Conversational program

[Boswell]

Have you tried ChatGPT?

OMG

 

[BGHansen]

; Drill first hole
G01 Z-5 F500 ; Move down to the drilling depth (-5 units) at a feed rate of 500 units/minute
G01 X2.5 Y0 ; Move to the first hole position on the circle (5-inch radius)
G00 Z5 ; Move back up to the clearance height

I'll have to look at ChatGPT, never thought of that! I do it regularly on my Tormach using a G81 drilling cycle. G80 to cancel the cycle. Off the top of my head, it goes something like this:

G81 X 1 Y 1 Z -0.5 R 0.5 (drill cycle enabled, hit the first hole at X 1, Y 1 to a depth of Z -0.5, then retract to Z 0.5)
X 2 Y 2 ; second hole location, same Z's
X 3 Y 3 ; third hole
G80 ; turn off drill cycle

Curious about your example, should the G01 Z move be before or after your first G01 X2.5 Y0 move?

Bruce

 

[Shotgun]

Don't know, Bruce. I didn't look too closely at the generated code, but I do know that so-called "artificial intelligence" is really artificial (as in shallow) and not so intelligent. It needs lots of hand holding. For instance, it set the units to mm for a problem specified in inches, then used the inch to specify movements. Also, I didn't specify a clearance height, or depth to drill. If this thing were truly intelligent, it would ask for clarification.

But, it's a good start, as it beats a search engine and then having to piece together examples. It is supposed to be "conversational", so I'm betting the output would get better if I (for example) said, "Could you modify that to use inches, and drill to a depth of 1"?"

Offtopic: Having worked with NLU (natural language understanding) for several years now, I can tell you that this is as good as it gets for "AI", and as good as it will ever get. The computer isn't intelligent, and has no concept of machining or any understanding of what you are doing with the information it is giving you. It never asks itself, "What is the user trying to do with this information?", "How will it be used?", "What will be the result of them accepting the information I give them?"
When asked a question, a human, without ever really consider that he's doing it, will extrapolate out from the answer they give to some end result. That's the true essence of intelligence. NLU should really be called NLI (natural language interface), because it is just a fancy way to get directions from the user into the computer, very much like moving a mouse to click a graphical element on the screen.

 

[deakin]

Offtopic: Having worked with NLU (natural language understanding) for several years now, I can tell you that this is as good as it gets for "AI", and as good as it will ever get.

sounds like when i am trying to get past the computers answering the phone and convince it that i actually have a phone or internet outage. not much better when a human answers tho