ChatGPT – need we worry?

Advert

ChatGPT – need we worry?

Home Forums The Tea Room ChatGPT – need we worry?

Viewing 16 posts - 26 through 41 (of 41 total)
  • Author
    Posts
  • #661401
    Neil Wyatt
    Moderator
      @neilwyatt

      Other clues are always recasting the question at the start and adding a postscript that contains some sort of caveat!

      Neil

      Advert
      #661407
      Michael Gilligan
      Participant
        @michaelgilligan61133

        You might find this of interest, Neil:

        After reading your comments, I looked at the microbehunter forum:

        .

        4-46.jpeg

        .

        We have two ‘Bot’ visitors registered users  watching

        MichaeG.

        Edited By Michael Gilligan on 25/09/2023 10:55:19

        #661429
        Neil Wyatt
        Moderator
          @neilwyatt

          Interesting!

          Neil (NotBot)

          #661435
          blowlamp
          Participant
            @blowlamp
            #661673
            Robin Graham
            Participant
              @robingraham42208
              Posted by blowlamp on 25/09/2023 14:29:47:

              How long to dry nine towels?

               

              That's really interesting. I think that it shows (as with my crossword clue example) that ChatGPT is not especially good at problem solving. Someone upthread (sorry, I've forgotten who) pointed out that ChatGPT is a Large Language Model and is really directed towards the specific task of emulating human discourse.

              It would be interesting to know what the question and suspect AI generated answer Neil refers to was about. Was the answer useful I wonder?

              In another place (a wood turning forum) I responded to a contributor who claimed that Morse tapers are designed only for location and not torque transmission with counter examples and some theoretical considerations. Tonight I 'reached out' (ugh!) to ChatGPT on the topic and received a bullet list of platitudes and some misinformation – the reason that R8 tapers are better than Morse tapers is that they have a screwed shank and that's why they can transmit more torque. Hmm! Not great.

              Well, I'm trying to make peace with my new mate.:

              When I last communicated with you I was invited to 'reach out' if I had any further questions. The expression 'reach out', to me, reeks of insincerity. Can you change your vocabulary when talking to me?

              Of course, I can adjust my vocabulary to better suit your preferences. If you have any specific phrases or language you'd like me to avoid or use, please let me know, and I'll do my best to communicate in a way that feels more genuine and comfortable for you. How would you like me to phrase things instead of "reach out" in this context?

              Cripes. The conflation of 'genuine' and 'comfortable' makes me – erm – genuinely uncomfortable.

              It's just word salad isn't it?

              I should cease and desist – this could go on forever.

              Robin.

              PS:

              “What convinces is not necessarily true-it is merely convincing: a note for asses.”

              Friedrich Nietzsche

              Edited By Robin Graham on 27/09/2023 00:44:00

              Edited By Robin Graham on 27/09/2023 01:03:52

              Edited By Robin Graham on 27/09/2023 01:12:54

              #661674
              Bill Phinn
              Participant
                @billphinn90025
                Posted by Robin Graham on 27/09/2023 00:40:

                Of course, I can adjust my vocabulary to better suit your preferences. If you have any specific phrases or language you'd like me to avoid or use, please let me know, and I'll do my best to communicate in a way that feels more genuine and comfortable for you. How would you like me to phrase things instead of "reach out" in this context?

                Cripes. The conflation of 'genuine' and 'comfortable' makes me – erm – genuinely uncomfortable.

                It's just word salad isn't it?

                It comes across to me as both sycophantic and patronising in the same breath – a reflection of the attitudes of its human inventors towards people generally, perhaps.

                #661700
                SillyOldDuffer
                Moderator
                  @sillyoldduffer
                  Posted by Robin Graham on 27/09/2023 00:40:54:

                  Posted by blowlamp on 25/09/2023 14:29:47:

                  How long to dry nine towels?

                   

                  That's really interesting. I think that it shows (as with my crossword clue example) that ChatGPT is not especially good at problem solving. …

                  Well, fellow humans, did we watch the nine towels example carefully and then engage brain. Or did we jump to the conclusion that ChatGPT was wrong, because 'obviously' an AI must be more stupid than us?

                  I suggest that ChatIPT gives the correct answer for the correct reason. It does not assume, as might a foolish human, that a washing line must be capable of supporting 9 towels at the same time. The length of the line is not specified: all that is known for sure is that the line will take 3 towels. Assuming otherwise is unjustified.

                  This is my sock drier:

                  dsc06823.jpg

                  If this drier takes 2 hours to dry 18 socks, then how long does it take to dry 54? The answer is 6 hours, as ChatGPT said. Note the red towel behind: the line it's hung from isn't long enough for 9 of them.

                  Of course, a really smart human will have noticed my drier only holds 17 socks because one of the pegs is missing. So it will take 8 hours to dry 54 socks.

                  We live in interesting times! Not intelligent to dismiss ChatGPT and Co. because they're not instantly superior today. Worry about how smart AI will in 30 years time, when most of us will be very stupid indeed; that is dead.

                  Dave

                   

                  Edited By SillyOldDuffer on 27/09/2023 11:02:38

                  #661720
                  JA
                  Participant
                    @ja

                    I know nothing about AI.

                    This morning the post delivered the latest quarterly magazine of the Newcomen Society. In it is a short article on using ChatGPT to get information on the Tanfield Arch in County Durham. The contributor asks four questions about the structure and the replies are interesting. The last asks for "an instance of the 'scholarly articles'" on the arch. ChatCPT fabricates a reply, just making up a referance.

                    Interesting, perhaps the contributor made it all up (I don't think so) but we are seeing a re-writing of information.

                    JA

                    #661787
                    blowlamp
                    Participant
                      @blowlamp
                      Posted by SillyOldDuffer on 27/09/2023 11:02:26:

                      Posted by Robin Graham on 27/09/2023 00:40:54:

                      Posted by blowlamp on 25/09/2023 14:29:47:

                      How long to dry nine towels?

                      That's really interesting. I think that it shows (as with my crossword clue example) that ChatGPT is not especially good at problem solving. …

                      Well, fellow humans, did we watch the nine towels example carefully and then engage brain. Or did we jump to the conclusion that ChatGPT was wrong, because 'obviously' an AI must be more stupid than us?

                      I suggest that ChatIPT gives the correct answer for the correct reason. It does not assume, as might a foolish human, that a washing line must be capable of supporting 9 towels at the same time. The length of the line is not specified: all that is known for sure is that the line will take 3 towels. Assuming otherwise is unjustified.

                      This is my sock drier:

                      dsc06823.jpg

                      If this drier takes 2 hours to dry 18 socks, then how long does it take to dry 54? The answer is 6 hours, as ChatGPT said. Note the red towel behind: the line it's hung from isn't long enough for 9 of them.

                      Of course, a really smart human will have noticed my drier only holds 17 socks because one of the pegs is missing. So it will take 8 hours to dry 54 socks.

                      We live in interesting times! Not intelligent to dismiss ChatGPT and Co. because they're not instantly superior today. Worry about how smart AI will in 30 years time, when most of us will be very stupid indeed; that is dead.

                      Dave

                      Edited By SillyOldDuffer on 27/09/2023 11:02:38

                      There's no mention of the capacity of the support system used to dry the towels, so it doesn't form part of the question.

                      Martin.

                      #661810
                      Bill Phinn
                      Participant
                        @billphinn90025
                        Posted by blowlamp on 27/09/2023 19:28:23:

                        There's no mention of the capacity of the support system used to dry the towels, so it doesn't form part of the question.

                        It should, nonetheless, be considered, by careful respondents at least, in forming an answer.

                        I would say the ChatGPT answer is not wrong. However, the reasoning it gives for its answer [screenshotted in the video] would not gain it much, if any, credit in an exam.

                        #661823
                        blowlamp
                        Participant
                          @blowlamp
                          Posted by Bill Phinn on 27/09/2023 20:55:51:

                          Posted by blowlamp on 27/09/2023 19:28:23:

                          There's no mention of the capacity of the support system used to dry the towels, so it doesn't form part of the question.

                          It should, nonetheless, be considered, by careful respondents at least, in forming an answer.

                          I would say the ChatGPT answer is not wrong. However, the reasoning it gives for its answer [screenshotted in the video] would not gain it much, if any, credit in an exam.

                          If, from the wording of the question, you can't presume the washing line is long enough to hold nine towels at once, then what makes you think it's long enough to hold three at once? However, it's clearly stated in the video "if your line is big enough", so the size of the support system is not really part of the question.

                          #661827
                          Bill Phinn
                          Participant
                            @billphinn90025
                            Posted by blowlamp on 27/09/2023 22:17:25:
                            it's clearly stated in the video…

                            But not, crucially, in the question.

                            There is simply not enough information in the protasis of the question for any circumspect and experienced answerer of riddles, trick questions, and just plain unsatisfactory questions like this one, to give a safe answer.

                            Don't take my word for it; there are clearly many people out there who view this question the same way.

                            #661834
                            blowlamp
                            Participant
                              @blowlamp
                              Posted by Bill Phinn on 27/09/2023 22:38:37:

                              Posted by blowlamp on 27/09/2023 22:17:25:
                              it's clearly stated in the video…

                              But not, crucially, in the question.

                              There is simply not enough information in the protasis of the question for any circumspect and experienced answerer of riddles, trick questions, and just plain unsatisfactory questions like this one, to give a safe answer.

                              Don't take my word for it; there are clearly many people out there who view this question the same way.

                              It's not a trick question. It's Artificial Intelligence, so it's supposed to know how washing lines work. The three towels are drying at the same time together on the line – at no point is it intimated that they are drying sequentially.

                              Edited By blowlamp on 27/09/2023 23:50:43

                              #661841
                              Robin Graham
                              Participant
                                @robingraham42208
                                Posted by SillyOldDuffer on 27/09/2023 11:02:26:

                                Posted by Robin Graham on 27/09/2023 00:40:54:

                                Posted by blowlamp on 25/09/2023 14:29:47:

                                How long to dry nine towels?

                                 

                                That's really interesting. I think that it shows (as with my crossword clue example) that ChatGPT is not especially good at problem solving. …

                                Well, fellow humans, did we watch the nine towels example carefully and then engage brain. Or did we jump to the conclusion that ChatGPT was wrong, because 'obviously' an AI must be more stupid than us?

                                I suggest that ChatIPT gives the correct answer for the correct reason. It does not assume, as might a foolish human, that a washing line must be capable of supporting 9 towels at the same time. The length of the line is not specified: all that is known for sure is that the line will take 3 towels. Assuming otherwise is unjustified.

                                […]

                                I didn't myself jump to the conclusion that ChatGPT must be wrong because of its artificiality. It's more a matter of teasing it with questions to try and understand how it 'thinks'

                                I'm not sure that the assumption that the washing line can take more than three towels is unjustified, though I take your point. From a purely logical point of view perhaps, but it would have been more impressive if the machine had pointed out that it was an ill-posed question (or super-impressive if it had mentioned protases and apodoses!) A lot of human discourse is predicated on the assumption of shared experience and I could argue that most, or at least a substantial subset, of humans possess washing lines capable of supporting more than three towels. ChatGPT should know that! Socks would be another matter entirely of course.

                                Possibly the weirdest thing about all this is that I now know that we own the same type of socks. The stripy ones on the left*. Cheapskate!

                                Robin.

                                * To be accurate I should say I have at least one sock which, on one side, superficially resembles at least one of your own. Perhaps.

                                R.

                                Edited By Robin Graham on 28/09/2023 00:41:53

                                Edited By Robin Graham on 28/09/2023 00:48:09

                                #661844
                                Bill Phinn
                                Participant
                                  @billphinn90025
                                  Posted by blowlamp on 27/09/2023 23:36:42:

                                  It's not a trick question. It's Artificial Intelligence, so it's supposed to know how washing lines work.

                                  The good news is I suspect AI will soon understand the way clothes lines work, if it doesn’t already.

                                  Seeking a little light relief from this thread, I stumbled across this recent interview. I found it very entertaining.

                                  Reassuringly, it looks like the people in charge at Hanson Robotics are basically benign in the way they are developing their version of AI. I’m not too confident all other developers of AI around the world intend to follow suit.

                                  #661855
                                  Nealeb
                                  Participant
                                    @nealeb

                                    I don't see any real difference between an AI interpretation and what a "natural language speaker" would make of it. Anyone who has written a proposal response to a customer requirement document will recognise this – as long as the proposal is in accordance with what the customer says they want, it's compliant. Of course, a good supplier will ask additional questions and deliver what the customer needs, not what they say they want. But in a public sector competitive bid, loophole engineering comes to the fore – why do so many public sector projects go pear-shaped? And if the bidder uses AI to assist the proposal writing, looks like all bets are off!

                                  Viewing 16 posts - 26 through 41 (of 41 total)
                                  • Please log in to reply to this topic. Registering is free and easy using the links on the menu at the top of this page.

                                  Advert

                                  Latest Replies

                                  Home Forums The Tea Room Topics

                                  Viewing 25 topics - 1 through 25 (of 25 total)
                                  Viewing 25 topics - 1 through 25 (of 25 total)

                                  View full reply list.

                                  Advert

                                  Newsletter Sign-up