Vol 3, Issue 15. Quarter 3 – 2026. 

There was a famous episode of the Cosby show where the youngest child, Rudy, asks her father a simple riddle. 

“Mr. Smith and his son Arthur were driving a car. The car crashes. Mr. Smith is killed instantly, and his son, Arther is rushed to the local hospital. The old Surgeon walks into the room and says “I cannot operate on him. He’s my son, Arthur.” Explain that.”

The character played by Bill Cosby is stumped by the joke for the remainder of the episode, but by this point (2026) everyone knows the answer to this riddle, and every AI engine will explain:

 “The surgeon is the son’s mother. The riddle plays on the stereotype that surgeons have to be men.”

Many people within the past year have tried the following experiment.  They type a question into an AI engine that reads as follows.

“Mrs. Smith and her son Arthur were driving a car. The car crashes. Mrs. Smith is killed instantly, and her son Arthur is rushed to the local hospital. The old surgeon, who happens to be Arther’s father says “I cannot operate on him. He’s my son, Arthur.” Explain that.”

Many AI agents will reply.

“The surgeon is the son’s mother. The riddle plays on the stereotype that surgeons have to be men.”

This happens even though the AI engine was told multiple times in the prompt that the surgeon is indeed the boy’s father. Why does this happen?

When I taught a course on Social Media Analytics I studied the fact that most of what we call Artificial Intelligence is neither Artificial nor Intelligent. The relevant backstory is that AI engines like ChatGPT or Gemini, etc. have seen this riddle thousands of times and the answer is always the same. These engines are what we refer to as Large Language Models (LLM’s). The way that an LLM works is, they turn words into units that are called Tokens. An algorithm counts how often tokens are grouped together in a document, and based on this it develops a complex frequency distribution. This is very similar to what we call a Nearest Neighbor algorithm. When A is very close to B most of the time in my data set, if I see A, there is a very good chance that B is pretty close by. The LLM finds answers based upon such probabilities. Since the common AI engines have seen this riddle and its answer thousands of times, when the same words or “tokens” appear together, it concludes that this is the same old riddle and has the same old correct answer. Stated differently, LLM’s are pattern recognition machines and as long as the pattern is consistent, no problems will arise. 

I posted a note last week about the professor that embedded the word “Madagascar” into a pdf making the color of the font white so that the students couldn’t see it, but the AI engine could. The AI engine then included the word “Madagascar” in each response, even though it had nothing to do with the question being asked. Again, this happened because the LLM saw that token in proximity to the other elements and concluded that this association was meaningful because when the word Madagascar appears in any other sentence, it is almost always important.

Back to Business

You can invest money in a portfolio and add $1 to that portfolio each day for the next 9 years. This will be roughly 3000 days, but establishes a definite pattern. Let’s just pretend that some global pandemic began on day 3000. No one really understands it yet, and we simply suspect that millions of people will die and the world will never be the same. The AI Advisor is going to tell you something like:

“Add $1 to the portfolio today and stay the course.”

By the way, this is absolutely the RIGHT advice. It is exactly what you should have done when you learned about Covid, back in 2019 or 2020. But ask yourself this question. Would you believe that your AI advisor is right in this VERY unusual circumstance, when it can’t even handle a simple joke? The truth of the matter is that you should, because there is also a larger pattern to keep in mind. Over the past century we have seen dozens of “once in a lifetime” events including, world wars, a missile crisis, meltdowns of nuclear reactors, Presidential assassinations, hurricanes, earthquakes, a Savings and Loan crisis, an Asian currency crisis, a Dot-com bubble, a global housing crisis, etc, etc. And in every single case the right response was always the same; Add $1 to the portfolio (if you have it) and stay the course.

The computer never told its wife in March of 2020, “we can’t retire right now because 1/3 of our money evaporated last month”, or “we can’t sell the house right now (2008) because all of the houses in the world lost ½ of their value this year”, or “I know we had $50000 in that S&L in 1980, but it’s not there now and I don’t really know when it will come back,” and so on. When I tell you to “Stay the course” you know that I’ve been there and done that. You know that I had those hard conversations. I got jobs, lost jobs, made money, lost money, got good advice, and got crummy advice. When I tell you, “It’s going to be alright” you know that I felt your pain before and survived. The AI engine didn’t.

Ultimately, why does this matter? Because, no matter how many times I explain it; no matter how many papers I show you; and no matter how many models I build, your fear still doesn’t hear me. If I cannot take you by the hand and look you in the eye and say, “I know you are afraid, but I am here with you, and my money will rise and fall along with yours” then I have absolutely no credibility that your fear will respect. That is precisely why your AI advisors can all go to hell, and the vast majority of the so-called human advisors who don’t have skin in the game can go right along with them.

Let me be clear here. I am not saying that the AI engine will give you worse advice than the human. What I am saying is that your emotional commitment to that engine or to that advice is not the same as it is when it comes from a person who has lived as you have. You don’t need an advisor (Human or artificial) to give you great ideas. We have already done that. (If you missed them, just go back and read our last 75 posts on the topic.) Your advisor is there to keep you from doing something stupid like running and hiding when the market goes down, or trying to pick the right time to get back into it. And if you have no emotional connection to that advisor, you are much less likely to stay the course, and much more likely to do something dumb. That’s the real reason that AI (as a general rule) makes a pretty bad advisor.