Some Thoughts about ChatGPT

Trying to get my head around ChatGPT, an artifcial intelligence (AI) program

I have been using ChatGPT, and writing about it on my Substacks. It sure can lead you astray, despite the incredible things it can do though its mystery algorithms.

I am a martial arts instructor, among other things. This information is publicly available. ChatGPT took that information and incredibly mangled it. At first, it had me being a martial arts instructor in California, doing incredible things. Not me. Next, when informed that I was on Vancouver Island, in Victoria, it had me travelling to Europe, giving seminars to the military. Nope, not me. So, its amazing algorithms can combine information it was trained on in the most mistaken ways, and present it as being authoritative.

The lesson is, unless you are knowledgeable in an area of inquiry already, you have no easy way of ascertaining the correctness of the ChatGPT responses. For instance, I asked it to tell me about the development of electric motors yesterday. I have been fascinated in the under-reported phenomena (at least from my sources) of incredibly powerful but small electric motors, as in hover-boards and e-bikes. Not being current on that technology, I have no simple way of knowing how factual the ChatGPT responses were.

So, you do have to be creative in drawing out information to your satisfaction, and just how you phrase your input may result in different replies. However, there is an essential confounder; every time you get it to respond to the same query, it will give different results,: sometimes similar, sometimes radically different, always unpredictable and random.

When you try to get it to present its results with some directions given as to format, it is hit or miss as to how well it will respond to your instructions on the first attempt; every subsequent regeneration will be different, often radically different.

It will get better, and maybe version 4 is already a lot better; I have not forked over the money for that. In the medium term, I expect that it will get incredibly better, unless the whole approach proves to be an interesting dead end.

Will this approach lead to general artificial intelligence? It is a good question, is it not? How would anyone know, since we have only the foggiest idea of how our own brain and body to allow us to think. It would be a mistake to believe that it is only the brain which must be considered, for instance, the nervous system extends throughout the body, and the role of the endocrine system cannot be ignored.

Then you have the most perplexing issue of consciousness. Is it essential to reasoning? Can it arise from hardware? Can it arise from algorithms and neural nets? I have been thinking about that, in an under-educated fashion, for decades. The only think that is clear is that theories abound, opinions are all over the map, and anybody who claims they have it all figured out is not to be believed.