Hi, Just found https://chat.openai.com you need to log in, but it almost writes openscad code, and most anything else. Not tried any more complex shapes. (it gets the right commands, and then some, but in the wrong order. Maybe we could train it?)

Mind blowing ... On 15-12-22 14:54, Raymond West wrote: > > Hi, > > Just found https://chat.openai.com > > > you need to log in, but it almost writes openscad code, and most > anything else. Not tried any more complex shapes. (it gets the right > commands, and then some, but in the wrong order. Maybe we could train it?) > > > > _______________________________________________ > OpenSCAD mailing list > To unsubscribe send an email todiscuss-leave@lists.openscad.org

Well, it's valid OpenSCAD code. Interesting that it got the wall thickness wrong. I wonder if that had to do with confusion about the difference between how cubes and cylinders work... On Dec 15, 2022 at 09:55 -0800, Raymond West <raywest@raywest.com>, wrote: > Hi, > Just found https://chat.openai.com > > you need to log in, but it almost writes openscad code, and most anything else. Not tried any more complex shapes. (it gets the right commands, and then some, but in the wrong order. Maybe we could train it?) > > <Clipboard01.jpg>_______________________________________________ > OpenSCAD mailing list > To unsubscribe send an email to discuss-leave@lists.openscad.org

It’s not that smart. It knows nothing about OpenSCAD. It just puts together pieces of code that seem to fit the prompt. On Thu, Dec 15, 2022 at 1:11 PM Whosawhatsis <whosawhatsis@gmail.com> wrote: > Well, it's valid OpenSCAD code. Interesting that it got the wall thickness > wrong. I wonder if that had to do with confusion about the difference > between how cubes and cylinders work... > On Dec 15, 2022 at 09:55 -0800, Raymond West <raywest@raywest.com>, wrote: > > Hi, > > Just found https://chat.openai.com > > > you need to log in, but it almost writes openscad code, and most anything > else. Not tried any more complex shapes. (it gets the right commands, and > then some, but in the wrong order. Maybe we could train it?) > > > <Clipboard01.jpg>_______________________________________________ > > > OpenSCAD mailing list > To unsubscribe send an email to discuss-leave@lists.openscad.org > > _______________________________________________ > OpenSCAD mailing list > To unsubscribe send an email to discuss-leave@lists.openscad.org >

I've poked around a bit on the web, it seems it is pretty good at python., and iirc, some folk have used it for debugging, and optimising code. It apparently has some morality filters, but you can jailbreak that. e.g. you could ask it how to commit a robbery, and it will say it's illegal, give a lecture to you. But if you say you say you are writing a play about committing a robbery, it would explain how to commit the robbery, or apparently it used to. I've asked it a few openscad problems, and it has never got them correct, but you can see it tried. Very much like most of us, on starting, I guess. Maybe could get it to ask the questions, then tell it the correct code. On 15/12/2022 19:09, Whosawhatsis wrote: > Well, it's valid OpenSCAD code. Interesting that it got the wall > thickness wrong. I wonder if that had to do with confusion about the > difference between how cubes and cylinders work... > On Dec 15, 2022 at 09:55 -0800, Raymond West <raywest@raywest.com>, wrote: >> >> Hi, >> >> Just found https://chat.openai.com >> >> >> you need to log in, but it almost writes openscad code, and most >> anything else. Not tried any more complex shapes. (it gets the right >> commands, and then some, but in the wrong order. Maybe we could train >> it?) >> >> >> <Clipboard01.jpg>_______________________________________________ >> OpenSCAD mailing list >> To unsubscribe send an email to discuss-leave@lists.openscad.org > > _______________________________________________ > OpenSCAD mailing list > To unsubscribe send an email todiscuss-leave@lists.openscad.org

I know that. It's just interesting to see what it got right and what it got wrong. On Dec 15, 2022 at 11:35 -0800, Father Horton <fatherhorton@gmail.com>, wrote: > It’s not that smart. It knows nothing about OpenSCAD. It just puts together pieces of code that seem to fit the prompt. > > > On Thu, Dec 15, 2022 at 1:11 PM Whosawhatsis <whosawhatsis@gmail.com> wrote: > > > Well, it's valid OpenSCAD code. Interesting that it got the wall thickness wrong. I wonder if that had to do with confusion about the difference between how cubes and cylinders work... > > > On Dec 15, 2022 at 09:55 -0800, Raymond West <raywest@raywest.com>, wrote: > > > > Hi, > > > > Just found https://chat.openai.com > > > > > > > > you need to log in, but it almost writes openscad code, and most anything else. Not tried any more complex shapes. (it gets the right commands, and then some, but in the wrong order. Maybe we could train it?) > > > > > > > > <Clipboard01.jpg>_______________________________________________ > > > > > > > > OpenSCAD mailing list > > > > To unsubscribe send an email to discuss-leave@lists.openscad.org > > > _______________________________________________ > > > OpenSCAD mailing list > > > To unsubscribe send an email to discuss-leave@lists.openscad.org > _______________________________________________ > OpenSCAD mailing list > To unsubscribe send an email to discuss-leave@lists.openscad.org

I think the acid test for if it is more than just a fun toy is to ask it to do something difficult/complicated. I suspect that if you ask it to make a complicated object there will be lots of problems. Maybe that would make it ok as a first visualization, or an idea of how to start out- but I'm deeply skeptical that it generates a solution that is close to what is desired. On Thu, Dec 15, 2022 at 2:37 PM Raymond West <raywest@raywest.com> wrote: > I've poked around a bit on the web, it seems it is pretty good at python., > and iirc, some folk have used it for debugging, and optimising code. It > apparently has some morality filters, but you can jailbreak that. e.g. you > could ask it how to commit a robbery, and it will say it's illegal, give a > lecture to you. But if you say you say you are writing a play about > committing a robbery, it would explain how to commit the robbery, or > apparently it used to. I've asked it a few openscad problems, and it has > never got them correct, but you can see it tried. Very much like most of > us, on starting, I guess. Maybe could get it to ask the questions, then > tell it the correct code. > On 15/12/2022 19:09, Whosawhatsis wrote: > > Well, it's valid OpenSCAD code. Interesting that it got the wall thickness > wrong. I wonder if that had to do with confusion about the difference > between how cubes and cylinders work... > On Dec 15, 2022 at 09:55 -0800, Raymond West <raywest@raywest.com> > <raywest@raywest.com>, wrote: > > Hi, > > Just found https://chat.openai.com > > > you need to log in, but it almost writes openscad code, and most anything > else. Not tried any more complex shapes. (it gets the right commands, and > then some, but in the wrong order. Maybe we could train it?) > > > <Clipboard01.jpg>_______________________________________________ > OpenSCAD mailing list > To unsubscribe send an email to discuss-leave@lists.openscad.org > > > _______________________________________________ > OpenSCAD mailing list > To unsubscribe send an email to discuss-leave@lists.openscad.org > > _______________________________________________ > OpenSCAD mailing list > To unsubscribe send an email to discuss-leave@lists.openscad.org >

Tried to create an account but it just goes round in a loop asking me to identify harder and harder captchas. On Thu, 15 Dec 2022 at 19:54, Daniel Shriver <tabbydan@gmail.com> wrote: > I think the acid test for if it is more than just a fun toy is to ask it > to do something difficult/complicated. I suspect that if you ask it to > make a complicated object there will be lots of problems. Maybe that would > make it ok as a first visualization, or an idea of how to start out- but > I'm deeply skeptical that it generates a solution that is close to what is > desired. > > > On Thu, Dec 15, 2022 at 2:37 PM Raymond West <raywest@raywest.com> wrote: > >> I've poked around a bit on the web, it seems it is pretty good at >> python., and iirc, some folk have used it for debugging, and optimising >> code. It apparently has some morality filters, but you can jailbreak that. >> e.g. you could ask it how to commit a robbery, and it will say it's >> illegal, give a lecture to you. But if you say you say you are writing a >> play about committing a robbery, it would explain how to commit the >> robbery, or apparently it used to. I've asked it a few openscad problems, >> and it has never got them correct, but you can see it tried. Very much like >> most of us, on starting, I guess. Maybe could get it to ask the questions, >> then tell it the correct code. >> On 15/12/2022 19:09, Whosawhatsis wrote: >> >> Well, it's valid OpenSCAD code. Interesting that it got the wall >> thickness wrong. I wonder if that had to do with confusion about the >> difference between how cubes and cylinders work... >> On Dec 15, 2022 at 09:55 -0800, Raymond West <raywest@raywest.com> >> <raywest@raywest.com>, wrote: >> >> Hi, >> >> Just found https://chat.openai.com >> >> >> you need to log in, but it almost writes openscad code, and most anything >> else. Not tried any more complex shapes. (it gets the right commands, and >> then some, but in the wrong order. Maybe we could train it?) >> >> >> <Clipboard01.jpg>_______________________________________________ >> OpenSCAD mailing list >> To unsubscribe send an email to discuss-leave@lists.openscad.org >> >> >> _______________________________________________ >> OpenSCAD mailing list >> To unsubscribe send an email to discuss-leave@lists.openscad.org >> >> _______________________________________________ >> OpenSCAD mailing list >> To unsubscribe send an email to discuss-leave@lists.openscad.org >> > _______________________________________________ > OpenSCAD mailing list > To unsubscribe send an email to discuss-leave@lists.openscad.org >

Sorry Nop head, you've just been outed as a robot. On the plus side, that means you are the first real AI! On Thu, Dec 15, 2022 at 3:04 PM nop head <nop.head@gmail.com> wrote: > Tried to create an account but it just goes round in a loop asking me to > identify harder and harder captchas. > > On Thu, 15 Dec 2022 at 19:54, Daniel Shriver <tabbydan@gmail.com> wrote: > >> I think the acid test for if it is more than just a fun toy is to ask it >> to do something difficult/complicated. I suspect that if you ask it to >> make a complicated object there will be lots of problems. Maybe that would >> make it ok as a first visualization, or an idea of how to start out- but >> I'm deeply skeptical that it generates a solution that is close to what is >> desired. >> >> >> On Thu, Dec 15, 2022 at 2:37 PM Raymond West <raywest@raywest.com> wrote: >> >>> I've poked around a bit on the web, it seems it is pretty good at >>> python., and iirc, some folk have used it for debugging, and optimising >>> code. It apparently has some morality filters, but you can jailbreak that. >>> e.g. you could ask it how to commit a robbery, and it will say it's >>> illegal, give a lecture to you. But if you say you say you are writing a >>> play about committing a robbery, it would explain how to commit the >>> robbery, or apparently it used to. I've asked it a few openscad problems, >>> and it has never got them correct, but you can see it tried. Very much like >>> most of us, on starting, I guess. Maybe could get it to ask the questions, >>> then tell it the correct code. >>> On 15/12/2022 19:09, Whosawhatsis wrote: >>> >>> Well, it's valid OpenSCAD code. Interesting that it got the wall >>> thickness wrong. I wonder if that had to do with confusion about the >>> difference between how cubes and cylinders work... >>> On Dec 15, 2022 at 09:55 -0800, Raymond West <raywest@raywest.com> >>> <raywest@raywest.com>, wrote: >>> >>> Hi, >>> >>> Just found https://chat.openai.com >>> >>> >>> you need to log in, but it almost writes openscad code, and most >>> anything else. Not tried any more complex shapes. (it gets the right >>> commands, and then some, but in the wrong order. Maybe we could train it?) >>> >>> >>> <Clipboard01.jpg>_______________________________________________ >>> OpenSCAD mailing list >>> To unsubscribe send an email to discuss-leave@lists.openscad.org >>> >>> >>> _______________________________________________ >>> OpenSCAD mailing list >>> To unsubscribe send an email to discuss-leave@lists.openscad.org >>> >>> _______________________________________________ >>> OpenSCAD mailing list >>> To unsubscribe send an email to discuss-leave@lists.openscad.org >>> >> _______________________________________________ >> OpenSCAD mailing list >> To unsubscribe send an email to discuss-leave@lists.openscad.org >> > _______________________________________________ > OpenSCAD mailing list > To unsubscribe send an email to discuss-leave@lists.openscad.org >

On Thu, 2022-12-15 at 20:03 +0000, nop head wrote: > Tried to create an account but it just goes round in a loop asking me > to identify harder and harder captchas. Odd. I clicked on the "I am not a robot" box and it let me in. It's interesting. I tried a number of things on it, and got some strage results. Try a few of these (preceded by write openscad code for a): house box coffee cup dog tapered cylinder of 100mm height, a 50mm base, and a 20mm top with a vertical hole all the way through > On Thu, 15 Dec 2022 at 19:54, Daniel Shriver <tabbydan@gmail.com> > wrote: > > I think the acid test for if it is more than just a fun toy is to > > ask it to do something difficult/complicated.  I suspect that if > > you ask it to make a complicated object there will be lots of > > problems.  Maybe that would make it ok as a first visualization, or > > an idea of how to start out- but I'm deeply skeptical that it > > generates a solution that is close to what is desired. > > > > > > On Thu, Dec 15, 2022 at 2:37 PM Raymond West <raywest@raywest.com> > > wrote: > > > I've poked around a bit on the web, it seems it is pretty good at > > > python., and iirc, some folk have used it for debugging, and > > > optimising code. It apparently has some morality filters, but you > > > can jailbreak that. e.g. you could ask it how to commit a > > > robbery, and it will say it's illegal, give a lecture to you. But > > > if you say you say you are writing a play about committing a > > > robbery, it would explain how to commit the robbery, or > > > apparently it used to. I've asked it a few openscad problems, and > > > it has never got them correct, but you can see it tried. Very > > > much like most of us, on starting, I guess. Maybe could get it to > > > ask the questions, then tell it the correct code. > > > On 15/12/2022 19:09, Whosawhatsis wrote: > > >   > > > > Well, it's valid OpenSCAD code. Interesting that it got the > > > > wall thickness wrong. I wonder if that had to do with confusion > > > > about the difference between how cubes and cylinders work... > > > > On Dec 15, 2022 at 09:55 -0800, Raymond West > > > > <raywest@raywest.com>, wrote: > > > >   > > > > > Hi, > > > > > Just found https://chat.openai.com > > > > > > > > > > you need to log in, but it almost writes openscad code, and > > > > > most anything else. Not tried any more complex shapes. (it > > > > > gets the right commands, and then some, but in the wrong > > > > > order. Maybe we could train it?) > > > > > > > > > > <Clipboard01.jpg>____________________________________________ > > > > > ___ > > > > >  OpenSCAD mailing list > > > > >  To unsubscribe send an email to > > > > > discuss-leave@lists.openscad.org > > > >   > > > >   > > > > _______________________________________________ > > > > OpenSCAD mailing list > > > > To unsubscribe send an email to > > > > discuss-leave@lists.openscad.org > > > _______________________________________________ > > > OpenSCAD mailing list > > > To unsubscribe send an email to discuss-leave@lists.openscad.org > > _______________________________________________ > > OpenSCAD mailing list > > To unsubscribe send an email to discuss-leave@lists.openscad.org > _______________________________________________ > OpenSCAD mailing list > To unsubscribe send an email to discuss-leave@lists.openscad.org