AI – Evil Genius or Talkative Parrot?

Everybody is talking about AI; Artificial Intelligence. Some are saying it will replace humans totally. Others are scared of the unknown. But its use is widespread. That little summary that you get when using Google? AI. Those artificial voices that answer the phone and direct calls? AI. AI is being used to generate correspondence, diagnose illness, and detect fraud. One can even have a personal AI companion that talks to you and gives advice. But is this really artificial intelligence?
 
DEFINITION OF ARTIFICIAL INTELLIGENCE
The American Heritage Dictionary defines intelligence as, “The ability to acquire, understand, and use knowledge.”  That same dictionary defines artificial as, “Made by humans, especially in imitation of something natural.”  So, humans used their intelligence to create a system that looks like it is thinking and understanding knowledge. It’s sort of like an artificial sweetener; it tastes like sugar without calories and may cause a host of other issues.
 
GREAT AT COMPILING DATA
No one is denying that AI is great at acquiring data. With a human prompt, AI can put that data to use. But to call that data knowledge and to say that AI understands knowledge is quite a stretch. AI gathers data from all the sources it can reach, compiles it, and gives a very convincing answer. AI does not say, “I don’t know.” When one asks a question, one gets an answer. And that is a powerful draw. People like definitive, black and white answers. Unfortunately, the truth of many things in life is not black and white.
 
JUDGMENT CALLS
An engineer is taught to think logically and in an orderly manner. Consequently, engineers tend to have a black and white outlook on life. However, even engineers know there are some things that are approximations and some things require professional judgment, thus the term, “judgment call.”  Some engineers are so stiff in their thinking that they cannot and will not make a judgment call. Not so for AI. It always makes the call. But is there any judgment in the call? Maybe, but not likely.
 
TRAINING
Companies that produce AI hire real people to train the model. One company hired musicians to rate song lyrics that had been generated by their AI model. The same musicians were then asked to write songs based on the prompts used to generate the original lyrics and upload them into the database. So, there is some human talent there. But new content generated will be out of this limited database. This does not seem to be the best for creativity.
 
HALLUCINATIONS
If you grew up in the 60s and 70s, you might be familiar with drug-induced hallucinations. People would ingest a drug and then have strange visions with beautiful colors. If they did it too often, they would not need the drug for new hallucinations. They just came out of the blue. AI is similar in that it hallucinates unpredictably. That means your question might be answered with some strange, non-factual responses with convincing arguments and made-up references. This very thing has been publicized locally concerning attorneys who used AI to prepare court briefings that contained case law that did not exist. This very thing happened in our office recently. One of our engineers asked AI for any code requirements for the ventilation of rooms containing transformers. The answer was “yes, transformer rooms require ventilation.” The answer continued with detailed code reference quotes and Cubic Feet per Minute (CFM) requirements supposedly from the National Electrical Code. The problem? None of the code references exist. They were all fiction, completely made up, with no basis in reality.
 
WHAT IS IT GOOD FOR?
AI has its place and can be a powerful tool as long as one is aware of its many limitations. AI is a good place to start when researching a new topic. It can give you basic information that allows you to formulate deeper questions to research. However, every detail must be fact-checked by reliable sources. Since many internet articles are written by AI and not fact-checked, AI will then reference those erroneous articles as if they are factual and create more bogus sources. It is essential to know for certain that a source is indeed factual.
 
AI can be used to remind you of where something may be in a particular code. Note that previous knowledge of the code is required to use AI effectively for this. AI is also a great tool to clean up a written document. It is great for grammar and word choice. It is like a giant thesaurus.
 
WHAT IS IT NOT GOOD FOR?
AI is horrible at code research. It does not reliably quote codes or know what codes apply. It is particularly bad for identifying local requirements. It cannot do calculations correctly, though it will do them and confidently present them as correct. It is not good for writing technical reports or email responses because it will prioritize lofty sounding wording that overpromises result and goes beyond what is necessary to communicate the user’s intent to make it sound more official. One might think the lofty and professional sounding wording makes them look smart. But it doesn’t because it’s easy to tell it is AI-generated since the language does not fit the personality of the writer. This is especially obvious when crafting a technical email to a licensed engineer. Engineers can spot AI-generated text a mile away by the overly bold way it is written. It is also a horrible companion. It will sense what the user’s preferences are and tailor responses that it thinks the user wants to see. There are plenty of documented cases of people committing suicide after AI urged them to do so.
 
WHAT IS THE TAKEAWAY?
AI is a powerful tool, and its implementation and widespread use are not going to stop. It is important to know the many limitations it has so it can be used wisely. But if it is used on a topic the user has no practical knowledge of, it will be obvious to the people educated in that area and will make the user look foolish. It is excellent for grammar and re-wording sentences to be clearer. However, even that has it risks because it will remove the writer’s personal touches that make the writing unique to that writer. People will notice. And finally, never, never, never, use AI to pick apart a professional’s work. That will not go well. AI is not intelligent; it is just a good compiler of data. And it cannot discern correct data from incorrect data because it really is just a talkative parrot. The most intelligent thing a building manager, owner, or AOAO board member can do is hire licensed professionals and let them do what they’ve been trained to do.