#metabrainz

/

      • CatQuest
        but apparently kids in norwegian school is also using it?
      • akshaaatt
        I’ve also been planning my trips to cities using ChatGPT
      • CatQuest
        what *IS* it?
      • mayhem
        I'm not very worried about chatgpt replacing programmers soon. will it make them more effective, yes.
      • CatQuest
        some AI that writes what?
      • mayhem
      • try it
      • CatQuest
        hmph, bet it doesnt work for me
      • mayhem
        I'll bet you it does.
      • CatQuest
        nope "your browser is out of date" :D
      • akshaaatt
        I use chatgpt for complex functionalities as well.
      • mayhem
        oh, not work in *that* way. yes, of course it won't work for you.
      • CatQuest
        omg jsut reading that excerpt
      • alastairp
      • akshaaatt
        We could probably ask it about how to integrate the apple music sdk to android and it could help any experienced android dev get it done
      • CatQuest
        are ya'll enurtypicals really falling for this BS
      • alastairp
        CatQuest: it's a natural-language "AI"-backed chat bot
      • CatQuest
        It’s difficult to say without more information <--
      • akshaaatt
        Damn that’s really smooth alastairp
      • CatQuest
        🙄
      • alastairp
        you ask it things and it tells you things
      • and it gets things wrong most of the time
      • CatQuest
        it sounds liek every annoying chat bot i've ever had the missfortune to interact with
      • alastairp
        akshaaatt: careful, because it's also very confident at telling you things that are completely wrong
      • CatQuest
        to much text. just tell me the proble don't wrap it in lots of "blah blah"
      • akshaaatt
        It seems to be correct for me. It actually learns as well idkh. I am able to prompt to the model about what it did wrong in the previous response
      • alastairp
        CatQuest: you can in fact tell the bot to respond to you with less text
      • akshaaatt
        I agree that it answers wrong things confidently at times
      • CatQuest
        see. this is where neurospicy people will turn out to havean evolutional advantage and
      • mayhem
        huh. oh. hah. oh yes. we need this.
      • CatQuest
        i don't really like it
      • alastairp
        e,g,
      • CatQuest
        you do? why?
      • alastairp
      • akshaaatt
        I don’t look at chatgpt as a know it all but an entity which summaries things like a human. There are times when us humans also say the wrong thing.
      • alastairp
        not actually a valid uuid
      • CatQuest
        it just repeats itslf
      • lol
      • i cna get that by searching mb
      • and typing *less*!!!
      • mayhem
        we need a "make a playable playlist from chatgpt results" feature in LB.
      • well, shan't be a lot of work with troi
      • akshaaatt
        CatQuest: google tends to link you the closest answers, where there could be many. But chatgpt gives you one answer at a time, which I tbh really like
      • CatQuest
        akshaaatt: why. do you nead. it to summarise "like a human" ? why need "it's hard to" blah balh "you could change if you" why need this extra fluff?
      • akshaaatt: I don't use google?
      • alastairp
        yep, openai API + some careful prompts + parsing results should be easy peasy
      • akshaaatt: but how do you know that the answer it gives you is correct?
      • mayhem
        I was thinking of just copy & paste for regular users.
      • CatQuest
        i'd liek if it was like "need more input" and "seems to be erro at (this thing)" it cna be solved with "blah"
      • also that.
      • i want to know why it thinks it's right
      • source of decision
      • alastairp
      • for those in the know
      • atj
        heh
      • CatQuest
        i thoguht "slash" was a refrenche to / in urls
      • huh
      • alastairp
        it's so confident with the answer!
      • CatQuest: yes, that's the correct answer
      • CatQuest
        .. wait?
      • oh
      • uh
      • uh
      • akshaaatt
        We don’t alastairp . For example, if you were to explain something to me, even if it were wrong, I would believe you in the initial fly. If I had more knowledge on the subject and I felt you were wrong, we would have had a discussion over it. I see the same happening with my conversations with chatgpt
      • atj
        best description for it i've seen it was "automated reply guy"
      • CatQuest
        20 somethings frighten me
      • :D
      • akshaaatt
        Lmao atj
      • mayhem
      • alastairp
        sweet
      • mayhem listens
      • akshaaatt: that attitude worries me, and I'm trying to articulate why
      • CatQuest
        i think it's because we assume a computer will give us truth but another human can be flawed
      • atj
        as usual, it's a tool that is useful for certain things and not for others
      • CatQuest
        becasue instincitvly we assume computers are not fed bogus
      • alastairp
        I don't think it's that exactly
      • CatQuest
        it already annoys me when an aswer is wrapped in many layers of text (like the initial bit about "what is slashdot" in that reply. like, i didn't ask about what it is, i asked aobut the name of it. it's unecessary information i don't care about)
      • alastairp
        consider something like the original google - when it very clearly used relationships between websites to build this network of trust
      • akshaaatt
        Makes sense
      • alastairp
        compared to internet/search results now where everything is trying to get to the top of results, and people fill in pages and pages of garbage to try and catch search queries to get eyes on a page to show you ads
      • Freso
        alastairp: Let me know when I should try the test chatlogs site again btw. :) (Still getting server error.)
      • alastairp
        I'll admit, I'm more likely to trust something at face value that I read off wikipedia, even though I know that has its own problems
      • so I can see the jump from that to trusting chatgpt
      • akshaaatt
        An example of what you're trying to say is that, if you had to search what a word means, you would prefer to be given the dictionary, which is sorted and look for the word yourself and conclude your finding.
      • However, for lazy people, someone just telling them what a word means is enough
      • CatQuest
        but that's what irc is for!
      • CatQuest kidding
      • akshaaatt
        Although I am hoping GPT 4 will be better at this
      • It has way more parameters under which it is being trained
      • CatQuest
        people need to watch some old sci-fi movies
      • alastairp
        well, I'd use a digital dictionary instead of a paper one. but I suspect I'd trust google or wikictionary more than chatgpt on that
      • CatQuest
        this is why i wanted the grove dictionary books guys
      • Freso
        You’re welcome. :)
      • CatQuest
        :)
      • (I love these books btw)
      • Freso
        Did you add them to BB?
      • akshaaatt
        When we write research papers, we end it with a References link where we link to all the webpages or articles that we referred to for our findings
      • What if chatgpt linked you to the reference alastairp>?
      • CatQuest
        not yet because I haven't figured if each tiny article (wich has credits of friter) shoudl be its own work (probably) but thats.. probably thousands of works /.___.\
      • now *THA*T i would want, i would consider linking the source mandatory in ALL AI things
      • Freso
        You could always add the edition(s) and worry about works later. I’ve done that with a couple of books.
      • CatQuest
        for mmsuic to art to chat ot whatever
      • alastairp
        Freso: thanks, but I've got to get some other things done today so may not be able to fix it. But I can check it myself later
      • CatQuest
        Freso: yea. probably a good idea. i shoudl do that
      • Freso
        alastairp: 👍
      • atj
        now this is a great use of openai: https://pbs.twimg.com/media/Fi2yZq1XEAEs9cp?for...
      • CatQuest
        (thoguh i have to figure out how to add somethig that is.. 3 editions but yhe same "thing")
      • like a cd boxset
      • sigh
      • perhaps the same EG?
      • alastairp
        akshaaatt: I mean, there's a good point. why doesn't it do this already? Wikipedia does it, Google does it (even if the material it links to might be suspicious)
      • atj: comments are in the wrong tone
      • akshaaatt
        Right. I think it has to do with what CatQuest said
      • CatQuest
        atj: now *THA*T is the sort of thing I'd like the thing for
      • atj
        alastairp: perhaps if they had specified including comments
      • mayhem
        alastairp: roulette sandwiches?
      • CatQuest
        russian roulette sandwiches
      • alastairp
        mayhem: I have food with me, thanks
      • mayhem
        k
      • CatQuest
        (sorry not sorry)
      • atj
        final point on openai, this is an example of how it can really help people: https://twitter.com/DannyRichman/status/1598254...
      • alastairp
        yeah, for sure
      • I saw another one showing a possible future where someone uses it to generate a nice email, and then someone else uses it to summarise the nice email into key points
      • thanks chatgpt
      • akshaaatt
        A friend of mine backed an internship by making their cover letter and sending an essay using chatgpt
      • You could also look into using chatgpt to help with your accounting work I suppose mayhem
      • alastairp
        that sounds like the worst idea I've heard yet, sorry
      • akshaaatt
        XD
      • alastairp
        you really don't want to be submitting legal documents generated by a computer with no insight into how it got to the answer
      • atj
        accounting is an area where one mistake could put you in jail, so probably not
      • if you ever have the misfortune to have to do accounting you'll understand :)
      • akshaaatt
        I didnt mean the legal documents part of it alastairp. I am not quite familiar with the work underneath sorry
      • CatQuest
        i'd be ok with all the "generated by computer" parts of all these AI's *IF* I had that data. ie, if the AI "showed it's work" thne it'd be great
      • atj
        CatQuest: the way these generative AIs work means it's impossible to "show it's work"
      • CatQuest
        that's really my only issue with it. that the source and origins and such is locked behind this text/image/music/whatever
      • pbryan
        That's the central problem with AI output, it's virtually impossible to reason about its behavior.
      • CatQuest
        atj: i say that's bogus
      • it's jsut hard
      • atj
        CatQuest: no, if you understood how they work you'd realise
      • CatQuest
        i can't us that argument whne writing a paper or math problem solution
      • pbryan
        Why did the Tesla swerve into the crash barrier?
      • atj
        it's like trying to remove the egg from a cake
      • pbryan
        I wish I had a good punchline to that.
      • CatQuest
        i don't buy it, sorry
      • atj
        well i'm not selling
      • CatQuest
        AI needs to be written to *implement* the ability to serve the source.
      • no i don't know the code behind it or how it "works"
      • but i'm.. clever enough ot imagine the gist of it
      • akshaaatt
        Tbh you add a voiceover to the chatgpt language model and you have another one of Siri/Google Assistant