I would feel hesitant about this, though I don't really have a serious interest in the tool beyond silly experiments I do on the bot.
I've experienced probably a dozen times in the past 30 days where I was trying to chat with the thing during "peak time" and ended up basically just fucking off to something else.
The sour taste that this has brought on drastically reduces my desire to give this thing my money. $29bn valuation, and you can't just serve me the thing on a webpage reliably? And now you want a subscription??
I don't get it. Especially when half the shit I see people using it for could be done correctly by using wolframalpha in the first place.
To get even more tangential: I think we are opening a wildly dangerous can of worms by interfacing with an agent whose responses aren't verified by default. Correctness is extremely important, and ChatGPT doesn't seem to be capable of being locked into any kind of "verified correct answers only" mode. If the average person I speak with in DC is anything to go by, than the people even dumber than that are NOT going to be capable of filtering fact from fiction when using this service, despite how well-known it is to "hallucinate" and literally just talk out of its own ass.
But yeah. Get your javascript snippets faster.... /shrug
I've experienced probably a dozen times in the past 30 days where I was trying to chat with the thing during "peak time" and ended up basically just fucking off to something else.
The sour taste that this has brought on drastically reduces my desire to give this thing my money. $29bn valuation, and you can't just serve me the thing on a webpage reliably? And now you want a subscription??
I don't get it. Especially when half the shit I see people using it for could be done correctly by using wolframalpha in the first place.
To get even more tangential: I think we are opening a wildly dangerous can of worms by interfacing with an agent whose responses aren't verified by default. Correctness is extremely important, and ChatGPT doesn't seem to be capable of being locked into any kind of "verified correct answers only" mode. If the average person I speak with in DC is anything to go by, than the people even dumber than that are NOT going to be capable of filtering fact from fiction when using this service, despite how well-known it is to "hallucinate" and literally just talk out of its own ass.
But yeah. Get your javascript snippets faster.... /shrug