Post AW82GgXLDoSHCad9fc by jgordon@appdot.net
 (DIR) More posts by jgordon@appdot.net
 (DIR) Post #AW7yCe2CBSqUSxsdgu by lauren@mastodon.laurenweinstein.org
       2023-05-28T22:27:09Z
       
       0 likes, 0 repeats
       
       ***** My view on Generative AI to date *****So far, I am singularly unimpressed with all of the Generative AI chatbots and search extensions I've experienced to date. The results are usually bland and poorly worded -- much like an "intelligent parrot" perhaps, with content obviously lifted en masse from other websites, likely without their permission in many cases.More worrisome, these generative AI systems' responses tend to be scattered with errors (and sometimes utter fantasies) that are woven in among sometimes accurate information. This is the worst of both worlds, because scattered misinformation binds itself to accurate information in ways almost impossible for the casual observer to even notice. The possibilities for "misinformation on steroids" in this tech are enormous and growing. -L
       
 (DIR) Post #AW7zS0pHpJoTS3jZ3o by jgordon@appdot.net
       2023-05-28T22:41:07Z
       
       0 likes, 0 repeats
       
       @lauren So radically different from my experience. I just asked ChatGPT 4 about unclogging a drain and got a quality list including options I'd forgotten.
       
 (DIR) Post #AW7zYcUug78C1bM5E8 by lauren@mastodon.laurenweinstein.org
       2023-05-28T22:42:25Z
       
       0 likes, 0 repeats
       
       @jgordon And where did it steal that list? Did you visit the site it stole the data from? Or did it even bother to tell you?
       
 (DIR) Post #AW82GgXLDoSHCad9fc by jgordon@appdot.net
       2023-05-28T23:12:38Z
       
       0 likes, 0 repeats
       
       @lauren I don't know where it got material it shared. I did not visit any sites other than ChatGPT. ChatGPT does not indicate sources.
       
 (DIR) Post #AW83ELs9BFykUHoV9M by privacat@dataprotection.social
       2023-05-28T23:23:28Z
       
       0 likes, 0 repeats
       
       @laurenI liken LLMs as a first draft tool, or occasionally a sounding board that also happens to summarize well.You can get some utility out of these tools, but it should always an iterative process, and just like any other nascent assistive tech, you should verify every time.Sometimes, it will get enough of an outline or skeleton together that I can take it, build on that, iterate with the LLM and make it my own. That will saves me quite a bit of time -- getting initial words/thoughts on paper can be a huge blocker.But I have zero expectations that it will do all of the work for me.
       
 (DIR) Post #AW84qoAwUxa9E6VS6q by privacat@dataprotection.social
       2023-05-28T23:25:51Z
       
       0 likes, 0 repeats
       
       @jgordonYou can actually get it to cite sources if you ask. 'please cite your sources and include links' usually works. Granted, sometimes those sources are confabulated, but some are accurate. @lauren
       
 (DIR) Post #AW84qp6N3RWG6DNJce by lauren@mastodon.laurenweinstein.org
       2023-05-28T23:41:39Z
       
       0 likes, 0 repeats
       
       @privacat @jgordon Hardly anyone will bother asking for sources. Basically, the data is being stolen from sites, and they don't even get traffic back in response like a normal SERP. It's that simple.
       
 (DIR) Post #AW8L1oBoNw6tipBE2q by jgordon@appdot.net
       2023-05-29T02:42:56Z
       
       0 likes, 0 repeats
       
       @lauren @privacat I do know that.
       
 (DIR) Post #AW8LPvvDHvjKeYv5s0 by escarpment@mastodon.online
       2023-05-29T02:47:20Z
       
       0 likes, 0 repeats
       
       @lauren I am repeatedly awed and astounded by ChatGPT, but I have never really tried to do something "useful" with it. I'm just amazed when I can get it to spit out a blues song about a certain word game I was just playing or generate a novel episode of a TV show.Perhaps it's a difference of expectations. I'm expecting an uncannily good natural language generator and understander, and that's what I get and I'm impressed. Sounds like some are expecting a fully sentient AI and are disappointed.
       
 (DIR) Post #AW8LTk9D0T9VIJHbRw by jgordon@appdot.net
       2023-05-29T02:48:01Z
       
       0 likes, 0 repeats
       
       @lauren Hi Lauren. I know that of course. It’s possible for two things to be true:1. LLMs are astounding technology that many find very useful (I realize you do not)2. Many believe the use of public web data to train LLMs is a form of IP theft.
       
 (DIR) Post #AW8LkHqvSY9uSicUOe by lauren@mastodon.laurenweinstein.org
       2023-05-29T02:51:01Z
       
       0 likes, 0 repeats
       
       @jgordon https://lauren.vortex.com/2023/02/14/giving-creators-and-websites-control-over-generative-ai
       
 (DIR) Post #AWBVRgEEBkT5ZnAIyG by berndporr@mastodon.social
       2023-05-30T15:23:48Z
       
       0 likes, 0 repeats
       
       @lauren ...and as any YouTube guru grifter it delivers it with huge confidence and eloquence so it must be true!