Yippee! I don’t have to work anymore! AI is going to take all our jobs!
Should we be terrified? Or should we be thrilled? Perhaps neither. Oh wait, didn’t I write this in my last article? Is AI more or less repetitive than old people retelling the same stories over and over again?
There is no intelligence in AI. That statement of course leads us into needing to define intelligence. Without diving into that topic, what I’m emphasizing is that AI lacks creativity.
Now you say: “Wait John, Stable Diffusion, Midjourney and the others are able to create amazing original images… art!”
Play around with it yourself and you’ll discover the failures outnumber the successes. It turns out that a million monkeys randomly typing do actually produce “Shakespeare” sometimes. If you understand the underlying technology this is actually how it works (but computers are better trained and far faster than monkeys). If you need a new conspiracy theory: What if AI image creation is nothing more than a huge database of so many images that you just think you have never seen the produced image before?
Is a toddler finger painting an artist? Perhaps more so than AI image generation! We could certainly debate the quality differences though!
One of the common phrases I’ve heard about development of AI is that “we don’t actually understand how it works”. This is an interesting point in time where humans are building things where no single person actually understands everything about how it works!
Computers fail at thinking, but they excel at making decisions. Perhaps this is what we actually want? A “Thunderhead”, such as Neal Shusterman imagined, to make decisions and run the world with flawless precision and no corruption. Wouldn’t that be desirable?
In my line of work (my primary career), software development, there is a lot of excitement about using AI to write software. Perhaps most people think of writing software as some magical arcane thing that us geeks do (Wait! Am I a geek?). Reality is that software languages are well defined. Compilers and interpreters are very precise about syntax. A semicolon in the wrong place means complete failure. In some bizarre languages (Yes, I’m talking about you Python!) even an incorrect space can change the logic of a program.
Guess what? AI loves precise syntax rules. With those rules it knows what to do. Unlike natural language, which is constantly mutating under the influence of human thought, where ambiguity is the rule of the day. AI struggles badly at managing to work with natural language. It’s taken an incredible amount of human intelligence to produce software (AI) capable of wielding natural language reasonably well.
The key there is “reasonably well”. Try a bit of sarcasm the next time you are having a conversation with Grok or ChatGPT. “Oi mate, could you be a little less clear?”
Another fascinating aspect is that courses have been created to teach humans how to write to AI. It’s called “Prompt Engineering”. Wait? Huh? Why do I need to learn how to talk to Grok if AI is so intelligent? Why can’t it just understand me?
Be careful, if you don’t express yourself correctly AI won’t understand you!
Unfortunately we haven’t reached Utopia yet!
Part One: https://mithel.substack.com/p/utopia-part-1
Part Three: https://mithel.substack.com/p/utopia-part-3
> It turns out that a million monkeys randomly typing do actually produce “Shakespeare” sometimes.
Ha!