At the start of most workshops I facilitate related to artificial intelligence (AI), I begin by getting a shared definition of AI. I ask the participants for their definitions and then compare them to other definitions I’ve collected (including from ChatGPT itself).
And then we look for themes.
Usually the top theme that emerges is that AI is a machine with human-like capabilities. And we go with that.
But what does “human-like” mean? Why is that a big deal? And how does that mean we should interact with it?
AI was specifically designed to be human-like, and developers are working to make it more and more human-like each day.
We can see that in anecdotal evidence like the following:
It’s more productive in December than it is in April, and the going theory is because it’s internalized the idea of winter break from its training data.
It responds to emotional manipulation, giving you better quality if you offer to bribe it or tell it that your job depends on it.
It generates ideas that impartial, outside judges find to be better (and more human-like) than human, trained students.
It outscores human doctors on scales of empathy and judgment and beats human advice-givers on the effectiveness, novelty, and empathy of their reappraisal.
And, ultimately, it’s not a human, and we need to walk the line between knowing it’s not human…but treating it like it is.
So what does that mean?
Treat it like it’s human
AI does not work like a search engine. You can’t just give it the right search term and expect it to give you a great response. The most effective way to work with AI is to start with a well-developed prompt, engage it in dialogue, and train it. Yes, that means that you should have a conversation with the AI like it’s human.
But not too human
But keep in mind that it’s not actually human and does not follow the same ethical, moral, and behavioral guidelines as people. It might give great advice, but, ultimately, it’s a computer and can’t be held accountable for any action. It might seem to be really empathic, but, ultimately, it’s a computer and doesn’t actually care about it. It might seem to be wise, but, ultimately, it’s a computer and it basing its response on data and algorithms and may not be considering all of the angles. As such, you, as the human, need to keep it in its place – that of not a human.
And treat it like it’s a human who needs a lot of guidance
But that doesn’t mean that it isn’t good for anything. Treat it like a human that needs a little guidance. The way its training works, it’s more interested in pleasing you than being right, so double-check anything it gives you for accuracy. Just because it exceeds human ability at one task doesn’t mean it can do all related work at human level, so make sure not to put the AI above you and other trusted humans.
The best way to work with AI is to recognize and appreciate it for what it is – a human-like technology that needs actual human guidance. Partner with it, experiment with it, delegate to it. Treat it like it’s human.
Comments