Will AI kill higher paying CNC machinist jobs?

[Superbowl]

I am doing some reading to try to learn about CNC mills as I plan to buy one at some point to go with my manual machines. One thing I picked up is that G-code programs, feeds/speeds, etc. for a part are set to the specific machines and can't be used on a different machine. It appears that setting up and programing a cnc job, planning tooling needs, off sets, feed rates, etc. is what takes the know how and skills. The guys who do this well are true machinists and make much more than an operator who just loads stock, pushes buttons, and cleans up.

This got me thinking about future machines. The coming artificial intelligence advances will surely soon be integrated into new cnc machines. Will they they will take over much of the value of the special knowledge and experience of a top machinists, leaving employment just for lightly skilled machine operators?

Imagine a machine that can plan ahead and choose tools and speeds for each different material and adjust itself on the fly. It can analyze a CAD file and choose the best stock cut plan, design a work holding plan, the best order of operations, the best milling cutter to use, etc. from its vast knowledge base loaded by the manufacturer for cutting the same or similar material. It can also learn from all it's past jobs to optimize a new job for maximum productivity. The machine now just tells the lightly skilled operator what do do.

A stand alone program may be able to take a CAD file do some of this for existing older machines. AI will surely eliminate many positions in the world today. Will skilled machinists jobs be threatened? CNC machines killed most skilled manual machinists jobs. Will AI kill many of the remaining jobs?

 

[CarbideBob]

Not in our lifetime.
Now it is all the news as it was in the 1980s and I was a real believer back then.
We built "The Factory of the future".. that went well.
I find all this talk/scare now about AI rather amusing.

 

[technocrat]

An AI machine would spend all day arguing on PM about the best way to do the job (and other things)
NC was going to take all the jobs when I was a lad, but there are still plenty of manual machines that make money. Someone has to draw the cad file and deal with the customer. Clueless engineer + AI = disaster. AI is just in fashion and can do a better job than clueless screen scrollers, so they are impressed.
As a second opinion, AI may be useful but it has a long way to go.

 

[john.k]

AI is taking jobs that are all talk ...........good for instance ,the ABC ...Australian Broadcasting Comission.......ABC pseudo news from AI ... ....must include lesbians ,acceptable abos ,disabled women ,evil/violent/stupid white males ......especially conservative politicians......Press go for 'puff pieces' to pass as news.

 

[gregormarwick]

I think that some of you are being a bit too casually dismissive in light of the massive leaps that ai/deep learning has taken in the last 1-2 years...

There are ai bots right now that can create truly incredible works of art based on a few sentences of direction, or write code that actually compiles and does what it was asked to, or build a functional website, again based on some brief natural language input.

Why do you think the big push exists to get all of us to put our CAD/CAM data in "the cloud"?

10 years ago we could surely say "not in the next decade, haha" - I don't think we can say that now with too much certainty.

 

[technocrat]

The big push to get things into the cloud is to remove your autonomy and make you rely on the supplier. No subscription = no working CAD.
The companies that run the the cloud are not in the cloud as such, it is their cloud. Sure their network is distributed and can leverage cheap cpu time and resources, but that is the same as having 2 pc's, only bigger.
The art is not art, it is just a copy, based on some learning about what we think is good art. Same as many art critics can't tell the difference between the real thing and a forgery. It's incredibly technically good based on the description, but not original. Same with the code writing, the learning from open source software such as github creates software often in violation of copyright/left, again not original. I agree that It is similar to the code you get from an inexperienced programmer.
The people who's job it takes were going to lose their job anyway, just like the crap manual machinists. In any downturn, they were the first to be fired.

 

[technocrat]

Could AI replace a person issuing QA certificates?
Would you trust the cert?
How would tell if it was fake?

 

[EmGo]

Same with the code writing, the learning from open source software such as github creates software often in violation of copyright/left, again not original. I agree that It is similar to the code you get from an inexperienced programmer.

I'm convinced fully 40% of websites are already created by ChatGPT ... and maybe 70% of open sores. They leased time on it for the gtk3 project, I believe.

 

[gregormarwick]

The big push to get things into the cloud is to remove your autonomy and make you rely on the supplier. No subscription = no working CAD.
The companies that run the the cloud are not in the cloud as such, it is their cloud. Sure their network is distributed and can leverage cheap cpu time and resources, but that is the same as having 2 pc's, only bigger.
The art is not art, it is just a copy, based on some learning about what we think is good art. Same as many art critics can't tell the difference between the real thing and a forgery. It's incredibly technically good based on the description, but not original. Same with the code writing, the learning from open source software such as github creates software often in violation of copyright/left, again not original. I agree that It is similar to the code you get from an inexperienced programmer.
The people who's job it takes were going to lose their job anyway, just like the crap manual machinists. In any downturn, they were the first to be fired.

You sort of missed my point and sort of confirmed it all in one post.

Sure there are multiple benefits to the software supplier to removing our ability to work offline.

About the art being a copy - yes, sort of, sort of not. That's what deep learning and predictive compute are all about - looking at what's been done, analysing it, recreating it, building on it. The images/codebases/websites/etc.etc.etc. being created by ai algorithms right now are not copies, they are unique, albeit derivative of everything they've learned.

How much of writing a cam program is repetitive, how many man hours are spent ever year recreating the same steps in any given piece of cam software across the entire userbase? The AI bots have learned all this already with all the examples I gave above, what makes you think they will not achieve the same thing with cadcam? The Fusion360 document base belongs to autodesk, how many millions of cam programs exist there? Do you really believe that there are not already deep learning algorithms sifting through and absorbing all of that as we speak?

 

[EmGo]

About the art being a copy - yes, sort of, sort of not. That's what deep learning and predictive compute are all about - looking at what's been done, analysing it, recreating it, building on it. The images/codebases/websites/etc.etc.etc. being created by ai algorithms right now are not copies, they are unique, albeit derivative of everything they've learned.

Totally disagree. There is not even a fraction of a brain in any of this current "Artificial Intelligence". What it lacks totally is "intelligence". It's nothing but a huge collection of gossip that it condenses then spits back.

It can't do a fucking thing except copy, badly.

 

[gregormarwick]

Totally disagree. There is not even a fraction of a brain in any of this current "Artificial Intelligence". What it lacks totally is "intelligence". It's nothing but a huge collection of gossip that it condenses then spits back.

It can't do a fucking thing except copy, badly.

That's maybe a touch harsh, but you're not wrong.

Point being, it's extremely good at figuring out how to replicate repetitive functions with minor deviations from a fixed process to achieve fixed goals. As someone who automates a lot of my cam programming, it's not hard to see how a contemporary ai could start making parts.

 

[EmGo]

Point being, it's extremely good at figuring out how to replicate repetitive functions with minor deviations from a fixed process to achieve fixed goals.

Okay, if you're going to define it as a really advanced version of roughing cycles, then I can go along with you. That's been the trend for years. I just can't call it "intelligence" because it will never come up with anything new, it's just a rehash of what 10,000 people have done before.

In that sense, I can see it replacing programmers. Or maybe not replacing, just reducing the number of people required, like cad reduced the numbers of draftsmen but didn't replace them, just made one guy do the work of ten.

Not as well as the ten either, but that's a different subject

 

[DDoug]

^^^^ Sounds like we've achieved the 10,000 trained monkeys at 10,000 typewriters.....

 

[TAIWA NUMBA WAAN]

AI won't kill CAM programming jobs overnight. But I think over the course of this decade, a few of the CAM software companies will integrate AI into their software in a way that automates the vast majority of the programming.

It will analyze the part, either a 3d model or even a 2d drawing, then generate as many roughing and finishing tool paths as it can to make the part from a known set of tools in the library. It will generate multiple programs, compare them on cycle time and other factors, then tell the operator what tools to set up.

It will be a subscription service of course, and they it will spam a shitload of tool and equipment advertisements to the poor guy who has to check it and run it. But if it makes functional, good programs instantly, people will pay for it regardless.

I think whatever company does this first will try to offer it for free and make money off the included advertisements and resulting sales. It will be like what Google did with it's search engine from 2000 to 2020, or what Amazon did with that stupid portable house speaker thing that tries to order dishwasher soap with your credit card information.

 

[hanermo]

I believe it is quite certain that AI will replace most machining jobs ..
but it will likely take about 20 years.

It´s about as hard as a self-driving car, and tsla has been spending about a 1B$/yr at it for 6-7 years, and is still struggling to make it work.

About 90% of machining is straight forward and could be automated easily.
But 1% of the total will wreck tools, machines, and blanks .. and these are extremely complex to identify automatically.

If someone gave me 6B$ I would suggest a well-functioning system for 95% of cases in 10 years, and a 90% probability of 99% success within 20 years.
90% of cases would work in 3 years for 10M$.

 

[gregormarwick]

Okay, if you're going to define it as a really advanced version of roughing cycles, then I can go along with you. That's been the trend for years. I just can't call it "intelligence" because it will never come up with anything new, it's just a rehash of what 10,000 people have done before.

In that sense, I can see it replacing programmers. Or maybe not replacing, just reducing the number of people required, like cad reduced the numbers of draftsmen but didn't replace them, just made one guy do the work of ten.

Not as well as the ten either, but that's a different subject

I still think you're underestimating it, and I'm not sure why you're hung op on "new and original". CAM programming doesn't involve anything new or original. It doesn't matter what insane outside-the-box approach you're taking, someone somewhere has done it before, and for 99.9999% of the tasks involved in writing a program for any given part of any given complexity, a million people around the globe have done it ten million times before you, in some kind of general terms. That kind of goal searching via approximate pattern matching is something that contemporary ai's (yes I know you don't like the term, I don't either, but it is what it is) are proving to be incredibly good at.

Again, I really don't think we're as far away from this as some of you are choosing to believe.

 

[GregSY]

AI will be the death of us all. Will it replace jobs? Yes! Soon, no one will need to work at all...we will achieve The Liberal Dream - dancing, smoking dope, using intravenous drugs, listening to music, and eating. One of more of those will do the trick.

Can AI create art? LOL. Of course it can. It's very difficult to make the argument that AI art is not 'real' when it looks better than the poop created by real artists. And we all know that 99% of all art is stuff one artist ripped off of another artist. Art in of itself is 99% a sham; a useless pursuit. See above re: dancing and music. Art will be among the first to crumble in the face of AI, not the last.

 

[DouglasJRizzo]

When I went to work full time in 1982, I remember hearing how CNC was going to kill machinist's jobs and everyone would be a button pusher. Fast forward 40+ years, and, well, it's simply not the case.
I don't think AI would be any different.

 

[empower]

i'm probably in the top 1% of machinist jobs, at least pay wise, i'm not getting replaced any time soon, much less by AI. but i do way more than just programming and running machines so...

 

[empower]

Totally disagree. There is not even a fraction of a brain in any of this current "Artificial Intelligence". What it lacks totally is "intelligence". It's nothing but a huge collection of gossip that it condenses then spits back.

It can't do a fucking thing except copy, badly.

agree, other than the badly part, thats one of the things it does really well, copy. as of right now it is completely incapable of coming up with its own original thought/idea. until they can do that, we're fine.