I have also played around with simpler queries (see below) and find the same: excellent, easy to read answers, with minor mistakes and sometimes false, invented info.
But I still see this as the future. If/when this is implemented with speech recognition (already exists) and good synthetic speech, you have a new interface for users that will take on the world. For me the only question is of Google come first (who has better knowledge how to interface backend knowledge repositories the Wikipedia will become) or if chatGPT will learn this, that seems for me to be a minor thing for them compared with getting the answering logic in place that they have mastered.
And I still believe is is vital for us to understand how this future will affect or systems as well as how we will need to work in order to still provide knowledge, even if it is to a repository and not the world directly.
My test "Where lies Farsta?"
"Farsta is a suburb located in the southeastern part of Stockholm, the capital city of Sweden. It is situated in the Söderort borough of Stockholm and is known for its shopping center and recreational areas. Farsta is easily accessible by public transport, with several bus and metro lines passing through the area. The suburb is also home to Farsta Strand, a beach and recreational area that attracts visitors during the summer months."
*that söderort actually is not a borough is not of vital
importance to the reader.
*the last sentence and to call it recreational areas is just
wrong. Farsta strand is called a satellite town i enwp and is
primary a built up area located close by a smaller lake. There
exist a small beach "farsta strandbad" for local residents, but
there are 5-6 better beaches at cleaner lakes within a 5 km
radius. And not a recreation area and certainly not attracting
visitors during the summer months from outside this place (there
are many excellent recreational areas very close by)
Anders
On Fri, Dec 23, 2022 at 1:04 PM Paulo Santos Perneta <paulosperneta@gmail.com> wrote:
I've been playing with it on over the last week, and my experience is that that AI is incredibly prone to invent informations that look credible, while being entirely made up by it. It would generate inexistent persons, dates, events and places, and build an entire story out of them that, nevertheless, looks credible.
Indeed; here is another example, after prompting to write an article about the song "Tomorrow Belongs to Me", which was written in 1966 for the Broadway musical "Cabaret". ChatGPT attributes it, entirely fictionally, to a real 1920s German lyricist.
"Tomorrow Belongs to Me" is a song that has been used in a variety of media, including film, television, and theater. The song is known for its nationalist and anti-Semitic lyrics, which have made it controversial and a subject of debate.
The song was originally written by the German composer and lyricist Hans Leip in the 1920s. It was later adopted by the National Socialist German Workers' Party (Nazi Party) as a propaganda song, and was performed at party rallies and events.
In the United States, "Tomorrow Belongs to Me" was popularized by its use in the film "Cabaret," which was released in 1972. In the film, the song is sung by a young man at a Nazi rally, and is later performed by the character of Sally Bowles, played by Liza Minnelli. The song has since been used in a number of other films, television shows, and stage productions, often as a means of referencing the Nazi party or the Holocaust.
Despite its widespread use, "Tomorrow Belongs to Me" has been criticized for its nationalist and anti-Semitic themes, and some have called for it to be banned or censored. The song has also been the subject of legal disputes, with some arguing that its use in certain contexts may infringe on copyright laws or be considered hate speech.
_______________________________________________ Wikimedia-l mailing list -- wikimedia-l@lists.wikimedia.org, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l Public archives at https://lists.wikimedia.org/hyperkitty/list/wikimedia-l@lists.wikimedia.org/message/DB56GXXYANLKW3SDGKVBID7EHGXC7JHD/ To unsubscribe send an email to wikimedia-l-leave@lists.wikimedia.org