Is artificial intelligence a machine we have begotten in our own likeness?
Let’s play a word game. Finish the sentence: ‘The cat is sitting on the ___.’ I imagine you had various options but the strongest contender was probably ‘mat’. However, if I add an earlier sentence that might change. So what word comes to mind if I write ‘Whiskers is in the front room, again. The cat is sitting on the ___.’ Maybe ‘settee’ is now your most likely answer, although there are other possibilities.
Human beings are good at these games. But can we program our computers to play them? It turns out that we can and amazingly so. You may already know this from your phone trying to guess what word you want next when writing a message (to save you typing it).
However, we have now gone far beyond those simple systems with the development of ‘Large Language Models’ such as Chat-GPT and Bard. Give them some words and they will estimate the most likely word to come next, and then keep repeating that process with the result that you appear to get a complete reply from them.
So you type ‘Where is the cat sitting?’ and then wait for the machine to fill in the next mostly likely word. It estimates that word should be ‘The’ followed by ‘cat’ followed by ‘is’ followed by… well, you get the idea.
How does artificial intelligence work?
So how does AI work? There’s not space here properly to explain, even if I had the ability to do so – good explanations are available elsewhere. But let’s attempt to grasp a tiny gist of what’s happening.