That gift horse… it’s been around for a long time now… Actually, when you think about it, AI is just a gift horse. That’s why you should really ‘look into its mouth’.
Where does that saying come from and why? A gift horse is one you are presented with, for free. Very old horses though originally were considered to be less valuable. Partly because as horses in farming they wouldn’t have power enough. For riding they might also be unequal.
In order to ascertain the age of a horse you look at its teeth, into its mouth, and with experience and knowledge you would know its age with great precision then.
A gift horse comes free. So, don’t look too closely, it’s exactly that: A gift we would want to be grateful for.
In business, people may appear to give something away. But there’s always a reason behind it.
In the case of the major AI applications and chatbots, such as Google, ChatGPT, perplexity.ai, Gemini, Copilot(s), Apple Intelligence and many more, apparently published almost daily these days – they come with a ‘long tooth’:
When you use them, they gather your data, any data you leave, by asking, by searching, by disclosing your whereabouts or even the cookies stored inside your browser. At least their own.
More importantly even:
-
- When you know about a subject well, such as your profession of long standing, you will easily be able to judge where AI goes wrong.
- On the other hand, you cannot possibly be completely sure, if you are still learning.
How Do I know? And Why?
I have documented AI successfully, with the support by a mathematician at the time and the Product Owner and Designer, IT business. That documentation even was mentioned specially by the US agency Gartner in their Magic Quadrant for Metadata Applications in 2019.
AI per definition does and can do these things:
-
- It is programmed for finding and using patterns of – and inside – knowledge in the shape of data.
- This kind of pattern recognition is based on mathematical probability calculations: Statistics.
- At first, known data.
- After it’s been trained, it will be let ‘loose’ on other, more and unknown data to again find patterns.
- The next step is the re-assembling of those patterns in a likely manner.
- Or making predictions based on probability: “How likely is it that the cases A or B will occur again, after they had occurred before such and such number of times?“
- It is programmed for finding and using patterns of – and inside – knowledge in the shape of data.
This means that
-
- in cases where you are sure of your ground or the expected results and can double-check them, you are save ‘trusting’ AI.
- Equally, when you ask about a known text or data such as an image or given page/file, it will use that.
- But as soon as you would ask it like a search engine to look for answers online, you may find results be faulty.
- AI will never ‘tell’ you so. Because it is not able to. It does not recognise meaning. It finds patterns in texts and words.
Reassembling them. - Therefore it will always ‘tell’ you in so many words that everything it ‘says’ is of course right. True.
- Online sources can be anything from a Wikipedia article through to an advertisement on some page mentioning the phrases you were entering in the chat’s search box/field.
- AI will never ‘tell’ you so. Because it is not able to. It does not recognise meaning. It finds patterns in texts and words.
The conclusion to all of this:
Look into any gift horse’s mouth that is an AI you can use for ‘free’!
