3 ms·
Hmm. Well, I find chatGPT similarly useful to Google even in this early stage, albeit for very different types of question. I'd say it's more likely to disrupt
by ben_w 4y ago
Hmm. Well, I find chatGPT similarly useful to Google even in this early stage, albeit for very different types of question. I'd say it's more likely to disrupt WolframAlpha than Google, if my usage is typical — while chatGPT does make basic errors 5% of the time, WolframAlpha just straight up doesn't understand my query 33% of the time.
ChatGPT and other LLMs will probably also disrupt recipe blog spam, with potentially hilarious results for all foods and drinks whose names are double entendres.
LLMs will also, if I correctly understand the difference between opinion columns and journalism, the writing of opinion columns in newspapers. And turning dry research into colourful column inches.
- armchairhacker 4y agoWith what I search (API info), Google makes errors almost 0% of the time. I don't actually know how many errors ChatGPT makes, however a silent error could lead to hours of debugging, so I'd rather just use Google. I do know that when I have complicated questions where Google doesn't really help, ChatGPT also seems to not help: it either produces something false, or a non-answer
- cloudwalking 4y agoI had ChatGPT hallucinate an entire iOS API, complete with docs, sdk version, and URL on apple's website. All of it fake news.
- ben_w 4y agoFirst time I've seen "fake news" refer to not-news. (If it's faking an OS, perhaps "fake GNUs"?)
- ben_w 4y agoFor me, when searching for API info, Google does sometimes fail by giving me links to discussions about obsolete versions. I've not quantified that as a percentage.
- spike021 4y agoSaid the same thing the other day in the post about ChatGPT being used for cheating. I'd say it's a more advanced type of WolframAlpha engine, which I used a lot in my college days a decade ago.