Down Another Rabbit Hole

I love learning about new things.  I love learning about old things too.  I also love learning that some of what I thought I knew was incorrect once I learned more about it!  That may seem strange -- learning you are wrong -- but I would rather correct my incorrect knowledge than continue blithely in ignorance.

I have only minimal knowledge and likely some misconceptions about AI as it is currently being presented, but I have dived into that massive rabbit warren with both feet.  I suspect a Level A Hazmat Suit might be in order, but I'm not at that level of education on the subject as of yet.  You might take this paragraph as a warning that the subject will be reoccurring at times.

I have missed a couple of days of writing due to unexpected circumstances, but I hope to dive back in on a regular basis again.  Certainly, there will be occasional disruptions but hopefully they will be kept to a minimum.

When I begin looking into AI it is immediately obvious that it will be inescapable.  I deal with it as I type this blog entry.  We have dealt with it since some of the earliest versions of Word Processing programs.  Spell-check, Grammar-check and the words that magically appear ahead of where you are typing in a sentence in anticipation of that being where you were headed are some of the most basic versions of it.

What we are calling AI (artificial intelligence) is, to date, machine learning.  We are quickly advancing in the capabilities presented by machine learning but, in my opinion, we won't achieve true AI until we begin to see creativity and original ideas that are beyond the scope of current or historical human endeavor.  What we are seeing today is simply better organization of data, accumulation of massive amounts of data and mining that data in different ways.  It is still necessary that a human ask the correct questions.  That is part of what makes it so dangerous in my opinion.

Exploitative uses of data are just as likely, perhaps more likely, than benign or beneficial uses of data.  Analyzing the incidence and prevalence of measles cases throughout the world could be useful in prevention, or it could be used to enhance the spread of harm through the weaponization of some pathogen.  There must be some level of moral control on the use of AI.  That is scary because who is there that is trustworthy to develop such controls?

We have a tree in our front yard that I assumed was a Bald Cypress (Taxodium distichum).  I confirmed that identification using iNaturalist which is my "go-to" identifier.  But when I looked up images of Bald Cypress using Google Search, I didn't see any that closely resembled the tree in our yard.  It made me question the identity.  So, I went back through the process again and came up with the same answer.  I did learn that there are over 400 varieties of cypress trees around the world.  I also learned that their wood is great for building due to disease resistance.  But because the images don't really match up there is still question in my mind that my identity is correct.  I think that is a danger we face with AI.  The answers derived through its use should be closely scrutinized.