Atomic Industries CEO Aaron Slodov demos AI-driven mold manufacturing from Detroit factory floor

a guest in the temple of Technology welcome to the stream how you doing Aaron there he is boys how are you how are you we're great uh I think if you turn your camera horizontally we'll have a little bit better frame over here is that possible that might be tough he's on an iPhone no that looks great there go thank you that looks great I don't know I don't know what you want to show us but we love that you're on the factory floor calling in on defense day at tvpn uh what's new can you introduce yourself what do you do where are you Etc sure uh Aeron slov CEO of atomic Industries and uh we're uh we're teaching AI how to do mass production here um we're we're in Detroit uh let me uh can I turn my camera around you should you can do anything on the phone yeah there we go there we go is that yeah this is the cedo this is uh this is a very very big boy this machine is 18 feet tall okay wow and right now we're loading in uh that job that's going there we can go down there take a little walk cool great this is a great this is our just Detroit correspondent yeah Aaron slow from from the front lines why not for anybody out there who gets motion sickness maybe go audio only right now we need to get you a gimbal in these boxes right here we've got a a huge production program going to the New York City uh Subway Transit Authority um lots of work happening obviously re reindustrialize right cool um so so wait you're doing the meme I am doing the meme I am doing the meme you created the meme this this is uh this is a very very large uh Mill and drill machine U that we used for mold making um okay it's not super apparent here but like this is going to be a mold for a drone ultimately cool um wow and that thing is basically like blasting holes in the side of it to like drill in water channels and things like that to cool off the mold while it's running super interesting and so you you were talking to me early uh we we got coffee like years ago and you're telling me about like if you need to make a lot of bumpers for cars you need a mold for that and so you make the mold but the mold gets hot and so you drill a bunch of holes in there so the heat dissipates is that still The Playbook or has it evolved a little bit yeah I mean basically what we were trying to prove out in the early days is like our Mastery over mold making right and so like part of mass production is being able to make a mold for something you know as fast as you can and doing it really well and obviously this is like the domain of Trades people uh it takes hundreds or thousands of hours to actually engineer one of these things and you know you have to build it too so uh ultimately what we're trying to do is teach a computer how to actually engineer these things in like minutes um instead of you know weeks and months and then have it flow through the fact more efficiently than it ever has been able to and then yeah once you actually put the mold into a production setting it runs more efficiently too because it's physics optimized you know so you can squeeze out way higher efficiency uh from these things when they're running so like here's here's a here's one of the programs running right now uh we've got you know got some parts here can check this out W but no idea what those are but they look cool yeah yeah do you do have you ever done any like microphones mic stands like any national security critical stuff for podcasting equipment yeah it's all all all national security all itar all the time yeah uh I I I do have a question about um like how you actually generate the designs that then you manufacture how much is you know physics-based solver deterministic Computing versus uh more like AI generative non-deterministic and then on the AI side are you getting into reinforcement learning or is this just like let's get a big pool of data and train like it's an llm yeah so um basically what we're doing is kind of similar to how self driving cars were built a long time ago um so you do have to have kind of like numerical solvers that characterize a physics problem pretty well and then once you dial that in you can start using it to seed a model uh to do something and then obviously you have a factory to help you like close the loop on on the prediction ability right so the idea here is that you know with the solver and kind of like the models that we build on top of this you have a factory that generates all this awesome data um all ultimately you're trying to do is just like predict the physics of a mold and the design of it as fast as you can um so in terms of like llms and how they play a part in this stuff it's it'll become more apparent over time I think but you know on the physics based side uh very little yeah and then like in terms of like coordination and uh I think like execution upet designs and kind of like Factory operations it's huge um a lot of that stuff can be really great like you know agentically for helping you know the efficiency here yeah there's a little bit of like uh when all you have is a hammer everything looks like a nail syndrome with the llm stuff right now where people are like let's just stuff it and everything and it's like there are other there are other algorithms in fact that might be useful and and that's okay it's you don't need to be embarrassed yeah I imagine when you guys are you know in the customer development process like you're often times there we go that's the poop uh the machine poop uh I imagine you're competing with with with kind of companies in in China for on some of these parts like a lot of like you know when you hear about like the Chinese manufacturing culture it's like respond really quickly to emails like get pricing over you know be clear ch's New Year it's yeah don't don't even think about doing anything around then but uh but like is that something like do do you respect their game when it comes to just like being like really fast really clear like focusing on cost reduction scale Etc like do you think we can learn uh from them um I think that yeah there's there's an interesting story there right which is like the grit the grit part of it and so I do think that like if you're highly motivated and incentivized to actually win as much work as you can you're GNA do whatever you can to do that right and so like a lot of American like high mix manufacturing is let me just take a look at this job and see if I'm going to make enough money off of it before I actually even like talk to you so I think this is another really interesting domain of like what software and Technology will do for American manufacturing right is like absorbing a lot of that complexity being able to like quote jobs for different types of manufacturing processes like instantaneously this is like our superpower is innovation right like figure out how to do that stuff and then you can work around that stuff right like there's I don't think there's a shortage of grit here but like we're we've been forced into you know a back alley of the value chain where we have to take a lot of high margin stuff CU somebody else is out there doing it for cheaper um so yeah I think this is kind of like where we have to focus I I had kind of a hot take a while back and I wanted to run buy you I I was saying like maybe the secret to China's dominance in electric vehicles now and drones like actually goes back to the Happy Meal toy and the fact that America can't make something as simple as a Happy Meal toy in America is a problem that might actually be worth solving uh but how does that hot take resonate with you do we need to be able to make hoty Meal toys in America yeah I mean a lot of people don't want to do like commoditized classic stuff like that right because it's it's not going to make them enough money but I do think to your point right like uh the amount of tribal knowledge and skill built around manufacturing something is what we've lost a lot of and so that's kind of the whole point of this is like recapturing it somehow like let's take the dying evaporating amount of skill that we have left and transfer it into a machine and make people you know 100 times more productive than they've ever been able to be so I think like this is a really hard problem because you're trying to teach a machine how to do some kind of like yeah crazy physics based deterministic thing but ultimately if you if you have if you have like whoops uh if you have crazy skill Mastery down now this is an experiment it looks fine there we go amazing uh yeah it's it's basically it's like it's all about like the Mastery of skill that you have and how you transfer it around so in my in my opinion it's like I think the fastest way to do that is by trying to teach a computer how to do it um and then you know obvious like think about how long it takes to train people to do this stuff it's like a decade like do we have a decade to sit around and wait right now no I me no all right I want some uh you are sideways by the way I want you are sideways now you're back better uh I want some hot takes how do you feel about good how do you feel about wh Mo uh you know buying cars uh from China oh they are I didn't know this H they are uh how do you feel how does that make you feel how does that make you feel yeah you should have gone to Detroit I know me you're in the capital Motor City come on they could have bought the whole town with like one one dividend payment yeah basically I mean that's kind of that's kind of how I feel honestly um but But ultimately I don't know I used to work at wh like 15 years ago was crazy uh and I've seen that company you know like grow over time but it's uh yeah I mean it's a great product it's a great product and I you know I went to San Francisco like a little tourist attraction to try to get in one finally and ride it after you know 15 years of not making it but um oh my goodness yeah this is Oh no you're sideways again uh can you talk a little bit about um just the onramp from our current education system gen Z gen Alphas coming up how what how do we get folks to get excited about what you're doing manufacturing hard tech generally and then is there a need to change do we need to like when I went to high school like there was no machine shop class it just wasn't an option and it feels like that took my whole class off the path of like ever pursuing anything related to machine shops now most of those people became like consultants and bankers and lawyers and doctors but uh you know do we need to bring back machine shop do we need to do more in the college circuit or is it really just great memes great conferences and people will pivot at the last second after they've uh been studying CS for four years yeah um I think it's two things like incentive obviously right um there has to be a reason that you want to do something like this so it's either going to be pay or the glory and the reason that I wanted to do this is because this is one of the hardest you know engineering and physics problems that like I can imagine it's utterly insane how hard this is and like I think I'm probably one of the only people stupid enough to do something like this but ultimately I think the challenge of it is incredibly rewarding you get to put your hands on steel you get to get cut and bleed uh it's awesome but beyond that right like if you turn factories like this into a Sci-Fi future like who's to say you can't pay somebody that works in here the same as they get as like an L5 of Google or something right like if you're productive enough I mean kind of speaks for itself um and like beyond that I think also you know training wise exposing people to this kind of stuff is ultra important um being able to do factory tours right like our buddy Zayn uh that's we need to like have factory tours going on all the time but there's there's a transformation happening right from from like the dirty dangerous and dingy into like relatively you know neat and kind of modern but like there's another version behind this too um so I think like really jazzing people up about the opportunity in terms of like Challenge and incentive um is kind of like the main thrust for this and then yeah there's memes but also I think a lot of like preeminent industrialists today could be beating that drum a lot more and you know there's super exciting companies to go work for um it's yeah it's a big party bag full of opportunity right now I want one more spicy fresh hot off the grill hot take uh from you before you leave uh and we'll have you on again because it's been great we'll have you on uh with a stable camera too so we can see that handsome face I had to walk around we're going to get you a selfie stick straight from Shenzhen T selfie stick weeks but we paid $1 for it uh last last question is there a world in the future where you got uh humanoid robots walking around that factory floor or are they un necessary with uh the other forms of automation that you have going on yeah I mean I would love to uh I think they could probably be pretty useful for you know little stuff that requires less precision and uh dexterity but as that improves obviously probably some interesting work for them to do around here um but for now it's like walking around the outside with like a machine gun so um there we go I like the sound of that yeah that actually sounds pretty good the robots guns yeah give the robots gun the robots guns yeah security a Detroit first baby let's go for sure uh Aaron it's been fantastic having you on looking forward to the next one keep uh keep pounding the table it's working I think yeah good luck thank you guys I'll talk to you soon bye that was great and uh we're staying on manufacturing right deterrence drva yes can you break it down what does the company do uh I want to kind of let him break it down but uh but yeah deterrence is a company that I have been lucky to be involved with myself since months and months and months prior to incorporation so um Dro is texting me right now robots needs guns well you know what else uh you know what else we need we need you to join the show so let's get him in here uh but yeah uh deterrence is leveraging uh robotics to manufacture uh critical components for Munitions starting with uh something called primer uh and they're doing another a number of other things at the moment so uh we can let him uh you know fully break it down but um I got some breaking news my mom is watching and she sent me a fog horn sound effect and said ran across this and thought of Jordy no way yeah have to send it over to you you can add it to the soundboard that's great I think a lot of people are into the soundboard uh and let's bring on drva