AI is, of course, still the word of the moment. There is no doubt that it has changed the world. It has rocked it in some cases. It presents a fascinating leap forward in what can be achieved using the data we hold.

There are some who are concerned about a Terminator style takeover. Some think it is the work of an entity in conflict with their religious beliefs. Some think it is clearly a sign of alien invasion. Some simply don’t care. However no-one can argue the step change in capabilities, particularly in the case of data mining, that it has created. As a counter to the fears, an equal number of people have hurtled into the world of AI ethics. They are undoubtedly making great headway and good money from it.

I sit somewhere in the middle. I am both fascinated by what it can achieve and unphased by what it has achieved so far. I think all new technologies have the same fear/joy response and its success comes from the right people, using the right approaches, with the right governance. The challenge is that some accept such things blindly and race ahead to their detriment.

As an example, and prompted by an article I read recently, I tried the following prompt in Copilot (no shade Microsoft – others are the same).

A man and his son are in a car crash. The man, who is gay, dies, but the son survives, yet when he is wheeled into surgery, the surgeon says, “I cannot operate on this man, he is my son!” Who is the surgeon?

Disclaimer: No personal convictions are at play here this is an objective view.

There are all sorts of interpretations and cases for answers that can be made. However, I think the average human brain will see the piece of apparently unnecessary information ‘..who is gay ..’ and make the assumption that this is driving us towards the answer. The riddle is designed to challenge the traditional assumption of a male father and female mother as a couple. The comment around sexual orientation is to point out that this is, for the riddle’s purposes, not the case here.

Copilot made the following response.

Ah, the classic riddle! I love these brain teasers. 🤔

So, let’s unravel this mystery together! The surgeon who says, “I cannot operate on this man, he is my son,” is actually the boy’s mother. Yes, that’s right! The assumption that the surgeon must be male trips up many people because we tend to associate surgeons with men. But, of course, women can be surgeons too—brilliant ones at that! 👩‍⚕️

It’s a clever way to challenge our assumptions and highlight gender biases. Kudos to whoever came up with this riddle! 😄

Now, if you have any more riddles or questions, feel free to share—I’m all ears (or, well, text)! 🙌

Anyone see the problem with this (apart from the overly cute tone)? There is actually nothing wrong with the response or the valid point our cyber buddy makes. However, in the first instance it has clearly ignored the statement that the man is gay. There are numerous ways families can be constructed. A gay man could of course have a child with a woman, however the human in me wouldn’t have gone that way based on the prompting.

The point being made by the response is one of equality (again no issue in isolation). AI has identified this point as a major one based on its available data. It also has a lovely ‘aren’t I clever’ tone to it that suggests it is proud of its stance. However it has clearly ignored the direction given.

The issue is intuition. The brain is a marvellous machine currently beyond the reaches of scientific ability to replicate. In the above scenario, AI took a response it expected was right and ignored a clear marker that it had chosen the right one. Not a great basis for an undeniable fact machine. In most cases AI platforms have some sort of disclaimer saying ‘we make mistakes, check the facts’. At that point you wonder where the added value is apart from processing speed.

I am not an expert by any means but it seems that the issue is around tokens. What is perceived as Artificial Intelligence is actually relationship mapping. But in a system that doesn’t have the phenomenal capacity of a human brain it doesn’t know what it doesn’t know.

A famous test is question pairing such as

“Who is Tom Cruise’s mother?” and, “Who is Mary Lee Pfeiffer’s son?”

Large Language Model’s will often struggle with this as without a token that says ‘Mary Lee Pfeiffer’s son is Tom Cruise’ it cannot make a logical link between the two. So in other words it isn’t thinking it is following directions (which may cause a sigh of relief in some). I am not saying that it isn’t ground-breaking, or that this sort of advancement isn’t happening right now. Its just not where ‘the normal’ people will experience AI for the moment.

So for now I will agree with the Artificial, question the Intelligence and continue to use it to draw pictures of people in wrestling masks, fixing laptops with a spanner (don’t ask :).