Ibrohimjon's blog

The most dangerous thing about using AI for idea generation or writing is the fact that it can be very limiting. It gives you a list of five options or writes you an analysis or a text, and when you read it, you convince yourself that if you had written on the topic yourself, you would say the exact same thing. But you did not, so you don't know if you would have had the same ideas.

When you don't develop an idea in your head well enough or you don't have a strong enough understanding of whatever it is you are thinking about, and you consult AI before you allow the thought to finish forming, you risk being shoved into a box. Whatever the AI told you will be what you will see. It's like visual illusions, once you are told what is hiding in the image, you cannot unsee it anymore, to the point that sometimes you can't even see the first image you saw when you first set your eyes on it.

I fear that is what AI is doing to most of us. When we have an academic paper to write and we give ChatGPT a prompt and it gives us the essay, we read and internalize it, and somehow in the back of our brain we convince ourselves that it was our idea, those are our words, our thoughts. But they are not. They could have been if we had written that paper ourselves, but now we will never know because our mind has already been biased. We have already swallowed the content AI wrote and assimilated it into our thinking system.

So we ultimately get stuck in this infinity loop of preexisting ideas, regurgitating facts that are not our own and that we didn't even write. We lose our power of individual, creative, nuanced thinking. We are moving towards homogeneity, and not in a good way.

Perhaps the most effective approach is to use AI more as a responsive collaborator than as a first-draft generator, consulting it after we've developed our own initial thoughts, using it to challenge our assumptions rather than to form them.