Crafting the Perfect Brew

On this month's episode, in our installment of the Data Pulse, we chat about a candidate for mayor from Wyoming who's suggesting a chat bot could run the government.

https://www.foxbusiness.com/politics/wyoming-mayoral-candidate-promises-let-openai-bot-govern-capital-city-faces-backlash-from-tech-giant

We then go on to chat with Ron Hockersmith of Amorphic Beer about the unique ways his company uses scientific data in the brewing process. If you're in the Milwaukee area, be sure to check them out!

https://www.amorphicbeer.com/

Thank you to Continuus Technologies for use of their space and equipment!

Transcript

Colleen Hayes: 0:00

Welcome to the Cream City Calculations Podcast. We're three colleagues and friends that love data and to talk about how data is impacting our lives. I'm Colleen.

Frankie Chalupsky: 0:10

I'm Frankie. And

Sal Fadel: 0:11

I'm Sal.

Frankie Chalupsky: 0:11

Welcome to the Data Pulse, your quick hit source for this month's most impactful data news. Today, we are going to be talking about a very interesting data topic. In Wyoming, there is a mayoral candidate who promises to let OpenAI govern the city of Cheyenne, Wyoming, which is Wyoming's capital. And it's a city of about like 65, 000 people. And it's interesting because he wanted to have this AI bot run for mayor, but then the city wouldn't allow him to do that because technically the person running for mayor has to be a registered voter. So he is changing his name, he used to be called Victor and the bot was called Vic. And now he is being called Vic and he's calling the bot Victor so that he can put the name Victor on the ballot.

Sal Fadel: 1:01

Yeah, I kind of love that, how he is skirting the system a little bit here, but it's also very concerning that he's going to 100 percent go and use a an AI bot, which is built on OpenAI by the way.

Colleen Hayes: 1:15

Yeah, that seems a little problematic. I just want to go on the record as saying VIC stands for Virtual Integrated Citizen. So that's what he's named this bot. He goes on to say that he feels like this is the best way to make non biased decisions. And I guess if there's anything we've learned about AI at all is that it doesn't usually stay unbiased for very long.

Sal Fadel: 1:36

Yeah, and he said he's helped train it or at least categorize it in a certain way. So automatically adds his bias, right? Like it's hard not to add bias into that. But I do love that he was getting the media to answer questions from it. People were actually engaging with it. Yeah. Yeah. Which it didn't come back with terrible answers in some of these situations, but I really think that it's funny that people are actually buying into that.

Frankie Chalupsky: 2:08

Yeah, I think I would not vote for AI to be running my city. What about you guys?

Colleen Hayes: 2:15

Yeah, I'm with you on that. I feel like it's a good tool but there's still got to be some human person monitoring the outputs for it, and I really wonder how much Vic or Victor will be doing that.

Frankie Chalupsky: 2:29

Yeah, and I thought about that too there's some really large companies that are using AI for very powerful things, but none of them are just letting AI run the company. And I feel like there's good reason for that maybe they have figured that out already. I think that could be very interesting for the city.

Sal Fadel: 2:45

So it's really about transparency here, though, if someone comes and says, hey, we're going to let chatGPT or. system run the government and full transparency versus some politician that is actually going on and using all this writing his scripts through it and using it to create policy and understand policy or band, which is better. Like one, the both of them are using AI. One is just. Open to the public, right?

Colleen Hayes: 3:14

I think the way he's doing it, like you said, at least he's being transparent about it, and at least he seems to, or this article implies, that he would be letting asking the question of the chatbot to make the decision versus if a politician trying to write maybe a press conference or a news release may already have the decision in mind that they feed into the chatbot to get their response back. I don't know which one of those would be better, but also didn't this article say like this is the open source, like this is the web facing chat GPT. So it's not going to just have his bias in it. It's going to have everyone else's. I guess I would have less of a problem with it if it was like a closed system where they fed in other political decisions that were made by that office over an extended period of time or something but if you're like going to the internet basically and like chat gpt is just one step above a google search So if you're asking google like what to do in some of these situations, I don't think you're too many steps away from getting really racist or bigoted answers to be honest

Frankie Chalupsky: 4:21

like feeding it private documents, like, how can it make decisions on something that shouldn't be shared? And then it's just reading it in. Yeah, I don't think that could,

Sal Fadel: 4:32

I would love to know how he reacts to a difficult response or decision that this makes, right? Let's cut. spending to working or any social program, right? By a ton. And if he doesn't agree with it, will he hold to what it said or will he choose his own path?

Colleen Hayes: 4:52

What if the solution is really controversial and like the question is how do we save on taxes? And the response is let's kill all the old people or something crazy. Oh my gosh. I don't think we're too many steps away from a chat GPT giving a response.

Sal Fadel: 5:08

I will say Colleen, a lot of bad ones went through my head, but I didn't say it, but I'm glad that you did.

Colleen Hayes: 5:13

Yep, yep, I'll throw that out. I'll be the one to do that. Oh my gosh.

Sal Fadel: 5:18

I also love, one of the parts in this article is, ChatGPT, because of their policy of you can't do political campaigning has canceled his subscription, or whatever you want to call it his membership Already and then he did it with a new name and a new I. D. and then called it Vic 2.0, and is doing that. And it's really like it talk. The article talks about. Not being able to actually govern people from doing certain things in this. And how hard it is if this person is just easily building an app or building a chat GPT kind of prompter or AI system how quickly you can skirt the system.

Colleen Hayes: 5:58

Yeah. He seems like he's definitely going around the terms and conditions there. Yeah.

Sal Fadel: 6:07

This is our sixth episode and we actually have another beer one. So we're really excited. We brought in Amorphic Brewing, Ron to show off some of his beers. But he has some unique kind of ways that he's creating this using I would say machine learning or AI in that aspect.

Ron Hockersmith: 6:22

We'll talk about how true that is.

Sal Fadel: 6:23

True. Okay. That works, but yeah, just tell a little bit about yourself. Tell a little bit about Amorphic.

Ron Hockersmith: 6:29

Sure. So I think those of us that founded Amorphic come from a different background and than is common. We're all engineers, one material scientist two mechanical engineers. We all met at GE Healthcare. Before that, I did things like design gas turbine engines for F 22, Joint Strike Fighter, 777, when I was growing up in Arizona. Then worked at GE Healthcare for a long time. That's where I think I became a lean Six Sigma black belt and then the numbers thing and the data started happening. And then after that, I actually ran the engineer organization at an international defence contractor called Leonardo that makes electric propulsion systems for nuclear submarines.

Colleen Hayes: 7:07

Wow.

Frankie Chalupsky: 7:07

I feel really, yeah, I feel really blown away right now,

Ron Hockersmith: 7:11

but then after all of that, I don't know if bored is the right word or I was tired of the Corporate world. Yeah, and just didn't want to have a boss anymore. Even yeah I mean I was at the point where my boss was never even in the same state of the country as me often, but I Just wanted to do my own thing so I entered a semi retired phase of life and so did my main business partner and We decided to make beer. It's something we had joked about 20 years earlier or 14 years earlier And we gave it a whirl and we essentially do things. There's another business owner in the Milwaukee area that owns a, quite a few actually drinking establishments. And he gave me some advice. He just happened to walk in during construction of the brewery. And he said, the number one thing I can tell you is don't do it like everyone else is doing it. If you think it should be done in a different way, because you learned that and you worked in a different environment. It's probably going to work here. So like when we hire people in an interview, they're blown away to like, why are you asking me these questions? It's this is the same way it would happen if you were at GE or international defense contractor. So that's how it works.

Sal Fadel: 8:13

So describing that a little bit further, like from setting up your plant, initial build understanding where everything is flowing. Did your engineering background help you design that and set it up in a completely different way? Or is it still in the same basic?

Ron Hockersmith: 8:27

I mean in the end every brewery looks mostly the same. We do have a couple of unique pieces of equipment One of which we designed and we thought we would be the first people like to have one in the world Except when I found someone to make it and sent them the drawings. They're like, oh Yeah, we even have a name for this now. You're the 71st person to ask us to make this My hopes of a patent for that went south fast But I haven't seen one in any other, Milwaukee brewery or any other brewery I've ever been in, so there's 71 somewhere in the world, probably more now, but

Sal Fadel: 8:57

Can you describe a little bit of what it is? Are you, it's proprietary ish?

Ron Hockersmith: 9:01

You'll have to come on a brewery tour to learn about that. It lets us make our IPAs better. Yeah.

Colleen Hayes: 9:06

Okay.

Sal Fadel: 9:06

That's awesome.

Frankie Chalupsky: 9:07

Did you just want to create beer first, or when you were like coming up with this idea, did you guys want to develop a brewery? Did you know? Which one

Ron Hockersmith: 9:16

when I first moved to Milwaukee? I love Milwaukee. I'm not from here. I'm from the Phoenix area. I moved here about 20 years ago, so I've lived most of my adult life here. And 20 years ago, I'm going to be honest, beer in Milwaukee was pretty bad.

Frankie Chalupsky: 9:31

Throwing shots right now.

Ron Hockersmith: 9:33

I came from Pasadena, California.

Sal Fadel: 9:36

You mean Schlitz is not good?

Ron Hockersmith: 9:38

I mean there was things like Sprecher Black Bavarian was pretty good. There wasn't even an IPA made in the entire Milwaukee area at the time. And I was coming from the west coast drinking things like Stone and Russian River. And That just wasn't stuff like that here. So I basically set up a pretty complex brewery in my basement for the first 10 or 15 years that I lived here. And that's where the passion came. And when someone asked why I did it, it was because I can't buy the beer that I want to drink here. So I make it myself. About 10 years ago, that changed a lot. There's been so many that have popped up and now there's plenty of good beer made in Milwaukee.

Sal Fadel: 10:12

That's really awesome. So overall, like I know it's a data show, but I obviously, We like beer as well. I feel like both of those are go hand in hand, beer and data. But yeah, like I would love to know a little bit of how data just drove maybe your decisions into going to Milwaukee market, the types of beers that you chose and overall, like how you distribute really.

Ron Hockersmith: 10:35

I'll give you a couple of examples. So this is more like business oriented data, and a lot of the data science and analysis that we do is. When people talk about data a lot these days, it's a lot of marketing focus AI. Yeah, or maybe that term is basically just been, you should just put AI created nothing but, but AI ours is more technical data. It's not like we do have data on certain customers and things like that, and that data certainly exists, but I would say ours is more like how the acidity or the pH or the sugar content changes or certain gases are created. Over time, there's hundreds of parameters like that during the beer production. And I think most breweries at least write some of that stuff down on a piece of paper. Ours, of course, all goes into a database and we're able to search and filter and graph and trend. And that's just the Six Sigma side of us. We treat the beer no differently than if we were making widgets in a factory essentially. But to start, you asked about why did we pick in Milwaukee? It's going to get a little nerdy, but I guess that's the point of this thing. So we're trying to decide, is it, should we do this? What kind of, how much beer could we expect to sell? It turns out how much beer every brewery in the state makes in both kegs and in cans is public data as part of tax records. It's on the Wisconsin Department of Revenue's website. It's not hard to find but it also only took about an hour to write a Python script to scrape every one they've ever had for the past 10 years. So now we have a pivot table and I can look at, and it updates every month and I can look what every brewery's production is. Just yesterday I was talking to another brewery in the area who we're good friends with and I mentioned, Hey, I think you filed your tax report wrong last month. He's how would you know that? I'm like your curve went all over the place. It's impossible to physically make that much beer in a month at your facility. So I was like, I'll look into that. So they might be firing their new accountant or something. So we looked at that, that data. I wouldn't call that big data. There's a couple hundred breweries in the database and there's One data point or two data points per month. But that gave us confidence. Like we want to pattern ourselves after a particular brewery in Madison, maybe mixed with a little bit of this brewery with 25 percent of this brewery, and we could see what could we expect to sell if we

Sal Fadel: 12:43

did you do that analytically or more kind of conceptually in your head I like these things. These parts of these breweries, or did you actually go in and say these have these elements and all that. Cause you said you map some of those features out.

Ron Hockersmith: 12:56

In this case, if it's nothing more than how much beer they sold. And it was more, I would say once we had, we were basically just doing visual analysis and looking at the graphs and then we were able to tell Hey, let's not do it like this brewery. Let's do it like this brewery. And I was like, yeah, but a little mix of that. But I guess we could have made like A figure of merit might be the right term where we take 25 percent of that and 50 percent of that and let's do the distribution Like this. That would only take one line of yeah script or one cell in a google sheet But we didn't even need to do that because it was obvious from the graph.

Sal Fadel: 13:23

That's awesome

Colleen Hayes: 13:24

So it was more about the data behind the brewery that you chose to emulate and less like the vibe of the brewery Because you liked it.

Ron Hockersmith: 13:31

There are other examples I would say the closest vibe that we did try and emulate is cerebral in denver You economically, in the business model, we mostly emulated working draft in Madison. And then there's another brewery called Cloudburst in Seattle that does some things that I really liked as well, in terms of the aesthetic of how the taproom works. Looks and just looks like a converted garage Also kept the cost low. Yeah

Colleen Hayes: 13:56

can you tell us more about your brewery and what the aesthetic is it's called amorphic, right?

Ron Hockersmith: 14:00

Yeah, which is a play on the word amorphous. It is actually a word in Merriam Webster. It's just a For a form that we thought sounded more scientific with the ick on the end. And my partner said that amorphous sounded gross. Like the word moist,

Colleen Hayes: 14:15

what was Lawrence Fishburne's name in the matrix? Was it Morpheus? Morpheus. That's what I think of. That's where my brain goes when you say amorphous.

Ron Hockersmith: 14:22

So we have a space in river west. It's on the north end of Fratney street, just north of Keefe. We have a couple of brewery neighbors. There's one a block to the north and there's one four or five blocks to the south We've got a 2500 square foot tap room fits about 100 people we have Patio on the front we have what some articles say is one of the most unique secret patios in the back called beer alley. It's just an alley with covered in vines Sandwiched between two old cream city brick buildings. It's only about eight feet wide And then we also have a event space. It's like filled with old industrial antiques and gears and things like things like that. Even our tap handles in the brewery are just a hundred year old tools that we've found while traveling.

Colleen Hayes: 15:06

While you're cleaning out the building?

Ron Hockersmith: 15:08

It might be one of those, but mostly it's like an antique shops and things like like that. People say that it's a mix of They say, these aren't my words, they say like nerd and industrial, rustic industrial is what they say.

Colleen Hayes: 15:18

Rustic industrial nerd?

Ron Hockersmith: 15:19

Yeah, I don't know what that is.

Colleen Hayes: 15:20

That's the aesthetic you're going for? That sounds right.

Ron Hockersmith: 15:22

That's what it ended up being. I think the first article that Bobby Tenzinillo wrote in On Milwaukee about us before we even opened, he like, it had the word nerd in the title. And we just, cause your background, we just got on a band. I think there was even something in there like these two nerds are starting a brewery. Thank you.

Sal Fadel: 15:42

Nerd is cool now. So it's okay.

Frankie Chalupsky: 15:45

It is a cool phrase now. So I would take great pride in that. All right, you have this great space and it sounds like you guys are doing really well. We want to know a little bit more about how you developed your beer recipe.

Ron Hockersmith: 15:57

Yeah. So another thing that's unique about us is we don't really make the same beer twice. So it's sometimes annoying that I spend a lot of time developing new beer recipes. And if we taste some of the ones that we have in front of us, you'll see they're all very unique in a different way. I didn't really bring any normal things from that perspective. But because we're making new beers so much some breweries, I've worked at other breweries in between more starting at Morphic and retiring from the corporate world. Some breweries are more art based, right? Just let's see what happens if we throw this thing in and let's see what happens if we don't put this thing in. Or maybe if we put something in at a different temperature or pressure, but that might be as far as it goes, but we have models that we've been developing that predict almost every aspect of the beers, even its aroma and flavor with mass before we've even Purchased the grain, so

Colleen Hayes: 16:51

I'm sure that makes a difference to where you buy the grain from.

Ron Hockersmith: 16:54

I strongly believe that it's true Maybe it's growing up in a on the west coast and depending on the time in california like california cuisine It's really nothing more than just use the best freaking ingredients you can possibly get. What do you know? It just tastes better Yeah, so I believe that's also true in beer I can tell you i've recently spoken to a couple of brewers who told me quote It doesn't matter where you buy your grain. I buy whatever's cheapest. It all tastes the same. Then, you know I will not name names from

Sal Fadel: 17:20

this is me on a side tangent from a a hops perspective Do you buy pellets or do you for consistency or

Ron Hockersmith: 17:26

almost all commercial breweries use pellets because the commercial equipment is made Specifically to handle pellets. Okay using whole cone hops is difficult without special methods There are some like sierra nevada is almost exclusively so before we get too excited about these models I would say that they are at best, like

Sal Fadel: 17:45

they're gradient boost models.

Ron Hockersmith: 17:48

They, they do correlate and they show general directionality, but when more depends, there's, we're still subject to the whims of mother nature and agriculture and the specific terroir of a certain lot of hops or something. And so even if. Even if I get a certification on this lot of hops, which I'm amazed at how many vendors are surprised that I'm asking for that and it shows the oil breakdown and All of those hops and in that particular a lot of hops like for example, there's an oil called farnesine Tastes like oranges It's these same oils are actually in cannabis as well. It's hops and cannabis are related minus the psychoactive yeah effects So if I'm trying to boost orange and pineapple, I can choose hops and lots that do that. And if I'm wanting to get strawberry and lemon, there's hops that have those components as well. And that's how we use those to predict, okay, this one should have a strawberry, vanilla, orange flavor to it. But sometimes the model says that, and then I'm tasting out of the tank and I'm like, okay, so this one's definitely not on that curve, the curve that the predictive curve, I'm like way off the curve. And. Something is different. There's so many factors that it could be that it's hard. So we improve the models and the predictions occasionally, but in the end it's because you're still guessing

Sal Fadel: 19:05

again, nerding out a little bit because your sample size is low. You said you only make like one. One type of beer or one like batch, right? Or you said, or a certain quantity, right?

Ron Hockersmith: 19:14

Nevermind, but there, so there are noise variables, but there's also like control variables that I know I did it different on beer one versus two, versus three, versus four. And in theory, there are established closed form solutions or how that does affect that. And so I can adjust for those factors and we do things like that. But all the things that I'll call noise variables, like you can't really, how hot it and humid it was in. on the western, on the eastern side of Lake Michigan and at this particular farm that year. I can't really see that in the data and it appears to affect the hops. It'd be on the finished beer beyond what. They currently provide in the certificate of analysis. There are people doing research trying to add like hundreds of more parameters. I don't think anyone will ever look at it if they do.

Frankie Chalupsky: 19:57

Do you think those people that are adding in like hundreds of parameters, do you think they're going too far? Is it over complicating the algorithm?

Ron Hockersmith: 20:04

I think what Flavor science, just from what I've read, is It's pretty immature. I would say some of these things, the human palate can detect things that are like on parts per trillion in some cases, and we don't have measurement devices that can. They might have a gauge on our measurement capability. That is. The whole size of what you're trying to detect. And so you can't even really tell if that compound is there. But when someone tastes it, they can. So we don't have measurements for all, say, 100 of these flavor compounds yet that I think are reliable. That's why I was giving you the hint that. I think they might be going too far too fast.

Colleen Hayes: 20:39

So can I just jump in and ask do you think we can open one of these and try one while we're talking?

Ron Hockersmith: 20:43

Yeah, we don't run out of time. I'm,

Colleen Hayes: 20:45

sorry We're all like looking at these beers sitting on the table and kevin our colleague is here in the room Just listening kevin is friends with ron and kevin if you want to come over and get one also Come on,

Sal Fadel: 20:55

kevin you're working

Colleen Hayes: 20:58

Can you tell us about what this beer is that you're pouring for us right now what it's called

Ron Hockersmith: 21:02

This is called They're Here, Aren't They?

Colleen Hayes: 21:04

They're Here, Okay, They're Here, Aren't They?

Ron Hockersmith: 21:06

It's an X Files reference. This is a, most of our lagers are Czech style in nature just because they have more flavor, I think, than German style lagers. This, it's just, it's an imperial pilsner, so it tastes like a pilsner urquell, except It's actually 17 percent wine. So we wanted to, a lot of times when you make something higher alcohol, it gets really sweet. And when it's really sweet it can, it wasn't the effect we wanted. So you can add sugar to make it drier, but we actually added Sauvignon Blanc grape must to this. So it's technically 17 percent wine and then aged in French oak for a little bit of time. But in the end, when you drink it. It's pretty simple and it tastes like a

Colleen Hayes: 21:47

I was gonna say it does taste like a little buttery. It's yeah it does taste like a Czech pilsner, but I get the kind of wine reference from it, too totally not that i'm like a wine snob or anything, but

Sal Fadel: 21:59

So that's good from designing beers and understanding it. First of all, my brother's a food scientist and he would He, I know he listens to this and he's going to nerd out hard because this is what he does. He loves doing it. And he went to Madison, which they had a beer science class. But like from this, like what things have you used to shape your beers in different areas? Like it, do you find that okay, it seems like the population likes IPAs, this is how I want to match these flavors and then picking out flavors based on that. Yeah.

Ron Hockersmith: 22:30

So there we do some of that. We look at like sales velocity, which is just how much you sell and what time period, of course untapped ratings, which I despise the untapped, but it is really, they bought their biggest competitor and ruined it. So they're really the only option. It is a very poor data set because number one, people are drinking. Just give you an example. The biggest correlation that we have found and others have found with untapped data. Is the how much alcohol is in the beer? There's a 73 percent r squared adjusted for The abv of a beer versus its rating way off.

Colleen Hayes: 23:01

That's crazy

Ron Hockersmith: 23:02

So if you make if you want to have a brewery with really high untapped ratings make everything 10 to 14 percent And you will probably have one of the highest rated breweries. Wow There are other factors in there that i'm, been working on and trying to prove if you offer flights at your brewery Yeah basically then people are encouraged to get things. They wouldn't normally order. Yeah, and then wouldn't like You So I believe that actually lowers your untapped rating by 0. 15. If you, the age of your brewery also affects your untapped rating. I've actually submitted a proposal to prove this with math and data scraped from untapped at the craft brewing conference in 2024 in April. And I was rejected. They gave me feedback on why I was rejected. And this seminar was rejected and their reason was that it's impossible to do what I said.

Sal Fadel: 23:50

What?

Colleen Hayes: 23:50

Okay, I'm sorry. I find that incredibly fascinating.

Ron Hockersmith: 23:55

So they don't, they must not listen to your podcast.

Sal Fadel: 23:58

Bear in mind, this is data science in general. Yeah. It's hard to explain it to people. Yes.

Colleen Hayes: 24:04

But that's crazy that it, that your score overall will go down by.15 points if you offer flights. But do you think that affects your sales? Do you think your sales go up of some of those other kinds that people might not otherwise try? And is it come out in the wash at the end?

Ron Hockersmith: 24:19

I don't think so. There, there are several kinds of breweries in the marketplace and for some let's just call them beer nerds. They are basically using untapped to decide where to go and what to buy. There are definitely, there's less and less of those because untapped I think is probably struggling. I don't have their financial numbers. I just know that less and less people are using it which makes it harder to use it for data. I think that those people do exist. If you're just a neighborhood corner brewery that mostly people from the neighborhood go to, then I think it doesn't matter as much for us. We generally are a destination brewery that people hear about around the Midwest and when they're driving through they stop in. So I think it actually does matter for a brewery like us.

Sal Fadel: 25:00

So not in just beer making, do you use it in just your advertisement? And kind of social media and all that as well? Do you use data to drive business as well?

Ron Hockersmith: 25:12

Would say not as strongly as I would bet some of Some other types of business, businesses do. But I think I can answer, give better examples of using data to finish the question you asked before. So yes, we use untapped data, we use how fast things sell then we use sensory panels, which is a fancy way of saying me and some other people drink and smell the beer and make some spider diagrams of, woodiness orange, and that's data. And then we take all of those and specifically for our IPAs, that's the one where we did the most work and developed our own little algorithm That we market as sentient IPA, and we released 1. 0 of that prior to ChatGPT becoming a thing. Yeah. It was a thing, but it wasn't like publicly available. We got quite a bit of like a local buzz about that and people were really excited to talk about it. Then about six months later, we released version 2. 0. And Chat GPT was a thing and everyone assumed that all that we did was go into chat GPT and say, tell me how to make an IPA. And then it parrots, then it stochastically parroted back the average of the internet on IPAs and that's what we did. So now we specifically market it as, this is not made with any LLM chat bot. That would not help you one bit with this. And it's the branding of it has become like making fun of, AI that is not super useful. And now we, so now it's a different type of people that are entertained by the way we've been branding that one.

Sal Fadel: 26:39

It's really fun to watch listen to the counter of everybody's Oh AI, or chatbots or whatever. It's actually like completely counter to that. I love that.

Colleen Hayes: 26:48

Can you talk about that? You were talking about your sentient IPA. I heard that you recently, your brewery won a, an award for that beer.

Ron Hockersmith: 26:55

Yes we won the People's Choice Awards. So there's Wisconsin IPA Fest put on by Third Space every year and this year 44 different breweries participated. And then there's certified judges do one, they judge the 44 IPAs and then someone wins that. And then there's the People's Choice which everyone, like the 600 people that come to the festival, all vote. We've won the one where the people vote two years in a row. We have, I've been trying to win both in the same year. But I struggle designing a beer because to please judges is very different than pleasing the people. So I guess I'd prefer to please the people versus please the judges. That way, but on this sentient beer, we didn't just take those three things that we just talked about with the untapped ratings, the sales velocity and the sensory analysis. Let's call those like our outputs are. Sorry, I'm going to speak a little like Six Sigma-y, which is adjacent. So those are outputs. And then there's there's 200 different inputs that are from all the data that we take during the production of the beer and as it ferments. And we just did the same things that like we did at GE back 15 years ago, which to be honest, used neural networks and space filling designs and. Like that except then it wasn't cool and it wasn't you couldn't chat with it. Yeah, it's doing the same effective thing, but instead of Three billion pathways based on every internet page that there is it was based on 10 000 Designs that we ran through a finite element simulation and then fed into it So we just applied the same sort of thing here And in the end, SentientIPA is, I would definitely not call it AI. Yeah. But I also don't call, I would say 99 percent of the things today marketed as AI. Yeah. I'm not even sure. I would consider a gen, a pre trained transformer, it can only learn once, right? Yeah. It will never learn again. Yeah. That I think is on the edge. Ours is not even that. It's certainly pre trained. It was not generative. I had plans to. do some machine learning and neural network things ran out of time and all in the end we really did was we took those 200 input parameters and these like nine output parameters and we essentially just did a multivariate non linear regression with like cross terms and some higher order terms you run into like overfitting problems with the data right so we started just throwing out terms that had low coefficient of influences and yeah eventually we were down to seems to be there's 12 things that matter And the square of a couple of them and a couple of cross terms and things like that. And then we started noticing that it did match. So then I took that model that we had developed, which was just like, I said, it's least squares, it's least squares fit multivariate regression, not, which I guess is like college algebra. In the end, but it's probably more college algebra than any other brewery has done. And then I'm like now I can predict the untapped score of any beer before we make it. So technically now I predict, oh, this IPA is going to be a 4. 16. In the air you close. I'd say we're like, it's not as good as it's plus or minus 0. 15. Oh, okay. But, so yeah it's pretty good. No, because I would say a bad IPA is like a 3.8. Okay. And a good IPA is a 4.20. So I would say, so there's not that much variance, the difference between bad and good.'cause humans don't like to use the whole scale of one to five. Yeah.

Colleen Hayes: 30:18

Yeah, you're right. I've looked at my untapped. I think I've, I don't think I've given any beers a five. No, because that's definitely,

Ron Hockersmith: 30:23

and you've probably given not a one either, right?

Colleen Hayes: 30:24

No, probably not. No.

Ron Hockersmith: 30:26

Maybe you should have more higher alcohol ones and then goes naturally

Colleen Hayes: 30:28

then, I just have a lot of fives.

Sal Fadel: 30:30

So that like you now can predict the year on top on tap score

Ron Hockersmith: 30:34

and how fast it's going to sell and how it's going to taste. And

Sal Fadel: 30:37

but like what you said earlier was like, you don't really believe in what they kind of show. And the numbers and the metric does that affect what beer you have or you're like, all right, loosely, like I still want to make this type of beer.

Ron Hockersmith: 30:50

So I always look at the data, but then because beer, whenever I give a brewery tour, I say that beer is a mix of art and science. Most breweries I would say maybe are 50 50. I always tell them we're like 95 percent science. I reserve that last 5 percent just to use my gut feel and say I smelled that lot of hops. I don't think it's actually going to taste like it says it's going to taste. We're not going to put that lot of hops in here. I'm going to use a different lot. I don't think that my fermentability is going to be like what it's predicting it's going to be based on that or my acidity or whatever. And I've actually been slowly developing like fudge factors effectively, but we've only made, we make about one and a half beers a week. We've been open for two and a half years. We've made like 170 beers. So I only, I have multiple data points from each beer, but really just 170 things to work

Sal Fadel: 31:37

How many, I guarantee you that there's other breweries that have not made anywhere near that amount of beers. So many beers.

Ron Hockersmith: 31:44

And it's a lot of work. It takes it probably takes me like four to six hours to design each one. Now, many of them are variants.

Colleen Hayes: 31:49

But this seems like a thing you really love, right?

Ron Hockersmith: 31:51

Yeah. To be honest Don't discourage anyone from visiting us, but if I could not have the tap room and just sit in the brewery and design things and then make them and see if my data that I predicted matches what it did there, I would love that. That's what I did my whole career. But now, instead of with airplanes and submarines and medical devices, it's with beer.

Colleen Hayes: 32:07

So it's not like you need a partner. You need somebody that works for you. That is like the king of the tour, the queen of the tour. And that's what they love to do is interface with the people.

Sal Fadel: 32:17

You're saying it's not like that?

Colleen Hayes: 32:18

No, I'm saying that's what you need. Cause it sounds like you're

Ron Hockersmith: 32:20

Occasionally when we have multiple people get sick or go on vacation, I've been forced to beer tend, not exactly my I enjoy it for a couple of hours.

Frankie Chalupsky: 32:33

Okay. So how much product do you waste or do you drink everything? No matter what, like when you're trying to develop a product

Ron Hockersmith: 32:40

we do have two breweries actually, and one is a, one is like 1 30th the size of our big one. So sometimes we will make beers there. And actually, a lot of those that are wasted, as you would say, just end up on tap in my basement. The museum of failed amorphic experiments.

Sal Fadel: 33:00

So when are you inviting us over? I want to see the algorithm.

Ron Hockersmith: 33:06

We made a pink stout using roasted beets and vanilla. And the first batch tasted like carbonated borscht. The Russian soup that went on tap at home. And I even had to dump it. But the cost to make a test batch like that is like a hundred dollars. The cost to make a full size batch ranges, but it could exceed 10, 000. Wow.

Colleen Hayes: 33:31

I think it's time to try another beer. Yeah. Sorry. I'm going to keep us on, on. Task with this.

Sal Fadel: 33:37

So as you're opening up, maybe tell a little bit about the beers. If you would mind.

Ron Hockersmith: 33:41

So since we just talked about sentient IPA, I actually don't have any sentient IPA because what do you know? You win the competition and then everyone just comes in sentient IPA. This is technically, I think the highest rated triple IPA in all of Wisconsin.

Colleen Hayes: 33:56

Thank you. And what is it called?

Ron Hockersmith: 33:59

It's called floofy chonks. We have weird names. It means like furry cat. Yeah. It's named after someone's cat and it's no, and it's in, it's cute. Cause it's in like a metal font. So it's cute and metal at the same time. I don't know what the term for that is.

Colleen Hayes: 34:15

Sorry, Sal, you got, yeah, it's 10 percent

Ron Hockersmith: 34:20

complex IPA. You got notes of there's strawberries and vanilla and wine All going on at the same time.

Colleen Hayes: 34:29

It's very smooth.

Frankie Chalupsky: 34:31

Yeah

Ron Hockersmith: 34:31

it Tastes more fruity than bitter. Yeah, and the vanilla adds a There's only a really small amount of vanilla in this, but it, you can probably tell it's there. And it really just makes everything so much more fruity. That is a factor I do not have any in any of the data. And I don't have add a, add a little tiny bit of vanilla at 0. 0002 percent vanilla to a beer. And all of a sudden the strawberries come out. That's like an interaction term. That I don't have represented in that particular case.

Sal Fadel: 34:58

I know like they do have a salt, but there are like I like cooking. So one of the big things is there's, it's called a flavor Bible and it's what complimentary things hit, pull out flavors and other things. So I thought, so what you're talking about is really interesting cause it's like, all right, how do I add a little bit of vanilla where you're not going to taste a ton of it, but it's going to pull out every flavor else. Like all the hops flavor that you were saying earlier.

Ron Hockersmith: 35:21

Yeah. You're saying salt is like that.

Sal Fadel: 35:23

Yeah. So it's yeah, sorry. Yeah. Yeah.

Ron Hockersmith: 35:25

Do potatoes taste like much until you add salt to them? Yeah

Colleen Hayes: 35:28

That's interesting, so do you have any new beer projects that you're working on at the moment anything you want to talk about or Plans like future things you want to try?

Ron Hockersmith: 35:37

I need to go on, I haven't gone on vacation in a while, but actually a lot of my inspiration comes from travel. Recently we released a tropical stout, a stout that is made to be drank when it's warm in the summer from a trip to Cambodia where they drink like really pale lagers and then 8 percent stouts that somehow taste good when it's 100 degrees and 90 percent humidity. We did that with co fermented coffee beans with banana and lychee fruit fermented with the coffee beans and then a coffee is in it. So when I travel, I just get inspired by these different things and sometimes it's that example, but other times I just go to a brewery that I haven't been to before and I'm like, holy cow. That is really good.

Sal Fadel: 36:16

And you can write them off every time you go to a brewery.

Ron Hockersmith: 36:20

There are some limitations. To be honest, when I first started this, I'm like, every vacation I take is now market research. I read some IRS guidelines and I'm not sure I can go that, that that far.

Colleen Hayes: 36:30

But maybe you can write off some of your food and drink. Yes. Expenses.

Ron Hockersmith: 36:34

I will not answer that question. Okay. Fair. Fair.

Sal Fadel: 36:37

Do you keep a notebook of everything? Like when you're Yeah. At a new brewery.

Ron Hockersmith: 36:41

I should, I do have a Google keep file where I note things that I think I'm detecting. And I'm even putting things in there that are like, we talked about some of the models earlier and I'm jotting down I'm going to guess this is say the mash temperature is 155. Wow. And I don't really know, but I'm guessing. And sometimes it's I see ingredients in there and I'm putting two and two together. Yeah. It's same way. Say, if you go to a restaurant and you have some dish that you really like. And you reverse engineer it. I'm just getting all the data that I possibly can from that experience. I'm mostly getting sensory analysis by drinking it and smelling it. And then I'm looking around and saying that they have this piece of equipment, so they can't be doing that. And they're doing, they have this, but they don't have that. I jot, I do jot occasional, notes down about that. And then I use that to influence what I do when I design the beer.

Colleen Hayes: 37:25

That's like fascinating to me. I was telling these guys when we were trying to set up an interview with you, Ron, that this is the like. lining up of all the nerdy things that I like. And I was like, so incredibly excited to get to talk to you today. And I probably have told every person that would listen for the last two weeks or whatever it's been. But let me ask you, do you have a favorite style of beer or favorite type of beer? What's your favorite?

Ron Hockersmith: 37:49

Sure. I always tell people normally like at a higher level, I always say I only drink three things and that's orange juice, water, and beer. No, four things, orange juice, water, beer, and tequila. Okay. Yeah. But then in terms of beer, if I, we make a wide variety of things like over fruited sours that are fluorescent pink and taste like a smoothie hard seltzers, but I personally don't drink those very often. I know I am noting you can mix those with an IPA 50 50 and they taste pretty good. Okay. But I would drink check Pilsner IPA. And rattanomyces saison. Rattanomyces is a wild yeast that's just in the air, which is actually what that one is.

Colleen Hayes: 38:30

The one in the bottle?

Ron Hockersmith: 38:31

Yes.

Colleen Hayes: 38:32

Oh, we should have started with that one. We don't have an opener. I was going to say do. Oh, perfect. All right.

Frankie Chalupsky: 38:38

Man carries an opener on him.

Sal Fadel: 38:39

To get back to the data set, how do you not add bias into the beers that you're choosing? Cause obviously you like certain types. Like, how do you not add that in when you're building it, or do you, because you're like, Eh, this is my beer brewery.

Ron Hockersmith: 38:52

I think I can understand why, like when you're building a model and now I'm thinking of examples back, and prior to beer, because I think my answer when it's beer is that I don't actually care if it has bias, and I might even want it to have bias. Yes. Which you anticipated, I think. Yeah. But before, when I'm, maybe I was dealing with Like finite element solutions of several million degrees of freedom in the late 90s in an airspace company and I'm pushing the limits of what the computers could do it at that time and it was already taking two weeks to solve or whatever and so I'm starting to chuck out some parameters in the optimization and sometimes I'm out of time and I promised the boss I'd have it done by some certain time and I'm like I could throw that out, but I think this one matters and They both have similar influence coefficients, so I can't use both. And then it's just a gut feel at that point. But maybe that's why you, your background and experience.

Sal Fadel: 39:43

Now I want to talk to you about all of your history. You seem really fascinating in general. But overall, this has been a absolute pleasure. This is. That never beats never is bad when you bring beer for us to drink during our podcast I do want to thank you and I think this has been absolutely wonderful.

Frankie Chalupsky: 40:02

Absolutely great. And are there any events coming up at your brewery? You want to showcase or?

Ron Hockersmith: 40:07

Yes good question So there's two I would say big events. One of them i'm really excited about so we started doing this thing called three minute thesis in the at our taproom We do it once a quarter It's basically, it's actually started by the University of Queensland Australia's international graduate student competition, where they have to present their thesis in three minutes to the general population instead of their advisors and their team members who of course are experts in the field and it helps them, communicate better. Yeah. I found this, I went to Caltech for grad school and during the pandemic I was you no longer had to fly to Caltech for like alumni stuff. You could just do it on zoom. And so they were participating in three minute thesis. And I got to watch people like Katie Bowman talk about imaging the first black hole and things like that as part of the competition and vote on it. So then we decided we want to do something like that, except I can't get Katie Bowman to show up. We just invite. Anyone can sign up and they get three minutes to present on anything they want to. It doesn't even have to be true. Some of them are not, one of my favorites last time was why Vanguard, the restaurant in Bayview, sausage bar deserves a Michelin star. There's been, there's a lot of food ones, there must be a lot of foodie crossover here. There's like how to make the best grilled cheese. I disagreed with that too. I'm not a fan of mayo on the grilled cheese. The greatest invention that Americans refuse to adopt, any guesses on what that might be?

Colleen Hayes: 41:29

The bidet?

Ron Hockersmith: 41:30

Yes. Oh, it's okay. That's good. Some Americans. Yeah anyway, so we do this once a quarter, and now we basically can't fit in our tap room. It's too popular. That's cool. So we are doing it at the, we're having three minute thesis all stars, and I've invited the 12 best presenters from the past two years of doing this every quarter. This includes professors neuropsychologists television personalities are all part of some of the best presenters that we've had. We're doing it at the Oriental Theater in the main hall. So we can, instead of fitting 130 and exploding into the parking lot at Amorphic, we'll be able to fit 550. And I'm hoping

Sal Fadel: 42:04

Can we get your winner on our podcast? Yeah, right?

Ron Hockersmith: 42:07

I'm sure you could.

Colleen Hayes: 42:08

Maybe we just show up to the event and just act really friendly. When is that event?

Ron Hockersmith: 42:12

That is on October 9th. You can buy. Since it's a larger event and we're working with Milwaukee Film we wanted to do it with tickets. So it is ticketed. The tickets are 10, but you get a free amorphic beer. So it's basically 2 or 3. There will be two bars set up inside the theater itself. With eight amorphic beers on tap. So we're trying to recreate the taproom experience In there, so that's the first one october 9th at the oriental theater buy tickets at mkefilm. org and then the next one is, it's october fest season almost I mean some breweries announced they released their october fest in june because they're trying to be the first one we're late. So maybe we're not going for first. I guess we're going for best then. Yeah Because we can't win first. Yeah We like I said, we don't really make German lagers. We make all Czech style lagers. I just from my travels the beer is better there Yeah, I know. I just offended lots of people but that's just what I believe so we're having Czechtober fest which we do every year and that's on september 28th We're right next to an urban farm right across the parking lot and they grow all their own vegetables and flowers and everything there and then They will be making some like side dishes to go and then we partner with Stefano's Slow Food Market in Sheboygan. They just, the butcher happens to live in River West by there and he makes these Czech style sausages that last year blew people away. So you can get some Czech style food, desserts, some Czech style beer all in an urban garden and there's some live music and things like that. That's September 28th.

Colleen Hayes: 43:37

That's very cool.

Ron Hockersmith: 43:38

It's basically the same as Oktoberfest, but slightly more flavorful. Yeah. Okay. That's basically how I would describe it.

Colleen Hayes: 43:42

That's how I feel about German food too. And I'm very German. Cool. Let me ask you real quick. I'm sorry, before we wrap up, is there anything we didn't ask you that you'd want to mention?

Ron Hockersmith: 43:51

I'm just a guy that has been playing with data my whole career and you guys like have a podcast and have it on your shirt and other things. I thought you might ask me questions and I would get,

Sal Fadel: 44:01

so that's part of it. We don't want to go. Too intrusive, a little bit, because it is your property and it's like your

Ron Hockersmith: 44:08

Yeah, like I probably wouldn't want to say, like for example, there are four parameters that turn out, if you want to have a high untapped rating, of course alcohol is one, and high cell fast, and good sensory results. And there's two of those that matter a lot, and there's two of those that I didn't know. Before I did this.

Sal Fadel: 44:23

Oh, really?

Ron Hockersmith: 44:24

And when I give brewery tours, sometimes I show that graph that I I blank out the axes and I don't tell'em what it is. I'm not gonna let anybody, like a shady news organization. Then they all just try and you're not gonna tell us the scale. Oh, is it this? And I'm like, you're on the right track. That's as far as I usually go, I'm like, this brewery should probably pay for this information. Yeah. Yeah. And imagine if I keep doing this in 10 years, maybe I should just write a book and teach everyone and then maybe people will just get.

Colleen Hayes: 44:45

Have your own podcast.

Frankie Chalupsky: 44:46

Yeah,

Colleen Hayes: 44:47

we'd be guests I don't know if we have anything that interesting to say actually

Ron Hockersmith: 44:52

If you drink a lot more during podcasts, then everything just becomes semi interesting, right? It just depends

Sal Fadel: 44:57

do you use clustering at all?

Ron Hockersmith: 44:59

No, so not in this case It's just there's not enough data. And yeah, like I said, we have 160 beers maybe 40 percent of our IPA is in the end, the models that we're using have maybe like 25 parameters that are going in back when I was designing things for x ray tubes. We did do that a little bit to to make that work, but to be honest, that was long enough that I hope you don't ask me any more questions about that. I remember using it and as you Katie researching it. Just kidding. All I could tell you is we're definitely not using it now. Yeah. But I used it 20 years ago in, I say

Colleen Hayes: 45:34

I don't remember things I did six months ago. So that's, yeah

Sal Fadel: 45:37

this has been really awesome.

Colleen Hayes: 45:38

Yeah, it's been great. Ron, thank you so much.

Sal Fadel: 45:40

Now I'm looking at all the Czech beers.

Colleen Hayes: 45:42

And I'm also thinking we need to organize like a group outing to amorphic and come check out some of, at least the beers, if not these events.'cause I feel like that would be really cool to participate in. For sure.

Sal Fadel: 45:54

Appreciate it. Thank you. Yeah.

Colleen Hayes: 45:56

Can we try this one in the recording just in case we want to add anything? Sometimes we end up talking about unfortunately great things.

Ron Hockersmith: 46:02

I would if you can yeah use a Clean one. Oh sure. Is this one the vanilla like actually adheres to the plastic?

Colleen Hayes: 46:09

Why don't I go get a couple more cups

Frankie Chalupsky: 46:11

And i'll take just a very small amount I don't usually drink at all. Yeah, you don't have to have, you don't have to finish it.

Ron Hockersmith: 46:18

People always at beer fest, they come up and they're like staring at the list of the six beers that we have at the beer fest. And they're like, I don't know what to do. It takes so long to read. Cambodian stout, co fermented coffee, lychee. What's that? Cucumber? Gin botanical kolsh with sea salt and lemon like what does it

Sal Fadel: 46:36

does that make it hard?

Ron Hockersmith: 46:37

Like people do get like they stress out Yeah at the tap room. I noticed whatever I this is data I guess at the tap room. I noticed whatever I put in the top left corner of the menu display sells the fastest generally Wow I'm like crunch the numbers to prove which is actually so it's only anecdotal so far, but I could prove it

Sal Fadel: 46:54

From a visualization perspective The eye moves, the first place an eye moves is the upper left hand corner. And like, when we're building dashboards or any visualizations, your most important or high level information should always be in your upper left hand corner. That I've never been taught that.

Frankie Chalupsky: 47:09

Which is why you put the logo there.

Sal Fadel: 47:12

They did it using eye tracking.

47:16

So here's where the conversation devolved as we tried more of the beers. So we're going to cut the conversation short and thank Ron Hockersmith again for his time. We really appreciate you coming into the studio and talking about your beers and your brewery and your scientific process with us. We absolutely mean it when we say we're going to be in to take a tour. And I know that I'm going to buy tickets for that three minute thesis. That sounds really interesting. I'm going to get another, nerd friend to come with me. So thank you again to Ron and Amorphic, and thank you all for listening to our podcast. We hope to see you next month on our next episode.