Video starts at 3:39

    • invalidusernamelol [he/him]@hexbear.netM
      link
      fedilink
      English
      arrow-up
      4
      ·
      edit-2
      1 year ago

      Agree that it’s useful, I use copilot a lot because I write a lot of templated code that interfaces with ArcGIS at work (arcpy) and it’s good at auto filling the arcpy function calls if I get my variables set right and document properly.

      It’s also good at generalizing data and extracting patterns any one person might not see.

      Using it with stuff like LHC data will probably make a lot of physicist jobs easier if it’s trained in a transparent and open way where peer review is involved.

        • invalidusernamelol [he/him]@hexbear.netM
          link
          fedilink
          English
          arrow-up
          3
          ·
          1 year ago

          Still think using “AI” as a term is pointless because there are multiple statistical frameworks that can be used to implement general regressive neural networks and there definitely not artificial intelligence any more than a barcode scanner.

            • invalidusernamelol [he/him]@hexbear.netM
              link
              fedilink
              English
              arrow-up
              2
              ·
              1 year ago

              So much of the discourse around this technology is focused on marketing terms and misses the underlying mathematics and engineering behind these projects.

              None of this is new, in fact the first Gosplan program essentially functioned using the same base math as GRNN except with only one iteration through the matrix and it took a long time to compute by hand. The only thing we have now that’s different is very advanced computational power that’s able to run those calculations in seconds rather than months.