Species X is a Project that Runs the Gamut from AI to GE
Feb 22, 2023
I heard about Species X from my colleagues Bill and Laura who visited you at the brewery not long ago. They told me about it and it made me curious to find out more!
Yeah, Bill and Laura came by. We all hit it off talking about nerdy stuff. I’ve actually been working on Species X for a few years now. It was initially incubated under Aslin, after founders Drew and Kai approached me about starting a contract brewery with them. We’ve since decided to take a shot at it with a physical location in Columbus, Ohio. And we’re breaking it off from Aslin to stand on its own. Once we find a location, things are going to move fast. So, I’m excited.
Species X is founded on a concept I don’t think I’ve ever heard of before for a brewery. It has a lot to do with experimentation with ingredients, and especially yeast, but some other big things, too.
Yeah, the majority of Species X’s focus first has been on stuff like genetic editing, spore germination, colony isolation, forced mutation, and hybridization; things that are at the absolute frontier in beer.
It will involve, and has involved a lot of experimentation by me so far, but we’ll incorporate new products and research like the stuff like Omega Yeast and others in the industry are doing. Whatever is at the forefront of what’s innovative.
In my lab back at Aslin, I did a forced mutation using UV light just based on a scientific paper I came across. As I was reading it, I thought to myself, ‘I could probably figure that out.’ So, I purchased the piece of equipment and tried it.
In another experiment with a Hanseniaspora uvarum strain I’d caught off of a grape, I liked its profile — it was like straight orange juice except for a strong ethyl acetate note, and I wanted to eliminate that note with the technique described in the paper. I tried it and it had an impact, but ended up making the note stronger.
I’ve also done simple plasmid transformation to make colonies fluoresce in UV light. That was pretty straightforward and cool. You can even buy kits for that online.
Yet another experiment I did was to isolate yeast from a dried sourdough starter that I got from a Welsh family. The starter was supposedly medieval in origin. It made a really fantastic hazy IPA, which kind of makes sense, since it is English in origin.
Doing all of that took a lot of time and can be expensive, so we’re going to continue to pursue those sorts of discoveries, and also incorporate emerging innovations from other research leaders in the industry — including whatever is in the pipeline over there at Omega that I don’t know about yet.
By the way, I wanted to add a funny coincidence — I read Laura’s paper about early dry hopping timing and haze formation, and how it clears up a beer literally just before she stopped in. We were making a West Coast IPA and because of that research we just put a little bit of hops with our usual house yeast in the fermenter, right during knockout to the cellar to experiment and it definitely did the trick. So, I thanked her in person when she came through the door. She saved us a huge headache. I’m always scouring for research papers, and Laura’s name comes up a lot. She’s done a ridiculous amount for the industry.
You’ve already produced three beers for Species X so far. Bill said they were tasty. I read some reviews that were pretty glowing, too.
Our first was Cantalouper. We used a genetically modified Conan strain that produced cantaloupe esters and flavors. To compliment that we rounded it off with some melon-like hops. That was a double IPA.
The next was Hyperthermic Arms Race. We use 22 different strains of kveik yeast, a Lithuanian farmhouse strain, and four different species of yeast: S. cerevisiae, S. pastorianus, S. bayanus, and S. uvarum. Then, in addition to that, I also threw in the medieval capture I got off of the sourdough culture that was dried. I isolated that, propped that up and put it in there with all the kveik strains. That was a mind blowing beer. It went through some really crazy fermentation and flavor profiles. There was a Christmas cookie phase, and then an all vanilla phase. I dry hopped it with like Sabro and Citra, or something like that.
Del Fresk was brewed with Torulaspora delbrueckii. The Torulaspora delbrueckii reminded me of a chardonnay, but it only attenuated 20%. When we threw in amylase enzyme, it turned out great: very dry with a lot of body even though it was like zero plato, I think because that strain produces a more than normal amount of glycerol. I added cucumber and lime to it because it had that sort of chardonnay aspect to it. I thought it’d be like a wine mixer. I really enjoyed how that one came out.
That’s kind of a cheat code to help yeast finish out: if you find a wild capture strain that tastes delightful but it’s an under attenuating species, throw in some amylase enzyme. It’ll dry out the beer, but if you want like a 7% Torulaspora delbrueckii fermented beer, you can do it. Same with Schizosaccharomyces japonicus. We also have other low attenuating strains that are good for NA beer stuff.
You mentioned another important aspect of Species X was related to AI.
We paused production because we realized we wanted to expand the project’s scope. I was getting really into AI and machine learning. At that point, I had reached out to a company in Europe who were experts and found out the budget for that sort of thing basically starts at like $50,000, so, it was like, oh, well, looks like I’m doing this myself!
That’s when I really dove in hard — machine learning, artificial intelligence, Python and R coding. I’ve been doing that for between a year and a half or two years. I have learned a lot. I’ve just been falling in love with keeping up to date and learning as much as I can and how to utilize it in every aspect of Species X moving forward.
I’m taking some really good classes online to receive a Certificate in Practice of Data Analytics from Ohio State University. And that’s also been an amazing help to set the basics and a good foundation for further study. It’s been great.
In simple machine learning algorithms, it’s just amazing what it can do when the data is clean and structured. The result of all the hard work of cleaning and structuring the data really shows up in the end product. I’ll be honest, it’s funny, there’s a similar ratio to cleaning and brewing beer and cleaning tabular data and coding algorithms. It’s like 80% cleaning. But the plan is to integrate as much automation as I can from the very get-go, and structure the brewery so that once more capital comes in and we’re moving, that we have direct reports everywhere where we want the automation to be, ready to plug and play. My approach is going to be rooted deep to extend throughout the entire process all the way to people’s reviews, so it’s gonna be a little different than just saying, ‘can you give me an IPA?’
So yeah, that’s what I’ve been doing for the past two years: on top of planning and executing the physical location, getting ready for that additional focus. What excites me about having both of those branches, very heavy tech and wet lab stuff is that those will cross paths at some point.
Can you give a little more detail about how you’re using AI for your recipe development?
My IPA model ‘studies’ all the IPAs I give it, and then matches up each aspect of the brewing process with customer reviews. I have a recipe generator that I put together that allows me to dial in specifics — ABV, number, quantity and variety of hops. Then I’ll run it. It’ll spit out thousands of recipes and pull the ‘best tasting’ to suggest to brew. It’ll have the entire recipe from production, the ingredients, timing, et cetera and it’ll make an inference on what it’s predicted to be rated as well.
We’re looking for stuff that is not already in use as much; we’re looking for the next big… X.
How does it sense or predict what’s best tasting?
It’s kind of like a hive mind based on the ranking of the beers it was trained on. And you can take the guard rails off. I have one model that’s crazy. I’m actually a little scared of executing anything on it because I took the guard rails off. I trained it on every single beer that I brewed. It’s not IPA-only or lager-only. It’s everything. It’ll spit out some styles that have never been created before, but still make sense. You know what I mean? So it’s cool.
I really liked the idea of giving people that they want, you know, and I mean, I find it really satisfying that my IPA model is at peak performance, and knows more than I do that I’m not seeing. And it’ll provide something so mind blowing. I wouldn’t have been able to do that on my own.
Say I come up with a recipe for an IPA, and I put it into my model, it will spit out a prediction for me. So I’ll be like, okay, this is a 2.5 rating out of 5, this probably isn’t going to go well.
Another thing — I actually did this with my IPA model the other day — I trained it, and asked for feature importance and it spit out the top 20 most important features of a recipe, according to reviews.
So not only will it produce spontaneous recipes for me, and enable me to dial them in, but you can also ask the model what the most important factor is throughout the entire recipe, what the thing is that influences the rating the most.
What did you find out?!
I can’t reveal it. But I will say, it’s one of the things that I didn’t think of at all.
Where did the name Species X come from?
I think of Species X as representing the next big organism in beer, meaning whether that’s artificial intelligence, or wet lab stuff with single celled organisms like yeast, or multicellular ones like hops, barley and fruit. We’re looking for stuff that is not already in use as much; we’re looking for the next big… x. Not, not what’s usual. And I guess in some sense you could also take x to represent extreme. Maybe some of that, too.
What else do you see on the horizon or in the future?
I think NA beer is here to stay. Those under attenuating yeast species, I’m gonna take those and integrate them into near beer and NA beers. I’m really happy about that. I am a huge fan of all the fantastic non alcoholic beers out there. So I’m looking to have Species X take advantage of that and utilize this growing market.
On the AI front, Laura and I actually talked about how it would be a great idea to do an algorithm that matches up with flavor profile. If someone had a chemical readout or breakdown of a beer’s esters and phenolics, for example, and matched it up with a rating that would be fascinating. I hope AI makes its way into more parts of our industry.
I think the next big thing in AI and machine learning is going to be quantum computers. IBM has a service if anyone’s interested to run code on their quantum computers via cloud. These things can compute everything all at once because everything’s at a different state at the same time. I’m still trying to wrap my head around it.
I think automation as well. Just in general, it’s going to be huge. Our industry is backbreaking work at times, and I think automation can improve work/life balance. It’s perhaps more accessible than ever before, or going to be very soon.
Last thing I was going to say is with, like ChatGPT, where you can just ask it and It’ll spit out all the stuff in plain English as opposed to what I’m doing now, which is code; it’s given a personality so that you can talk to it. That’s going to be important.
I want to do a lot more research and and help astronauts and people who are pushing the boundaries of human civilization to just give them a beer, you know what I mean? A fresh, well made beer.
One of the goals I have that’s maybe even further in the future is to start thinking about brewing in space, as crazy as that sounds. You have Blue Origin’s Orbital Reef coming out. It’s going to be a private ISS pretty much. And you’ve got a couple of other companies with similar plans. NASA is planning a moon base with Artemis.
In the past, beer has always been a pretty important part of being able to grow and survive as a civilization. I really want to be a part from orbital space stations to the Moon and to Mars. I want to do a lot more research and and help astronauts and people who are pushing the boundaries of human civilization to just give them a beer, you know what I mean? A fresh, well made beer. Fermentation and space, and fermentation in food is going to be important for long voyages.
Species X has even done some investigation into whether hops can grow in Martian soil simulants.
That’s a great image. I can’t wait to see the future as you envision it.
I know. I’m excited. I might be like a futurist at heart. I just I love the possibility of what we can do as an industry. And having people like Bill and Laura to bounce ideas off with when I go through this journey into the unknown. It’s good to have people at your side to help navigate.
Related Collections