Looking For More?
Watch this episode on YouTube
Connect with the Manufacturing Industry here www.amtonline.org/events
Discover the past, present, and future of American manufacturing with www.mmsonline.com/madeintheusapodcast
Explore, watch, read, learn, join, and connect at www.imts.com/
Tune in to the AM Radio podcast www.additivemanufacturing.media/zc/am-rad…o-podcast
For the latest in Manufacturing Technology news www.amtonline.org/resources
Produced by Ramia Lloyd
Ramia Lloyd:
Welcome to the TechTrends Podcast where we discuss the latest manufacturing technology, research, and news. Today's episode is sponsored by Modern Machine Shop's Made in the USA Podcast. I'm Ramia Lloyd, and I'm here with?
Elissa Davis:
I'm Elissa Davis.
Stephen LaMarca:
Stephen LaMarca.
Benjamin Moses:
And Benjamin Moses. Guys, I'm sick of home automation. I am. Well, I'm sick of how simple and how easy people think it is, and how effective people think it is.
Stephen LaMarca:
And then how quickly it breaks.
Benjamin Moses:
Everything breaks in my house.
Stephen LaMarca:
I'm honestly impressed you've made it this far.
Benjamin Moses:
So I have two kind of scenarios to walk over. So I do have a Roomba, or a robot vacuum cleaner. It is a different brand, and it gets stuck on everything. It doesn't like the carpets. The transition from the floor to the fireplace thingy area, it gets stuck on. It gets stuck on my Nerf bullets everywhere. So I walk down every morning, and I can hear it beeping somewhere. So it's like playing hide and seek at 6:00 in the morning with where did the robot vacuum cleaner fail.
Stephen LaMarca:
I love the idea of a Roomba, and I know they're only getting better, and I do really want one, but being that our house is three lazy people and a 80-pound Goldendoodle that does shed. Don't let people lie to you about Goldendoodles, they do shed. There's tumbleweeds everywhere. And using our Dyson, which is a handheld cordless Dyson that I've retrofitted to take Craftsman power tool batteries. So the battery lasts longer, but now there's a new problem with that vacuum cleaner, the chamber that collects all the dirt and debris, not big enough. And I know a Roomba, if the Dyson can contain one floor's worth of Goldendoodle tumbleweeds, Roomba is going to beep all the time. I'm going to get notifications to my phone [inaudible 00:01:59].
Benjamin Moses:
So one thing that the one I have, it'll actually go to the base station and evacuate itself into another container.
Elissa Davis:
That is fancy.
Ramia Lloyd:
That is kind of cool.
Stephen LaMarca:
And that is really cool, actually.
Ramia Lloyd:
I wish I could teach my dog that.
Stephen LaMarca:
The other problem though, again, just to fortify any notions that we're not lazy, is I will walk the Dyson down to the garage, go out the garage, and empty the dust collector outside, specifically, so none of the dust gets back inside. The problem is where I've been emptying it is right next to my AC condenser units outside, and the Goldendoodle dander has clogged the fins. So we lost upstairs AC for a bit because it just couldn't exchange the heat. Don't be a homeowner.
Benjamin Moses:
Also, one thing I've been doing is I've been using digital picture frames, and I have a series of them, like different phases of Amelia. So I've got one in our family room and one in our music area, and it's supposed to have a program scheduled to turn on and off during certain times. We're not watching it in the middle of the night or at home, so it's supposed to turn off.
But then I bought the cheaper version, let's be honest, I'm pretty cheap. It turns black, and it just doesn't go back to the program. So I was like, "Okay, I can solve this. I'll do a hard reset." So I figured hard reset works. So I bought a digital plug, and I have them all running to a digital plug that I can turn on and off through a schedule. So that is my solution to cheap digital frames that don't work. I do a hard reset that does a digital reset every single time.
Stephen LaMarca:
I think it would be really funny if you just came home and the picture frames all had the blue screen of death.
Benjamin Moses:
It's close.
Stephen LaMarca:
"This is not what I want to show off."
Elissa Davis:
My grandma loves her digital picture frame because my mom and her kids can send pictures to the picture frame.
Ramia Lloyd:
That's so cute.
Elissa Davis:
She has pictures of us, or my mom will send her pictures from things, because they don't really like to leave Virginia Beach very much. For important life events, she'll send pictures to her, like my little sister's graduating next week.
Benjamin Moses:
I do want to hit it on, because you did talk about a quick trend of garage doors for your bread now you're coming back.
Elissa Davis:
Yeah, yeah, yeah. I saw a video, and it was like this really nice kitchen. It's like a garage for your appliances, and it's like a door that lifts up. So you have your countertop, and then behind the countertop against the wall, it's like a door that lifts up, and then it's like a compartment for all your kitchen appliances. Right?
And I remember watching that and I was like, "My Nana had one of those in her kitchen, in her house that was built in the '50s," except it was just one of those wood, not accordion doors, but garage doors. And that's where she kept the toaster, that's where she kept the mixture. But the way you described your grandparents, it literally goes from the countertop all the way to the ceiling.
Stephen LaMarca:
The cabinets extend all the way down from the ceiling to the countertop, and it was caddy quartered in the corner of the kitchen. And yeah, it was just a little garage door that you manually opened and closed. They put liquor in bread, and I didn't realize it was supposed to be for appliances.
Elissa Davis:
I think it can be for anything.
Benjamin Moses:
I mean, the two things you need in life are liquor and bread.
Elissa Davis:
Yeah.
Benjamin Moses:
They do.
Elissa Davis:
But I'm like, okay... I mean, don't get me wrong. Style, in general, is cyclical because it's just one of those things where it's like, okay, so these were in the houses in the '50s. They're just coming back into style as an automated thing, because I would love to have somewhere to store all my kitchen appliances, but we've been doing that. This isn't a new thing. So people are getting all excited about it and I'm like, "This isn't a new concept."
Benjamin Moses:
I think the concept of copy and pasting is going to come up a fair amount in this episode. I do want to hit on one thing. Steve and I haven't talked about firearms in a while, but we are in the season of gardening and yard work. And-
Stephen LaMarca:
And I got a leaf blower.
Benjamin Moses:
And do you have a leaf blower?
Stephen LaMarca:
Our HOA covers all of the landscaping stuff, so I don't have to mow lawns, but I do have a leaf blower to blow all of the schmutz and detritus that is blowing into my beautiful garage.
Benjamin Moses:
And leaf blowers are allowed, even electric ones?
Stephen LaMarca:
Even my brushless electric one that's battery operated, it gets really peaky, high, and sharp. And years ago, when I started at AMD, there was this cool, not necessarily firearms company, but firearms industry company called OSS. Now, they've changed the name to Huxwrx, H-U-X-W-R-X. They're suppressor manufacturer, a silencer manufacturer, but they don't make traditional silencers that use baffles, which create a lot of back pressure but they suppress a lot of the noise. They muffle a lot of the noise.
Because effectively, it's just a car muffler, which is a reflective muffler, as opposed to an absorbing muffler, which is typically an automotive called a glass pack or a resonator, because they're packed with fiberglass glasspack. But they've been using the same science to make firearm suppressors, but because a lot of firearms today are semi-automatic, too much back pressure can be not nice to the movement of the firearm.
So this company OSS, basically looked at a jet engine, and were like, "Okay, let's redesign a jet engine. So something can go down the middle of it, but I want all of the plumbing around the outside of the jet engine, which is football fields of tubing and piping, and just room for gas to expand at a controlled speed without trying to bring it to an abrupt spot or stop." And OSS, now Huxwrx, created what is called the Flow-Through suppressor. They're not as quiet as a baffle system, but they're much better on the firearm and still hearing safe in some examples.
Johns Hopkins was like, "Hey, these cute engineering students, they're very young, which is why the baby engineers were like, 'We're at one of the greatest schools, and we can't study because we're annoyed by all these darn leaf blowers keeping the campus super pretty.' And how can we make them quieter?" So I'm going to allege and say that one of them might know something about firearms and be like, "Hmm, can we do this Flow-Through suppressor design and put it on a leaf blower?" And that's effectively what they did. I'm calling it right now. Allegedly, that's what they did.
Benjamin Moses:
But it is a fascinating lateral-
Stephen LaMarca:
But now, we don't have to talk about guns. We can talk about suppressors for leaf blowers.
Benjamin Moses:
Anyone can buy. It is interesting. I mean, we talk about the lateral transition of technology, so getting to a niche market of firearms, not defense-related, but common firearms to leaf blowers, and translating gas optimization. We were talking about gas optimization.
Stephen LaMarca:
And there's no $200 NFA tax stamp for it.
Benjamin Moses:
Talking about that for a while, so that is a cool transition. And students who looks into a new product are cool.
Elissa Davis:
I just feel like it's the definition. Don't me wrong, this is a great invention, but it's the definition of a first-world problem to be at John Hopkins University and be like, "Oh, that's too loud. I can't study." Props to you guys for doing it, but I'm just like...
Stephen LaMarca:
I 100% get where you're coming from. But as one of the owners of one of these electric leaf blowers, they're not that loud, but where they are loud is in a super high frequency. They're piercing, and then they have the audacity. My leaf blowers are loud on the normal setting with a throttle all the way open, but then there's a button to put it in hyper mode. And I can hear my hearing just deteriorate. And again, it doesn't feel loud, but... You know how a dog whistle is such a high frequency? You don't hear any. I feel like it's doing that to me, but it's taking my hearing away.
Elissa Davis:
But they're not operating the leaf blowers. They're sitting in their doors hearing the leaf blowers.
Ramia Lloyd:
Honestly, no, I kind of respect it because I would like to give them to our HOA company. Every morning at around 7:30, right before my second alarm goes off, there are just leaf blowers everywhere. It was after they've mowed the lawn that I've slept through, but it's the leaf blowers that give me every time. And I'm like, "Love you, guys, but give me 30 minutes at least. I just need a little bit more sleep."
Stephen LaMarca:
It's really important to emphasize the Flow-Through technology because if you think about using a baffle-type suppressor on a leaf blower, it's called a leaf blower. It's supposed to blow the leaves out of the way or in a particular direction. And the blowing act requires volumes of air to be pushed, and a baffle system would slow down the speed of that air. As for the Flow-Through design, it allows the air to keep moving. It dampens the peaky vibrations that make all the noise.
Benjamin Moses:
Speaking of US-based technology design, Ramia, can you tell us about today's sponsor?
Ramia Lloyd:
That was a beautiful connection. Today's sponsor, tune in for Modern Machine Shop's Made in the USA Podcast to explore manufacturing issues faced by companies making an intentional choice to manufacture in the US. Featuring commentary from OEM leaders, Made in the USA blends its nearly century-long expertise with a unique audio storytelling experience to shine a spotlight on the past, present, and future of American manufacturing. Find Made in the USA on Apple Podcast, Spotify, and all major podcast platforms. Follow Modern Machine Shop on Twitter, Facebook, and LinkedIn.
Benjamin Moses:
Thanks, Ramia.
Ramia Lloyd:
Anytime.
Benjamin Moses:
We got some testbed updates.
Stephen LaMarca:
We do have some exciting testbed updates.
Benjamin Moses:
Tell me more.
Stephen LaMarca:
Chloe and Sharab managed to get a working agent and adapter on the new pocket NC. It frankly took a little bit longer than we thought, but I never know the kind of roadblocks they're going to run into when it comes to software stuff or just computer stuff, Linux. But Chloe was very celebratory in announcing to me, "Hey, we saw data come off of the robot now too. We don't know what that data is, but we found where the data's coming from, and that it is data from the robot." So that's exciting.
More development on the robot, so we've got the $115 cable connecting the gripper to the robot. The gripper works. We know what we want to do with the testbed in terms of automating it. The first step is opening and closing the enclosure of the pocket NC, which does not sound exciting. But because we've done a lot of the preliminary boring software stuff and the very seemingly not difficult but ended up being super difficult, just plugging stuff in act, now that that's out the way, we can finally start practicing with the kinematics of the robot arm, which frankly I've been doing. So everybody else has been putting in their blood, sweat, and tears just so I can finally play with the robot. Hashtag privilege.
So I've been working on finding the movement path that the robot is going to do to open and close the enclosure. And Ben, you've made a design for some soft jaws that will go on the gripper because, while we have the gripper and it's finally bolted in and plugged in and working, the gripper does not actually have things to grip anything with. It's got joints, but it has no fingers, for the lack of a better term. It has no soft jaws on it. So we need to have soft jaws manufactured.
Georgia Tech, Kyle Saleeby was very willing, and it's ready at any moment for us to send him a file and make us some quality soft jaws, probably with some cute branding that he's going to put on there as a surprise for us too, because he's amazing. But Ben's figured out a design that we're going to do for the soft jaws, and we've also determined that we are going to submit the design to the Fairfax County Public Library system who has a makerspace of an undisclosed location. They don't want people showing up, I don't think, but they do say that, "You pick the library branch that is closest to you, and we will deliver the finished part there, but don't come to us."
Benjamin Moses:
And it's useful to go through the experiment, because it's been a thousand years since I designed anything CAD, so this is more of a test lab print, and then Georgia Tech will come up with the production version. So check, fit, and finish. So we're going to 3D print the fingers, make sure the interfaces work. The interface is going to work for the raw material and opening/closing the door. So once we verify, that works for like $20, then we can reach out to Georgia Tech to see if they can do. We do have to modify the design a little bit because-
Stephen LaMarca:
What's $20?
Benjamin Moses:
I have a feeling the cost of the fingers is going to be like $20 or less.
Stephen LaMarca:
In hardware?
Benjamin Moses:
No, no, no. Printing the gripper, they're going to charge us. Library will charge us.
Stephen LaMarca:
The library said it was free on the website.
Benjamin Moses:
They'll charge it for the material.
Stephen LaMarca:
Oh, okay. They do?
Benjamin Moses:
Mm-hmm. They charge for the material, not time.
Stephen LaMarca:
Gotcha, gotcha.
Benjamin Moses:
Yep. And we do have extra hardware, so we have it.
Stephen LaMarca:
That makes me feel better because I was under the impression that the entire thing was going to be free, and it's like, "This is not sustainable." I'm glad my taxpayers are coming back to me, or taxpayer dollars are coming back to me, but this doesn't seem sustainable.
Benjamin Moses:
It could be Fairfax taking a different approach because they have a lot of different rules in Loudoun County. So Loudoun County, when I show up, I have to pay for raw material used.
Stephen LaMarca:
Okay.
Benjamin Moses:
So yeah, this iterative approach, we did as much simulation as I could, verifying everything will work, but we need to physically see it because the path of the robot, it's kind of hard to simulate that now with all the models that we have. I think it makes sense to get right to the physical space and start doing pathing.
Stephen LaMarca:
This is interesting because this is America, everything costs money, but at the same time, it's like you would think a government organization, even state government, would be like, "Hey, if we're providing a service, we have a price tag on it." You would think that would be one of the first things that they put in there. I went pretty thoroughly through all of the documentation that they have on the website, and I didn't see anything about cost.
Benjamin Moses:
We'll find out.
Elissa Davis:
Maybe that's how Loudoun County stays the richest county in the United States.
Stephen LaMarca:
I did see the one line in the, not FAQ, just the guidelines, terms and conditions. That said, "Right now, you cannot bring your own filament. You use our filament, and you don't even touch the printer. We have operators to do this. But right now, it costs money. It's just on a first-come, first-serve basis. But if demand increases and backlogs get more and more dense, then we may put a price tag on it." But as of right now, I think it's free.
Benjamin Moses:
I'll find out. So we got hardware coming in, got nuts and bolts, or bolts and dowel pins, a drill. Since we're 3D printing it, I undersized the clearance hole for the dowel pin because I don't trust the 3D printer to get it to the size that we need. So we're going to drill that back out, and then glue the dowel pins in. And that's one of the modifications when we get to Georgia Tech, it's modifying the design so they can go to more machining tolerances and standards as opposed to 3D additive plastic printing.
Stephen LaMarca:
Yeah.
Benjamin Moses:
Cool, I'm excited. I want to get back to getting to automation. So there was an article or a roadmap published, "The 2024 edition of US robotics roadmap points to need for more federal coordination." This is published on The Robot Report. And this is published from, I think USC help organize this. And it's basically taking a look at robotics in the US and what are the challenges, our robot's current state as of this year. And to be honest, I have a lot of things I disagree with overall. So there are three main bullet points.
The first one, it calls for more federal coordination. I don't think that's really needed on a couple layers. It's asking for more federal coordination, so you already have the ARM institute, and you already have MEPs at every local, at every state. Federal coordination adds cost oversight, overhead, redtape, where the underlying problem is more of adoption as opposed to technology generation. I think that's where they're headed at the federal level, is that I think they're headed more towards technology development for robots. Which actually gets to one of the underlying things which I disagree with a lot, it talks about robots. Robots is not the problem. Robots is a subset of automation in general.
So I think their approach on focusing on... It's very myopic view of focusing on an underlying technology where there's other ways to increase productivity. And that's one of the other bullet points it talks about later, can the US regain robotics leadership? I'm like, "What is the point of gaining robotic leadership?" I think the focus should be on, can the US get back to higher levels of productivity through automation? I think they're missing a bullet on a lot of bullets here.
Stephen LaMarca:
I think there's two things that I want to bring up. Number one, when it came to the guide or the roadmap did list other competitive countries, they didn't mention South Korea. How do you not mention South Korea, not mention them as a robot powerhouse? Which they are. They are growing faster than any other country, and you didn't mention them at all? That just questions the credibility of this roadmap and who wrote it. Sorry, I don't mean to get spicy.
And then the second thing is, okay, you're calling for all of this federal coordination that you know is going to slow things down, and people are going to question that it's going to slow things down. Why aren't we transparent about risk? We need to be transparent about risk here. So you can say that, "Okay, these are the risks, but this is the risk that it would be avoiding, which is the monopolization of a robotic or automation technology," which is absolutely what we should be worried about too. We don't want literally everybody, including the US federal government, paying royalties to some company that has the monopoly on some sort of automation technology. That is absolutely bad. We are heading in that direction. There's no doubt about it. But you're not going to solve risk with introducing more risk.
Benjamin Moses:
Right.
Elissa Davis:
Well, I might get canceled for saying this, but I think it's a very American way to look at robotics, to really focus on the competitive aspect rather than the collaborative aspect. Because all of us working together can get a lot more automation and robotics put into place rather than competing and seeing who can do it first. But that is just a very American way of looking at it, and I think that's been true since the space race. But yeah, I think there's a lot more to be achieved being collaborative than competitive.
Benjamin Moses:
And they hit on another... The last bullet that I saw in the article was a national strategy for robotics. And I get the sense that what they're trying to do is copy and paste what other nations are doing. That works well for them, but that doesn't translate to the, as you mentioned, competitive, the capitalistic side of the market where what worked in Europe or other countries that are heavily focused on national businesses. Everything here, there's not a national business. There's individual companies that are thriving to grow and compete against each other.
And I think we're missing the boat a little bit in terms of how things work and transitioning that in to a more effective environment. Because, Steve, to your point earlier, we talked about the CHIPS Act, and that has presence and value in Intel and some other companies that are making factories, states are getting money to offset subsidies and things like that for the local level. But a lot of the lessons learned from the automotive transition from the '70s and '80s are kind of being missed a little bit. And I feel like this is another boat where we're copying/pasting what other people are doing, but not translating it to what should work in the US. Yeah, interesting article. I give it a quick read. I do want to talk about 3D printing in space though.
Elissa Davis:
Yeah. So we have officially started 3D printing metal in space, on the International Space Station.
Benjamin Moses:
Nice.
Elissa Davis:
So this is an article from TCT Magazine, and... Let's see. So the European Space Agency, the ESA, has declared an additive manufacturing first as a giant leap forward for in-orbit manufacturing. So they did an S-curve test line, which is literally just exactly what it says, which is just 3D printing the shape of an S in metal, but they did it. And that was in space, which is pretty cool. They are using the direct energy deposition process in space.
Stephen LaMarca:
CNC welding?
Elissa Davis:
Yep. So in stainless steel wires, heated by a high-power laser, it operates in a fully sealed box, preventing excess heat or fumes from escaping. And the process is overseen entirely from the ground.
Benjamin Moses:
That's really cool.
Elissa Davis:
So it is remotely operated the entire time. All the onboard crew has to do is open a nitrogen and venting valve before the printing starts. So you don't have to be a 3D printing expert to go to the International Space Station because they're taking care of it from the ground.
Stephen LaMarca:
That is so cool. Lean manufacturing in space.
Benjamin Moses:
I think this hit on a bunch of really interesting topics. So the impact of gravity on 3D printing in space, that is very important. Printing at point of use, that's something obviously we're very interested in. But also the fact that they're remote manufacturing. It's not the fact that they have astronauts printing, which technically they are, the way they're operating the machines. But realistically, all the data and everything is being zipped up and transmitted to the Space Station to print remotely. And I think that's a very fascinating scenario.
Elissa Davis:
So the project known as Metal3D started in 2016 when the ESA awarded the contract to Airbus Defence and Space. They spelled the European way with the C for defense. And so while polymer 3D printers had been on board the International Space Station and printing parts since 2014, this metal technology arrived on board the International Space Station in January. And the aim was to create the first metal 3D printer to operate under microgravity.
Stephen LaMarca:
Thank God, it's Airbus and not Boeing.
Benjamin Moses:
I mean, fire in space is scary. Let's be honest. A hot laser in space, I mean, the risk and thought processes to get to this point is fascinating to me.
Ramia Lloyd:
Yeah.
Benjamin Moses:
That's cool.
Ramia Lloyd:
But now they can repair satellite parts and stuff, all that stuff.
Stephen LaMarca:
It's really inspiring. It makes your imagination race wondering what's next.
Elissa Davis:
It took less than 10 years to get there.
Stephen LaMarca:
When are we going to 3D print stuff on earth from a satellite?
Benjamin Moses:
Other way around?
Stephen LaMarca:
No, I mean, the material, the energy sources in a satellite, and it's making something on the ground. A hundred years out? I'm going to freeze myself. I want to see it.
Elissa Davis:
I feel like we're not that far from it.
Stephen LaMarca:
Really?
Benjamin Moses:
I'm more interested in you freezing yourself.
Stephen LaMarca:
I feel like we're not that far from it, but we're really far from it in terms of government oversight and bureaucracy.
Benjamin Moses:
Steve, tell me about AI and LLMs.
Stephen LaMarca:
Okay, so last week, David Anderson reached out to me on LinkedIn being like, "Hey, saw your ChatGPT article about using ChatGPT in manufacturing. I wanted to reach out to have a nerdy conversation with you about LLMs and AIs, but with a focus on manufacturing." And David Anderson, he works for a... I think the company's called Engora, or that's just the name of his team's LLM. But he's a resident at Autodesk on the East Coast, so up in Boston -- visited there during the pandemic, season one, road trip with Steve, just check it out. And he's there working on this LLM like ChatGPT, but specifically for manufacturing. So here's some of the bullet points from our conversation.
Engora is an AI assistant for manufacturing and engineering, of which there's an enormous amount of documents, books, just information and data sets in general, that instead of being ChatGPT or Gemini, which has the access to a blocked off timeframe of the internet, Angora has just been feeding their chatbot, their LLM... What is the name? Where is it here? Their Chatalog with everything that they can get their hands on with respect to manufacturing. So every edition of the machinist's handbook is thrown into this LLM, so you can ask it all of the things.
One of the demonstrations that he did for me was he asked Engora, "Make me a table of all of the specs of the different types and sizes of socket head screws, socket head bolts." And it made a table, and then I was very impressed by that. And I was also like, "This seems a little something that ChatGPT could do." And then he threw the curveball at me. It was like, "All right, now make me this table in JSON," and it did that. And it is cool because they don't have the... Well, his team, which again, is just a residency at Autodesk right now, doesn't have the computational power that OpenAI has, or Google, or Microsoft has. So it took a little bit of time to think.
And one of their next steps is, "Oh yeah, we're also working on creating an error. So sometimes the thing will just start thinking, and it will keep thinking, and it will never tell you that it ran into an error. It'll just keep thinking, it'll just leave you twiddling your thumbs. We're working on building that."
Benjamin Moses:
So the domain-specific knowledge set that that has-
Stephen LaMarca:
You told me about this word.
Benjamin Moses:
I think it's very important because it builds confidence. And knowing that if they use a machine's handbook, you can ask, "What's a clearance hole for quarter-20 bolt?" And then, obviously, it'll give you a starting point and you can check the machine's handbook to verify. You should always trust, but verify. But the fact that being able to convey domain-specific prompts to the tool and knowing that the training set is from that training set, I think, is very important. And especially for manufacturing and the domain-specific vocabulary that's required.
Stephen LaMarca:
Yeah. I love the idea of being able to go to an LLM, a chat, and be like, "Okay, I'm using a quarter-inch, 0.030 corner radius, three-flute end mill, and I'm trying to machine some brass to make a watch dial. What speeds and feeds should I be using?" And then it spits out the speeds and feeds. It was like, "Oh, sorry, I need to specify this isn't machining brass. I'm doing real standard brass." And it changes up the speeds and feeds for you.
But more importantly, then I forgot to talk to him about this. It would also be really cool to see an LLM that can also cite its sources. It can tell you where it found it. Because I remember when I started college, when I started my undergrad, I actually went to... Part of orientation was some teachers pulling us into a massive classroom, a lecture hall, and basically yelling at us about, "Hey, don't use Wikipedia. Wikipedia is not a legit source, so don't be writing your essays up in here and using Wikipedia." But Wikipedia has sources. If I use Wikipedia, obviously I'm not going to cite Wikipedia, I'm going to cite Wikipedia's sources because those are legit. Wikipedia might not be. And yeah, "Somebody can influence bias on it." It's like, "Yeah, if you're an idiot, if you're dumb." So I'd really like to see an LLM that can cite sources.
Elissa Davis:
I just wanted to say that as someone who has tried to edit our IMTS and AMT Wikipedia pages, they're real strict about it. I tried to copy stuff from our website, and they were like, "You can't do that." And I'm like, "I work for the company. It is not like I'm a random person using this information. I represent the company, and I'm putting it on there. And I'm quoting it, citing it, all that stuff." No, they took it all down.
Stephen LaMarca:
"I'm LinkedIn verified."
Elissa Davis:
Yeah. No, they took it all down. So Wikipedia, because there are people who literally spend their lives, they're just editing Wikipedia.
Benjamin Moses:
Moderators are real.
Stephen LaMarca:
And frankly, I'm glad they are.
Elissa Davis:
Yes.
Stephen LaMarca:
I'm very thankful that they exist.
Elissa Davis:
But that's part of why Wikipedia shouldn't be this big taboo thing because my professor in college actually said, she was like... Shout out to Dr. Lynnea Chapman King, one of my favorite professors. But she said, she was like, "Look, don't cite Wikipedia. But use Wikipedia as a starting point, and then go through the sources that Wikipedia has, and use those sources because those are probably going to be primary sources."
Stephen LaMarca:
That's all good professors have to say. That's the difference between a professor that's just trying to get a paycheck, which is like, grow up. That's a bad way to get a paycheck. Or somebody that genuinely wants to teach and make education better.
Elissa Davis:
She's a great professor. Yeah.
Benjamin Moses:
So you do hit a really interesting point back to AI, that is still a problem on the largest set of applications of artificial intelligence. "Show me your work." Back to the math teacher asking, "How did you get your answer?" Everybody wants to know that. And I feel like in the next couple of years, we'll get to that point because there's still lack of trust, in general. The only way you're going to build trust is have the model start showing you. Or to your point, Steve, if there's error, like if the LLM or AI tool can't find the answer for you and it kind of guesses, "Tell me you're guessing." And that's where I hope to see more iterations, specifically with the domain-specific stuff.
Stephen LaMarca:
It'd be great to have David at Anderson on this podcast because he's very down to earth, and these are all points that would resonate with him a lot. He wanted me to also mention that they will be releasing a public beta soon that anybody can use with the caveat, not really caveat, but with the disclaimer that's like, "Listen, you shouldn't use this to program anything. Consider all of the results that you get as a guest. Do your due diligence to make sure what it's telling you is correct. And please, for the love of God, do not use this to program anything. Don't ask it to make lines of G-code for you. It can, don't do it though. Or if you have it, do it. God forbid, just don't throw that line of G-code into a machine without simulating it first."
Benjamin Moses:
Yeah. Awesome, guys. Where can they find more info about us?
Ramia Lloyd:
amtonline.org/resources. Like, share, and subscribe.
Stephen LaMarca:
Bing bong. Yeah.
