Chatbots are just bloviating idiots so calm down!
Yeah, that’s the ticket! (One for those who remember SNL in the 80s)
I’ve been reading about people panicking about chatbots being the harbingers of AI taking over all the jobs, but I think it’s a bit soon to worry. They’re good at producing text that makes grammatical sense and is internally logical, but still can’t work out if something is true.
I tested OpenAI out by asking it for help in completing a task in South Park: the Stick of Truth (it came out in 2014), and it got it wrong, despite there being a plethora of guides out there. It got some parts sort of right, but just made others up completely.
I said it was wrong and to try again, but it just made up something else vaguely plausible but incorrect. I gave up at this point as it was clearly just going to keep throwing shit until some of stuck. It might eventually have worked out the correct answer, but I wouldn’t know when this was unless I knew it already. There is no intelligence that I can see, just brute forcing a million figurative monkeys to type on a million figurative typewriters.
I think a big issue will be that click bait articles will be able to be more quickly produced as it can be automated, and who cares if it’s true or not? OpenAI was quite sniffy when asked to write a click bait article about Prince Harry, noting that “as an AI language model, I cannot generate clickbait articles or any other type of content that aims to mislead or manipulate readers”, but how long will that last? I tried a few different prompts but didn’t manage to get it to write anything fun about the Prince and Princess of Canada.
Chatbots give the illusion of intelligence, but I’m not impressed at all. Of course, I had to ask it to write an article about how OpenAI isn’t really intelligent as if written by a human and it seemed a bit too full of itself.
While OpenAI is a remarkable achievement in the field of artificial intelligence, it is not yet capable of replicating the complexity and versatility of human intelligence. We still have a long way to go in developing machines that can truly think and reason like humans.