be intelligent while in the background, it’s a search engine on steroids coupled with a sophisticated speech parser. My hat is off to the people who built these. I know a few programming languages and I can imagine how difficult it was to build something like this. But let us never forget that it’s just algorithms printing text based on the input iut gets from us. It’s greatest attribute is it’s learning ability. Chat-GPT uses feedback from users along with results from the web to increase it’s knowledge base. I can imaging a big part of that is separating the wheat from the chaff. After all, there’s a lot of WRONG information on the internet as well.
As has always been the case with new technology, bad actors will try to use this AI for nefarious purposes. Let us never build an “AI” machine that cannot be unplugged. Literally. Let there be a master power cable and if it’s not plugged in, the whole damn data center sleeps. That would probably take the rule of law to enforce but I think it’s necessary.