This article was written by Don VanSyckel, the club president, as a part of "The President's Pen". This article appeared in the January 2024 WYSIWYG newsletter.
Artificial Intelligence (AI)
by Don VanSyckel
SHCC has had a presentation of ChatGPT and this month we have another presentation simply titled "AI". Over the last six months I have been to several Artificial Intelligence, AI, presentations. It has become apparent to me that AI software can be a good thing or a very misleading thing.
AI background, like other software packages such as word processors and spreadsheets there are more than one AI software package out there. I have hear people state that the output of AI is repetitive and bland and that an AI output might make a good starting point but it needs the personal touch to edit it into something you'd be proud of.
A couple of factual points about AI output. First, AI systems do not invent any new information, they combine information that is found "out there" someplace. Some AI systems search a specific (limited) number of web sites to harvest information from. This limited number of web site makes the collected data more manageable but it also limits it in scope. This harvesting can not be done instantaneously so the pool of information that a given AI package is using is dated. Whether it was updated yesterday, last week, or last month, it is not up to the minute. This lag might have little to no impact on a report from an AI package. Information about the first US moon rocket is years old and new information is not currently being added. On the other hand, a report about some current event such as bank branch office closings could be missing the latest information due to the information harvesting lag. I suppose some AI systems are constantly harvesting in a cyclical manner, it just depends on how long the harvesting cycle is.
Another point that isn't mentioned often is who is writing the AI software package. We are all aware that if you use Google for web searches you soon are inundated with ads related to the searches. While simple, this is a form of AI. Actions are taken based of observed behavior. Suppose an AI package used by your doctor to diagnose you has had a bias built into it to recommend the drugs from a certain pharmaceutical company. Now the drug your doctor prescribes for you might be the actual second choice, even the third choice, or lower down on the list appropriate for you. AI can also be use to bias news articles supplied to you in all type of areas, politics being just one area. AI can supply you with positive article about one candidate and negative articles about the other candidate, thereby skewing your opinion. It has been documented that some of this has actually happened. The other bias is built into the list of web sites that a particular AI system uses to harvest information from. The AI vendor will include web sites leaning in the direction they want you to go and not include web sites that do not lean in their chosen direction.
Another point that I've heard more than once is never, never, ever put any personal information into an AI package. The AI package can "keep" all the information that you put in. You know how *#%^@ (fill in your own word) people are, they just don't think or they are totally unaware of what they are doing. You don't want to give the hackers a place to start in their attach against you making it easier to steal your identity than that of the next guy.
The above is not to say you shouldn't use AI, but you should use it for some things and not for other things. Like with any tool, lawn mower, hand drill, or automobile, you need to use AI properly to stay safe.
End of Article