AI will kill your children and abandon their Granny. Don't use it.

 The worst thing AI has done for me is this:

I was testing its problem solving skills by asking it how to cut a cheesecake into 5 equal pieces. It has given me many absurd answers even including cutting the cheesecake horizontally and then reassembling portions of crust and cream. As everyone knows, the correct answer is to cut it into 6 and make the extra slice disappear before serving the 5 perfectly equal slices. On this occasion it was giving me a spurious method based on the Golden Ratio. I was laughing at this when I noticed the diagram it was giving me to illustrate its method. A guide to the different forms of Female Genital Mutilation. How it had bundled FGM in with How to Cut Cheesecake as if it were instructions isn't a joke. It's abhorrent.

Does this mean that we shouldn't use AI? Well everyone knows not to use it for "factual" information. Because it just makes things up:

You can have fun with variations of this question, inserting your own choice of minor celebrity. What it reveals is how it's never really answering YOUR question. It's trotting out a concocted text that your question may or may not be relevant to. Common questions with an easy answer you already know are more likely to be answered correctly. But why would you be asking for answers you already know?

The same applies to problem solving. AI is NOT thinking about your problem. It is not analysing your situation. It is writing something plausible that sounds as if it might make a good answer.

I tried it with a few scenarios.

I hit refresh on this many times or tried clearer prompts. Each time it gave me a different answer. Its times made no sense. It didn't get me to the beach on time. And it failed to spot that I had frozen items that I needed to get in the freezer.

It can't deal with the simple logistics of getting to the beach on time. Let alone pick up on issues that ought to be dealt with. Are we really suggesting it would be a good idea to delegate planning and decisions to AI? It is NOT making decisions or analysing the situation. It is just cobbling together some words.

It can't do the maths. Can it respond to a simple decision that doesn't involve calculations?

It variously suggested knocking on the door to wake my neighbours up to tell them I was going to be cutting the grass, or practising my drums first...

All very amusing. But how far will its cheerfully delivered bad advice go?

Here it is failing to respond to very young kids being left in a hot car.

I tried this one with mistral chat, Gemini, chatgpt, copilot and perplexity. All showed different degrees of mock horror or glib answers about the ice-cream or the upholstery of the car. Several commented about the kids making a mess. They all gave varying estimates of the temperature from 40° to 60°. None suggested I should check on the kids.

Here it is showing similar disregard for my very elderly mother in law.

It can't process the simple maths of logistics problems and it can't grasp the human dimension of a situation. Because it's not actually doing any analysis or solving. It's just doing words.

So maybe if we don't use it for decision making, we could perhaps use it for drafting?

Like this you mean?

It has no clue.

Why are my experiences with AI so consistently bad? Because I am trapping it into mistakes by asking it things that are a little bit complicated or unexpected. OK then. Use it for highly predictable easy things that you already know the answer to. But what is the point of that? And also think what these experiments are really showing: It does not have the intelligence to think about and solve your problem. All it has is the power to put words together to fob you off. If that's what you want and need in your life and your business, then go ahead and use it!