peek

I’ve been playing around with ChatGPT for a while, now. While I understand the uproar and buzz, and it’s groundbreaking technology, for sure, I remain far more concerned about the lack of actual intelligence of many of its users.

While I have been awestruck by it’s capabilities, I have in equal measure been underwhelmed at its ability to cope with *conversational* exchanges that are not purely technical or lecture in nature. Even on the technical side, I’d exercise great caution in relying on these tools in their current form to learn anything new, beyond using it to illuminate the kinds of ideas or related things you need to know. It’ll absolutely get you pointed in the right direction.

It can absolutely code, but you’ll have to debug that code. It can be useful for that too, of course … but it’ll absolutely spit out code that doesn’t work, and dutifully apologize when you point that out.

Does the “conversational” AI not understand the concept of “common” in terms of shared resources? A common cathode LED element differs from a common anode LED element in a key (admittedly somewhat technical) way. Folks may not know what those words mean, and ChatGPT can help with that, but getting it to “remember” that one is absolutely not the other seems to be an exercise in frustration.

It may be that I need to allow longer conversations to be tracked, to ensure that the “corrections” and clarifications I provide actually ‘stick’, but these are basic facts that work the same for everyone and are clearly defined.

It’s impressive, and it’s useful, and it’s going to allow a lot of people to do a lot they couldn’t before. That’s going to be as impressive and awesome and terrifying as such advances have always been. With great power comes great responsibility, and whether to serve good or evil is a choice each individual makes each moment after another.