Is AI the Future of CNC Programming? | Machine Shop Talk Ep.105

[@ArthurField]

Ian, man, I love the possibilities this opens up for machine shops. It makes me also think of the possibilities with quoting jobs and reducing the turn around time on an RFQ. I agree with your comments about finding the benefits and everything you said about the taking jobs mentality. I've been to shops and worked in shops that avoided cnc for 30 years for the same mentality, and the end result was them closing their doors.

[@timmontano8792]

I was self-taught in CNC programming. When I was traversing the learning curve, I would have loved to have had a software like this. It would have made learning much easier for me. When I retired I sold my CNC mill thinking I'd never have much use for it anymore. Boy was I wrong. Now I've been dreaming of getting back into CNC milling some of my own personal projects but, I've already forgotten alot of what it took me a long time to learn about CNC machining. If I ever decided to buy another mill, it's good to know that there's something like this available. I notice you didn't mention anything about pricing or availability for anything other than the two CAM software's you mentioned. I use SolidCam so I guess I'd be SOL.

[@lasgusgeras3766]

I am currently introducing fusion, auto tooling, smart library to my manufacturing team using AI to analyze KPIs and OEE, create macros and much more.

[@jimgourgoutis]

Thanks for the great video! I am really curious to learn about how they got the training data for the machine learning models? Did they program (thousands?) of parts themselves, or did they get it elsewhere, or??? Also, I work in an ITAR shop and sending our parts into the cloud would be a big no-no -- do they have an on-site hosted solution in the works? Thanks!

[@spacechannelfiver]

The have their own factory, so a lot of this learning came from their own data producing parts for customers.

[@kubetail12]

I would imagine it has to be an option because things like ITAR. I just think your average shop doesn't have the resources to build a large GPU server do the AI Inference for CAM Assist. AI inference just means running or executing the AI model. I imagine CloudNC only devotes so much resources to a single session. CAM Assist might go a little faster if you were running on more GPUs on a local network.

[@ArikaStack]

I'm a machinist of 10 years. I left the industry to pursue other things, and the burnout was stopping me from simply doing it in my freetime. A year ago, I said AI would come to shops in some form within 2 years, and you'll be able to put a CAD file and stock into a machine, and it will offset, angular workshift, generate a program that won't crash, and pick the proper tooling based on the tolerances and material. In 5 years, I'm willing to wager you won't even need a CAD file anymore. Alot of machinists think they're immune to AI. They have alot of immunity, true, but they aren't, and they're looking at it wrong. No, your job won't be reduced to a machine's lackey, at least eventually it won't. Okuma's view on automation is not the future. Your job will be, if you're lucky, a teacher of sorts for AI. Yes, AI will replace prototyping and lean jobs. Your job, isn't under fire yet due to the costs, and when it will be, AI would have replaced so many aspects of the workforce that we'll need to look into radical solutions, quick.

[@ArikaStack]

yes, currently, you can have a programmer put together a Renishaw program that can do almost everything I already mentioned. I guarantee you that, one, that process will be pushbutton, and that two, a probe and math won't be the only things machine will have. Once manufacturers realize how much of a headache AI alleviates, they'll make things like a LIDAR array that 3d scans the machine in real time if adequately shielded. People in this trade often forget that the biggest obstacle to automation is R&D for programming. I've seen, and had to fix some pretty crazy, complicated programs that execute things like engrave a date code based on the machine's global for date and time, and I've seen that same program break because a different machine, from the same manufacturer, stored that global in a string (not a float) on the controller, and this older machine had an extra space being read when it called it back and converted to integers. Something about bits, parsing and emulating a NC controller on Windows (which I'm not against, Mitsubishi). I had to trace that program with a pen and paper to figure out what it was doing. I think that program had 5000 lines, and a negligible amount were the actual paths. Even if we stick to traditional gcode, AI will spit out 5000 lines of code with few errors much more rapidly than a person.

[@mattdroneon58]

Your ignorance is probably why you got burned out. No need for a CAD file in the future? 🤯 how the heck are you going to build anything without a model? I have been doing this for over 40 years I have gone from feeding tape into a flexwriter to having my programs automated to a very high degree. There are so many variables that one would have to be stupid to just plug an untested program into a machine and just walk away. AI has been implemented throughout my life. I still need somebody to catch the crash before it kills the machine on the shop floor. Good luck with your burn out.

[@ArikaStack]

Yes, don't walk away from the machine. Whatever. I have this mindset too. I'm not ignorant when I tell you that AI will generate a program AND part based on a prompt, and that program will run perfect 99% of the time, while the other 1%, the machine will have optical, simmed, and load based anti crash systems in places. I hate how in this industry I have to either explain my entire life story or not be taken seriously. For "angular work shift" I was talking about g68. You will literally be able to THROW an op1 down on a magbase and make an op2 out of it. I used to be ignorant like you. I'd call other people ignorant for saying things I disagree with. Sometimes they were. Other times, I learned how to put #5205=0.019 at the start of a multi pallet 5x program instead of... well, I could put many past examples here. None of them I am confident you won't argue against until I exhaust every step of why the decision was made. Good machinists are rare, and it's hard to give people the benefit of the doubt because there is already a world of difference between lean 5x work, which is what I cut my teeth on, and handwriting a program on a controller in a job shop to deal with a rev change, which I also cut my teeth on. AI IS coming, and it's coming quick.

[@ironwill8596]

I agree I'm a second operation operator for 20 years on brown sharpe screw machine and more and more jobs are going to cnc with robot arm because one operator can run 4 to 5 machines at once them my with only about to run two brown sharpe's.

[@williammorris1763]

just means homegamers are going to fill the gap and home cnc will be a boom for a time.

[@nutela6868]

I haven't been programming for long. When I was 18-19 I was still writing G code at Work and now with Fusion programming is easier than ever but AI is the new level, I'm sure

[@iloidi]

It seems like a good solution! It is a product that requires a long learning curve, but nowadays a good mix of macros and templates can standarize and optimize your processes and eliminate many "silly clicks"

[@2_dimes]

In my industry, we prodeduce approximately 1500 custom programs per shift. For more than 20 years, we have been doing this. One of the advancements was to have the boring and tool paths for the hardware that is mounted. So if I move the hardware to tool path moves. Many of the programs are 4 axis. With the volume, we are unable to qc the programs. Over the years, we have qc to parametric programs and have been able to increase our project velocity by having to tooling predetermined based on material and tools. Even the change over on the machines is optimized by common hold down.

[@MichaelSmith-lm5sl]

Thank you for sharing your experience and insights. It's impressive to hear about the scale of your operations and the advancements you've implemented over the years. Producing approximately 1500 custom programs per shift is a remarkable feat, and your approach to optimizing tooling paths and changeovers clearly demonstrates a high level of efficiency and innovation. The use of parametric programs to streamline quality control and increase project velocity is particularly noteworthy. By predetermining tooling based on material and tools, and optimizing changeovers with common hold-downs, you’ve managed to significantly reduce downtime and enhance productivity. I’m curious to know more about how you’ve handled the challenges associated with 4-axis machining at such a high volume. Additionally, it would be interesting to learn about any specific software or systems you’ve integrated to manage and automate these processes. Your ability to adapt and continuously improve over two decades is truly commendable and serves as an inspiring example for the industry

[@2_dimes]

I appreciate the discussion. In our industry, approximately 90% to 95% of our work is parametrically described. This means our products are defined through parameters such as size, material properties, and specific features, all managed via SQL and a user interface. At the backend, it's all about logic statements concerning each component's features, hardware requirements, and more. For instance, as objects increase in size, they may require more hardware to secure them, resulting in a greater number of hold-down points and adjustments to our bolt hole patterns. Factors like pressure levels on plates also influence bolt selections and configurations. All this engineering logic is encapsulated within algorithms. For instance, if we have a 6-inch pressure vessel with a specified PSI, the software calculates the necessary bolt hole pattern adjustments, possibly upgrading from 5/16 to 1/2 inch bolts and adding additional holes accordingly. Similar considerations extend to thread engagement depths and adjacent hole specifications. These processes, conceptualized and refined over two decades, have seamlessly integrated into our operations. Today, our focus is less on deliberating over these details and more on our sales team specifying customer requirements, such as PSI ranges for pressure vessels. The software then autonomously computes the requisite configurations and associated actions, even accommodating variables like pressure gauge selections and their corresponding threading requirements. Shop tolerances are based on a din standard in a standard in the shop regarding thread sizes holes depths and other features. This cuts down on the number of engineering hours to constantly Define the same feature work. Once a customer approves a quote, it moves to our production engineers for final review and onward to the production floor following engineering approval. Each part's routing through our shop is determined by its designated work category and the capabilities of assigned machines. Specialized components may follow distinct paths, often involving more traditional machining methods. This approach, though representing the vast majority of our operations, underscores our reliance on algorithms and logic statements akin to AI, albeit on a smaller scale. Our parametric approach defines rules that adapt to diverse material sizes, properties, and end-use requirements, ensuring efficiency and precision across high-repetition, variably-demanding products. When considering new products, our aim post-sales is swift integration into production, typically within four business days, followed by efficient workflow management to achieve dock delivery within two weeks from client release. Procurement remains seamlessly integrated within our SQL database and ERP system, ensuring real-time management of material demands throughout the production lifecycle.

[@2_dimes]

@@MichaelSmith-lm5sl kzbin.info/www/bejne/oIfZmZt4e6mJmtUsi=Jiup8zvgWDfZAYHI The large white shroud was done with the above mentioned methods

[@crazytrutas]

This is great. Thanks for the video man.

[@iansandusky417]

Thank you very much for checking it out!

[@jasonruch3529]

Anytime you can cut setup with programming time you have a big win😊

[@iansandusky417]

My thoughts exactly!

[@MRunar007]

hope it get opensource version out after some time

[@173roberto]

Awesome technology, I think thats the future in CNC programming. We will be more time in front of the machine instead of being sit in front of the computers. Mazak also tried this some years ago.

[@iansandusky417]

My thoughts exactly! I keep going back to the point of how point-and-click CAM software at one point seemed too good to be true. You mean, you just select your faces to be machined, and the software will program the radial chip thinning adaptive clear approach to maximize engagement? You can't possibly trust that! This feels very much the same to me.

[@rapturefox7068]

Been a machinist for one month. 12.5 hour shifts. Mandatory overtime. Hardinge and Mazak. I freaking love love it!! I should’ve done this 20 years ago. I have cybersecurity training and coding so the G and M code is so easy and caveman could do it

[@jonathanhare9970]

It's interesting for sure, but I fear it will not be as useful in an industry that many customers require you to protect their data, uploading something to the cloud could potentially violate that requirement.

[@iansandusky417]

That's definitely something to keep in mind - I'm sure the CloudNC people would know more about how this works in order to navigate situations like these.

[@thisismyaccount456]

With Fusion being a cloud based software this can become problematic for some users who, like you say, have certain customers that need their data protected. However, with the beta release for Mastercam underway, as well as Seimens NX and Solid Edge CAM on the horizon, you wont encounter the issue of the cloud based storage on Fusion.

[@OtisFlint]

AI is the future of everything, it's only a matter of how long it's going to take.

[@iansandusky417]

It’s not going to solve every problem - but people thinking there’s any positive to being resistant to it are going to be in for a shock. I keep going back to the CNC example - certain things are going to become prevalent because they’re efficient - and it’s always better to be open to taking a look and seeing what works rather than getting steamrolled by it!

[@hunterb108]

I think as long as it serves as a good tool for modern day problems I have no issue with it tbh.

[@kubetail12]

What I notice at my workplace is that CAM programmers like to do things their own way and sort of don't get along to the extent that I consider it a bit toxic. My company would want engineers to build our own version of CAM Assist in addition too our daily duties, then it ends up in the pile of things might work if we could just hire that one engineer who knows something no one else does.

[@michaelespiritu2908]

Can’t wait for this to release to mastercam with 3+2 option

[@pb68slab18]

Artificial Intelligence just might save us because genuine human intelligence is getting very, very hard to find!

[@rapturefox7068]

😂epic

[@AnthonyGriz]

That's awesome. That is essentially the process for which I left a recommendation on the Fusion 360 forums some years back. At the time I simply just wanted a way for Fusion to recognise my available tool library, determine the basic parameters for speeds and feeds based off my homebuilt CNC router attributes, to do the bulk of programming for me, so I could spend more time focusing on other work I needed to get done. Will definitely be looking into CAM Assist, thank you!

[@iansandusky417]

My pleasure, thank you very much for checking it out!

[@Lwimmermastermetalart]

When CNC came out everyone did say it would eliminate jobs. However when I first started to purchase I grew and hired more ppl. Problem today is there simply aren’t enough good ppl in the trade. Why we HAVE to rely on technology to keep producing affordable parts. AND to keep work here rather than China

[@iansandusky417]

My thoughts exactly. It's no different than the old-school G-code NC programmers being skeptical when CAM software came out. "So you mean you're just going to trust this new-fangled Mastercam to pick up an edge and create a contour without making the NC code yourself? I'm not so sure..." - and look where we are now!

[@rapturefox7068]

Boom!!

[@kw2519]

The pay ceiling is the issue. It’s hard to get much higher than $40-$50 an hour. It’s been making me look at other industries to increase my pay ceiling.

[@Lwimmermastermetalart]

@@kw2519 $50 an hour….I remember when my shop rate was that much lol. Yes we are the most underpaid trades ppl there are. Why? I don’t know. We certainly know more than plumbers and electricians and look what they make. How about a doctor . My wife and I had a motorcycle shop for years after I retired and sold my machine shop. One day I had a heart surgeon in to fix his bike. I had joked with him and said I should charge the same as you. He came back with this. Well you’re working on the heart, being the motor while it’s not running. I work on a heart while it’s still beating. Good answer Doc. Just a little humor to a rather sad situation.

[@JClaytonWaltersIII]

This is amazing!

[@mattr7442]

1st point "lack of skilled labor": Manufacturers are themselves to blame. Getting rid of apprenticeships, outsourcing, & not seeing manufacturing as a competitive weapon. I have visited more than two dozen shops around Midwest USA; maybe three of those had any apprenticeship programs in place. 2nd point "automation". You need an educated, skilled & experienced labor to run automation effectively. Too many variables here for even an MBA student to understand. 3rd point "AI". You still need an educated & skilled labor to teach AI the 1st steps for it to start running. AI won't start working until you teach it the initial routine. 4th point. Same thing happened with CAD based CMM programming billed as a new revolution that anybody even HS grad student could do. Not so fast. Even with all that CAD help, you still need skilled CMM programmers preferably with manual (mechanical) inspection & GD&T experience to do this effectively. A newbie even straight from 2 -year college with applied science diploma & after level 1 programming course is not able to do this.

[@georgebaragan]

The Ai is here to stay so I believe it is just going to improve and that is happening very fast. I am 100% in 3 years, and we will see machines with Ai implementation in the controller. I can't wait 😊

[@me2609-l2i]

they already exist, just aren't that great

[@tomcat7661]

As you say, its here, you can't ignore it because your competitors certainly wont.

[@ipadize]

8:12 "i have welded this tool to the table" 💀

[@iansandusky417]

It was a Freudian slip because I’ve definitely just about welded a tool to the table before 😂😂

[@jeffwinkelman6574]

Mastercam has Feature Based Machining, basically that, but the setup of it to do good things is a bit. But most programmers will tell you, it takes as long or longer to go over someone else's work to prove it out. Not exactly working faster, or really easier.

[@iansandusky417]

Mmm, I'd disagree on that - going over someone who's abilities I'm not familiar with's work is definitely going to take some time, and it would likely be faster to start from zero. When it comes to CAM Assist, it behaves in very predictable ways and programs in the manner which you set it up, so review is very quick - but yes, it should be reviewed. You really need to give it a shot to see it for yourself - it sounds too good to be true, but it's incredibly easy and intuitive to use. There's a link for a free trial in the video description if you'd like to give it the ol' test drive yourself and report back!

[@MrAarms91205]

It takes time to get in advanced level. Future is all in AI.👍

[@Juxtaposed1Nmotion]

dooo it, take it!

[@rowdyrager-9403]

AI would be nice for me less learning but, I would still need to review what was generated which is basically what I do now while I am learning. My programs are done for me buy another machinist. Then once I get a better understanding of CAM ill program my own stuff.

[@kw2519]

Learn g code. Being able to read it on the fly will be very useful for learning what you’re actually doing, rather than relying on the cam software to get it right.

[@Bigwingrider1800]

I'm pretty quick as it is. No need unless free I may play

[@martindekker8203]

As more Machinists use this AI the more it's going to learn so it's only going to get better. As a hobbyist I would definitely use this and learn all the paths it uses. I don't believe it will effect jobs but instead will increase productivity and help us create more very complicated parts to mill.

[@iansandusky417]

Bingo - this is my same train of thought! There’s a link for a free trial of the software in the video description if you’d like to give it a whirl and see what you think!

[@derekfriedrichs]

It’s not there yet. But it’s coming. Can you say minimum wage…I test drove CAMASSIST yesterday and while it’s no where near being helpful today, it will be soon. I prefer to use 2D adaptive roughing. Right now there’s no option for that. Everything is 3D. It’s coming though. I can’t emphasize that enough.

[@me2609-l2i]

it will be a little while. we have a AI cnc mill currently in the shop. Now this is one company, but so far very limited, not the greatest programs, and tolerances very suspect. I'm sure in one way or another it will no doubt be a big factor in the future, but as a programmer currently i'm not threatened by what i see to date. And that could change in a minute with technology obviously. But i remember being intimidated by 3d printing, thinking this may be the end of the toolmaker. That was a while back, and that fear has been gone for a while now. Right now, alot of what we got for our $ was snake oil in terms of AI programming.

[@raider1628]

How are you liking fusion compare to mastercam?

[@iansandusky417]

I’m a Mastercam man through and through at the end of the day - I just “think” better in it but I do try to keep at least semi up to date on both

[@bogdanyelenkov4440]

​@@iansandusky417I've never worked with fusion 360. But I've worked with mastercam, and I can definitely say that mastercam are the worst software I've ever seen.

[@jasonruch3529]

@iansandusky417 i think mastercam has way less tool path errors

[@lilchirp4846]

Can you model with mastercam? Can you switch to and from design to cam? Make quick changes? Add new features? I'm not familiar with mastercam but for the reasons listed above I see an advantage to using solutions like solidworks, inventer with cam plugins or fusion..

[@iansandusky417]

@@lilchirp4846 You can definitely do all of the above - but I will say that I find Fusion easier and more intuitive for modelling. Fusion is a lot cleaner and modern, and I do tend to like the way it approaches toolpaths and such better in a lot of ways. To be honest, if I had started 17 years ago on Fusion (if it had existed) I probably never would have bothered learning Mastercam. The real issue is I don't see any 'real' improvement in Mastercam with each of their yearly releases, and I only see Fusion getting better and better. One of my biggest gripes with Fusion is how backwards they are with the ability to store your own files locally.

[@tone3560]

I think ai cam assist would be great for people that run 1000's of smaller parts. For large jobs like myself in the mold making industry it would be an absolute pain to dial it in for one off parts to the point it would be easier for me to just program it myself. Also when you get into 3+2 axis there is so many variables of efficiency that I believe ai is no where close to computing. So in retrospect smaller, simpler parts, 3 axis ai cam assist would be great to dial in. More complex parts, larger, one off, no redundancy in contours or tooling to a degree, 3+2 axis would be a nightmare at this point.

[@willysnowman]

The cream of the crop!

[@iansandusky417]

😂😂😂

[@richhuntsd12]

Very interesting video Ian. Is this an independent company or is it part of Mastercam or Fusion? We’re you a little nervous when you ran the part on your machine? Pretty awesome I would say. I want to check it out.could be very helpful possibly

[@iansandusky417]

Thank you sir! I was a bit nervous of course, but as it exports all of its operations directly into Fusion, you can see and verify that it's not going to do anything silly prior to running. They are an independent company as far as I know! I'd highly recommend checking out the free trial linked in the description of the video if you'd like to play around with it for kicks!

[@mattdennison201]

I wonder if it us true Ai, learning more as you use it, or just a cloud templates applied to any model that is uploaded. We will see....

[@iansandusky417]

That's a good question - I'm not exactly sure the back-end but I'm sure the CloudNC people would be happy to answer any questions!

[@monsieurb24]

you can make template under fusion 360 you can do the same thing with mastercam feature recognition ....

[@iansandusky417]

For sure, FBM has been in MC for quite a while, but I notice a lot of differences between the two. I'd recommend grabbing the free trial in the video description and trying them side by side on a part so you can see how each behave - it's pretty neat to play around with.

[@nileshtonage7048]

Very Informative video, I'm from India

[@iansandusky417]

Thank you for checking it out!

[@ash26yo28]

Awesome video 👍🏻👍🏻

[@iansandusky417]

Thank you very much for checking it out!

[@tylergilbertson4086]

Not 100% sure but isn't that what mastercams FBM feature does?

[@iansandusky417]

FBM does something kind of similar - not quite the same though, and I find certainly not as comprehensibly. There's a link to a free trial in the video description if you'd like to try 'em out side by side!

[@tylergilbertson4086]

I'll have to check it out of I start doing more job shop work. Right now all my programs are just needing tweaks for products I make.

[@dejanvanevski4399]

Can AI clean machine,lube machine,work with vmc from 80 and fanuc om.Pure sci fi😊

[@jamesharrison5004]

CNC programming via AI should be a relatively easy thing to do. The programme knows the shape, it knows the stock size, the material, the machine and what tools are available.

[@Theo-at-CloudNC]

That’s what we thought when we first pitched this to investors nine years ago 😅 We said it would take two years to build a working prototype. They said we were sandbagging and would probably be done in 12 months. 8 years later and a team of 160 we finally got CAM Assist out. This is legitimately one of the hardest problems in computer science.

[@jordan-eu8db]

@ville5695"we know the machine" I'm sending a huge............. nope. There are so many factors, and one of the things in this video that speaks volume to the perception of the influence AI can have here is 1 click. It may be that this is just going to be a "skilled trade" for a long time. Conversational programming, helping fill gaps for lack of skill, has been around, that does not solve the skill gap issue. at the end of the day, the idea that machining can be, basically a Mcdonalds job, because of AI, is a long ways away.

[@heliarcweldandmachine]

i can finally get into cnc ! great video mate

[@iansandusky417]

Thank you very much for checking it out!

[@andrewdafoe4365]

How much is it gonna be?

[@iansandusky417]

I believe all of their pricing is on CloudNC’s website - there is a free trial as well!

[@lilchirp4846]

Didn't find pricing, how much will it be?

[@SamMiller-x4f]

Does anyone see a possibility of the part that you ask A.I. to help you with get sold to your competition?

[@SamMiller-x4f]

If you don't understand this you're gullible.

[@poetac15]

I think you are missing the bigger picture. Those cases will be a small portion of the overall addressable market.

[@iansandusky417]

I wouldn't say it's any more dangerous than any other kind of Cloud-based storage - but I'm sure there are many levels of data security involved - I'm sure the CloudNC people can answer any questions you may have about it as it seems like it would be a fairly common concern I'm sure they've considered.

[@Theo-at-CloudNC]

Yeah absolutely no chance of this happening.

[@jasonruch3529]

Coulda made it do more deburring😂😂😂😂😂😊

[@philirwin2028]

Interested but no sure how secure the cloud will be.

[@TheMadHatter626]

My thoughts exactly. Considering this AI is still very much alpha stage I do see this approach would be beneficial for testing and refining. However I very much would like to see in a few years when it's more stable to be able to be done locally for people who are concerned about security or working with parts that have NDAs.

[@bradmorris1981]

As a programmer this is a bit scary. Hopefully its a ways out and very expensive. I'd like to keep my job.

[@iansandusky417]

I genuinely don’t think it’s going to replace a single skilled programmer - but I do think it’s going to help you do more with less headaches. I look at it like the finger cam NC coders scoffing when CAM came out - “you mean you’re just going to trust it to do a contour? How do you know it’s going to work?” - I really don’t see a difference in this before long in level of trust

[@phily-ue6et]

For very simple parts, as per your model, parametric/strategy/macro programming has been available, and widely used for years. Cam systems are really designed for complicated parts.

[@iansandusky417]

There's a free trial in the video description above - I challenge you to throw whatever part model you might like at it! I think you'd likely be surprised.

[@johnfredrickson6987]

Is there more information on how this incorporates AI/machine learning? This seems like just a plug for feature-based machining for Fusion 360, with "AI" being used as a buzzword for marketing purposes. NX has had this capability for a long time.

[@Theo-at-CloudNC]

It's more of a massive scale optimisation, simulating millions of ways of machining every surface and volume, striking out the dumb ways, and putting together a sensible strategy out of the remaining solutions. Simple in concept, but the precision requirement and physics problems make it crazy-difficult in practice. It's very, very different to FBM which is basically just a big macro library you have to spend six months setting up per shop before it gets useful. We wanted this to be useful immediately on download, and immediately able to handle any milling machine and toolset. Siemens considers it a massive upgrade and is integrating us into NX as we speak.

[@MrJohnnyFred]

Thanks for the reply Theo. Does it learn from changes made by the user after processing?

[@grougrouhh1727]

you can do more but your value also goes down you can do more but you can have less fun of course AI is better for productivity but it remove quite a bit of the fun

[@mattwa33186]

It's important to keep in mind that this type of AI is a crowdsourcing/plagiarism model. What it's doing is figuring out how other people do this job, or this step of a job, and doing that. The AI isn't a machinist, it just has the ability to mimic a composite machinist made up of everyone who uses the software. That means it's going to get better over time as long as it continues to get feedback from real machinists or at least people who inspect and approve parts. It also means, as another commenter pointed out, that this functionality won't be available in environments where keeping proprietary information private is required.

[@iansandusky417]

I think that's a big claim - CloudNC actually has a full manufacturing facility with a ton of CNC machines that they used to develop this. While it does use Cloud-based AI to process, I'd hesitate to say that it's 'plagiarizing' anybody's work as it doesn't pull in anyone's finished existing CAM programs to generate code - it pulls exclusively off the models. I'm sure the CloudNC folks can speak to it more than I can, but I'm sure there are levels involved to protect information. I wouldn't say it's any less safe than Fusion's native ingrained feature that stores all of Autodesk files on a Cloud. Of course, if you're doing controlled goods, ITAR parts, or NDA protected information - you'll need to do your homework to make sure you're adhering to all the requirements as with using any software.

[@Theo-at-CloudNC]

Yeah we don't actually do any learning on customer data. a) partly because everyone hates that and b) because it's impossible to know whether a program is good - you can't see if it successfully machined the part, your dataset would be full of dud at best and dangerous at worst programs. So we have to develop it using our own machines, machinists, work to close the feedback loop. You can't develop tech like this without your own factory.

[@jasonruch3529]

Still not mazatrol😂😂😂😂 jk

[@dirtboy896]

I first saw this on Patriot Machinists instagram about a month or two ago.

[@jameswalslager9281]

It's not ITAR compliant.

[@jamesr3884]

Those machines are always of money

[@RFT2112]

Machinist of 30 plus years. I would never recommend anyone get into this trade, there is no money in it! Plumbers, electricians, and garbage men make more in a very short amount of time on the job. On top of that dealing with managers that think you run a vending machine that takes no skill.

[@kisspeteristvan]

it DID take jobs, maybe you're too ignorant to see , but it is what it is , we have to accept it . I remember my coworkers talked about working in the 1970's , everyone had a simple task , but thouslands of people worked there , now to produce the same parts we only need houdreds . So what happened to the thoulsand of jobs , were they NOT taken? they WERE . But regardless , that's the past and we need to focus on the present , maybe future .

[@mrechbreger]

This part is relatively simple, I wonder what AI is doing there. The milling procedures are still handled by fusion & co. I modified FreeCAD to handle parts easier for me not that automatic but pretty good. In any case still need to do all the tool setup first, apply the settings of the corresponding machine etc. I don't see the advantage of AI here yet - plus you're very likely uploading your object to them, I'm pretty paranoid about my data.

[@iansandusky417]

That setup you're talking about took me literally five minutes, no lie. There's a free link to download a trial in the video description - why not whip up a fake part model and take it for a spin to evaluate it for yourself?

[@mrechbreger]

@@iansandusky417 You know in freecad you can easily program those things yourself. Analyse the part a little bit, roughing, facing, drilling. Once the tools are set, the stock is fixed it's fairly easy even with freecad and it's just python coding. It would not work for my purpose with oversized parts where I have to adjust parts, or where I have to turn around parts. I guess it will have some purpose for some things.. I'll keep watching it.

[@JV_CNC]

Is this really AI ? In my opinion it is just CAM as usual in which you need a lot of input to become the wanted output. With CAM assist there is a little bit lesser input and that's a good reason to further develop it until it really takes decisions as a human does. I'm looking forward to further evolution. At this moment It is very dangerous to give this tool to a not skilled programmer. Accidents will happen for sure.

[@iansandusky417]

Yeah, this isn't meant to replace skilled programmers - it's meant as a tool to help them avoid the busywork and tedious parts of programming. Much like how CAM seemed scary and prone to error when the original NC programmers were introduced it, this is the same idea. It's a tool in the toolbox to do more with your time, it's goal isn't to do print to parts with zero human checking for sure!

[@THEJOEBECK]

“No,” is the answer to this question… at least in this lifetime.

[@GunnysPicks]

This won't help for security sensitive parts. Everyone is rushing into the cloud, without any worries that all your data will be "out there" in cyberspace.

[@Theo-at-CloudNC]

Yeah this has always been an awkward problem, I think we'll need to do an on-premise box in the future to help with this, but right now the product is evolving so fast the box would be out of date every two weeks. It's coming in the future, but I can't provide a timeline for this one.

[@Bicifaidate]

artificial intelligence is so fascinating as it is terrible...

[@Habkoan]

It's nothing new here... Featurecam did it 10 years ago.😅

[@VictorHernandez-nt3tw]

Whatever happened to featurecam? I used it back in 2009 and it was easy and yes like you said, you can set it up to program automatically.

[@Habkoan]

@@VictorHernandez-nt3tw Autodesk bought it in 2019. Now it is not the same. Some parts are now in Fusion360.

[@nathancoy8066]

I don't see how it's much different than CamWorks TechDB or SolidCAM's iMachining. Been around over a decade. It does 80% and you still gotta go in there and babysit all the stupid shit it does.

[@Theo-at-CloudNC]

Yeah so iMachining is SolidCAM's name for the adaptive/trochoidal roughing strategy, or a continuous engagement toolpath. You're right nothing new there, but it did revolutionise rough machining, esp in harder materials where parallel countours would just overengage and trash your tools, or force you to use the lowest required F&S over an entire toolpath, monstrously innefficient. We use SolidCAM, Hypermill, Fusion and Mastercam at the CloudNC factory and oh boy do these adaptive strategies do a lot of dumb shit. Figuring out how to proactively prevent the dumb shit when CAM Assist is determining a strategy 'before' it gets output is something we sink a lot of software engineering hours into and aren't done with yet. TechDB is just a macro database/feature based machining approach i.e. you tell it if you see feature x with y properties in z material apply this series of steps that you or I have to create. It takes months to make one of these databases useful, and they're really not robust or scalable across medium to high part variability or into 3+2 axis. Pretty much all the corporates stopped developing feature based machining approaches a long time ago. You can still buy them, but they're not seeing any meaningful software work under the hood for many years now. I want to be able to download a plugin, and be getting good machining strategies out in minutes, not months regardless of which machine, toolset, part or material I throw it at, and for it to be improving all the time. It's not perfect and never will be, but I'll take working backwards from something 80% done over starting from scratch every time in batch work.

[@iancornwall1388]

Looks like I spent the last 12 years building a skill set that will become obsolete. Suppose it’s time to get an engineering degree, or go flip burgers for the same money in 5 years. Companies will see the capacity this has and hire fucktards to load tools into a machine and hit cycle start. I don’t see this increasing productivity, I see companies cutting middle class wages out of their budget.

[@iansandusky417]

I mean, would you consider the guys who built the ability to program purely in G-code / NC prior to CAM software having wasted their time? No, they just spend less time and make less errors using CAM software to be able to do more with their time - this is no different. It's about taking the time-consuming busy-work out of programming. A skilled programmer still needs to be able to look at what's happening, make adjustments, and retool anything they'd prefer to their liking. I'd rather not have to select every edge and fiddle with every chain if I don't have to!

[@iancornwall1388]

@@iansandusky417 I understand what you are saying. And I do agree. But 5-10 years down the road companies will cut programmers out of the picture. This software will be 99% accurate instead of 80%. Why hire someone for a decent wage when you can hire someone for half the money and do the same thing.

[@Falcon4646]

You've got very little to be worried about. Look at the part they made, it's 95% straight cuts and all 2 dimensional cuts. It's going to be a very long time before AI takes in considerations such as process planning, optimization, process improvements, fixturing considerations, debugging, drawing revisions, process tolerances, and a whole host of other decisions you make every hour. I spent 35 years programming machines and changed as CAD systems changed, and while the systems got better and more robust, it still required someone to drive the bus. Even in additive machining, it still requires someone to drive the bus. Design Engineering still requires someone to drive the bus.

[@iansandusky417]

​@@Falcon4646 1000% agree with your statement that it will always take someone behind the wheel, and that's a good thing. While the part I did was a pretty simple one, I think you'd be surprised how powerful this software is with more complex geometries. There's a link in the video description to a free trial if you'd like to pop some more interesting models into it and give it a shot yourself to see how it works first-hand!

[@Falcon4646]

@@iansandusky417 I used to joke around with my boss when he'd tell me I was a "necessary evil". I would laugh at him and remind him that until they add a key on the keyboard that says "Make Part"... someone like me would be around. :-) It may happen someday, but for now it's pure science fiction. I retired a couple years ago so I'm now out of the loop, so all I can suggest is to keep up with the technology changes in systems and perishables, and don't fight it. Stay at the forefront and you'll be fine. By the way, I worked with Unigraphics in the aerospace field since the early 90's and saw some pretty cool changes with FBM. My success was based on getting it right the first time. Best wishes for a long and rewarding career.

[@jarodgonzales4734]

Short answer is no and long answer is no, you can’t set up a machine with AI end of story

[@usrenmae]

Running it for a while. Its nothing special. Its no better than templates or fbr. The toolpaths it picks are often difficult to edit because it uses a lot of 2d vs doing 3d.

[@Shilohfan69]

Bull 🐂