Sam Cartford - Interpretable AI and Stock Market News Monitoring

The conversation this week is with Sam Cartford. Sam is doing cybersecurity by day and is the co-founder of his own startup called Babbl, which he is working on anytime else. Babbl is a website that tracks stock market news sentiment, it aggregates and runs sentiment analysis on thousands of finance articles to distill high-impact information and ultimately help investors supplement their current research. This saves them time, reduces bias, and helps them gain insights. It was recently accepted into the beta fall 2022 cohort and was a part of Twin City Startup Week where it was listed as a standout by the Business Journal.

If you are interested in learning about how AI is being applied across multiple industries, be sure to join us at a future AppliedAI Monthly meetup and help support us so we can make future Emerging Technologies North non-profit events!

• Emerging Technologies North
• AppliedAI Meetup

Resources and Topics Mentioned in this Episode

• Sam Cartford on Twitter
• Babbl
• Twin City Startup Week
• CrowdStrike
• Robinhood
• Coinbase
• Red Wing Ignite
• GPT-3
• Stable Diffusion
• Sebastian Ruder
• Neuro-linguistic programming
• Beautiful Soup

Enjoy!

Your host,
Justin Grammens

Transcript

Sam Cartford  0:00  

You know, with all the tools that are available out there, you know, really easy to start, the best way I think to learn is to start playing around with something that's interpretable. Right? So not going straight for some deep neural networks that you know, do something crazy, like simple diffusion. For example, start with something like a text generation, statistical method, find a way to implement it, if that's your cup of tea, and kind of work through. Okay, I'm going to start with this, I'll realize what the problems are with this approach. And then let me find the next better one. So that kind of self learning with something that you know, you actually have a model you can play with at the end of the day. There's so many domains and so many new ones are coming online every day. It's more fun to just get started.

AI Announcer  0:39  

Welcome to the conversations on Applied AI podcast where Justin Grammens and the team at emerging technologies know of talk with experts in the fields of artificial intelligence and deep learning. In each episode, we cut through the hype and dive into how these technologies are being applied to real world problems today. We hope that you find this episode educational and applicable to your industry and connect with us to learn more about our organization at applied ai.mn. Enjoy.

Justin Grammens  1:10  

Welcome everyone to the conversations on applied AI Podcast. Today we're speaking with Sam Cartford. Sam is doing cybersecurity by day and the co founder of his own startup called Babbl, which he is working on anytime else. Babbl is a website that tracks stock market news sentiment, it aggregates and run sentiment analysis on 1000s of finance articles to distill high impact information and ultimately help investors supplement their current research. This saves them time, reduces bias and helps them gain insights. It was recently accepted into the beta fall 2022 cohort, and was a part of Twin City Startup Week where it was listed as a standout by the Business Journal. Congrats, Sam on your success. And thank you for being on the program today. Hey, thanks, Justin. You know, I gave a little bit of highlights with regards to where you're at today, maybe you could give a little bit of a short background. You know, I know you graduated from University of Minnesota with a computer science degree sort of how what sort of brought you into this field since since graduating from school.

Sam Cartford  2:01  

Yeah, graduated with a degree in computer science focus in artificial intelligence, machine learning and straight into the pandemic, you know, I was on a beach in Georgia, or Alabama when I figured out that it wasn't going to be going back to class. So right after that I was lucky enough to get a job doing cybersecurity. And around the same time also, you know, turned back to a research project that I worked on before we find out co founder. And we turned that into something that we first were able to use ourselves, I saw a lot of value in it and knew a lot of people were like us out there who are invested in stocks, but don't want to read all the news, and all of the clickbait and all of the bodhran articles. And so now yeah, we put something out that anyone can now go and sign up for which is huge. We have almost 200 users now. And we're working on some enterprise deals for some folks in the publishing space.

Justin Grammens  2:53  

Very, very cool. Well, good. And so you know, you you're kind of doing the cybersecurity stuff during the day, like you said, you got it, you got a job. And in that space, do you see me sort of just kind of riffing on his, but you're seeing AI being used in that space as well. I mean, do you guys touch any of that stuff?

Sam Cartford  3:09  

Yeah, a lot of the machine learning and AI that's baked into cybersecurity all lives in the tooling space. So I don't get to do any of my own unfortunately. But things like CrowdStrike come along with some some really good machine learning under the hood. It might be just stats deep down, who really knows, but it does work.

Justin Grammens  3:26  

And I assure you, I know I have a friend who actually just got he got hired by CrowdStrike, maybe I don't know, three or four months ago or so he has an engineering manager. So I should definitely ping him maybe have him on the program in the future. Now that now that you mentioned that. What's interesting is you're still pretty young in your career. Yes. Right. So you've been you've only been out of school. You know, I've been out of school for, you know, more than 20 years now. So to me, since you've only been out for since the pandemic, it's still pretty young. But you know, what, what, I guess most people aren't investing money at your age. Do you think you're a little bit of an anomaly in that space? And how would this become a passion, like if you always kind of wanted to be an investor or been an investor

Sam Cartford  4:01  

I was seeing in high school and in college, the financial freedom that comes along with good investments early in life, and it seemed like a low cost way of, you know, making sure that you know, it's an insurance policy on retiring eventually, I'm not that much for anomaly anymore, to when you look back at GameStop kind of flood of investors, which is kind of who needs some new tools. So folks who need someone to help them read the news and to get better information. So we saw a 7 million new investors between 80 to 24 years old, come online, in 2021 alone, and the new cohort was younger, more diverse, and from lower socioeconomic brackets than than ever before. So these folks you know, increased returns on investment means you know, really increased quality of life so that's what we're trying to get in front of

Justin Grammens  4:51  

Oh yeah, so age doesn't really matter, I guess background or experience doesn't matter. And, and are you thinking about like, you know, when I started thinking about like Robin Hood, and you know, just didn't just like Coinbase and somebody other place where people just kind of throwing money in and not really fully understanding what they're doing and a lot of and a lot of cases, right.

Sam Cartford  5:07  

Yeah, it's an expensive way to learn about investing through through getting burned. But it's one of those steps you don't touch twice.

Justin Grammens  5:14  

Yeah. Tell me a little bit more about about Babbl, then, you know, you guys are doing sentiment analysis. I mean, I know that a lot of proprietary stuff going on. But maybe you can give us a general feel with regards to sort of the overall structure of the system and sort of how it works from an AI ml stamp.

Sam Cartford  5:29  

Yeah, yeah. So we take in around 30,000 news articles every week, in addition to a few 100,000, tweets, and some Reddit posts, and we take all that raw text. And we you know, a lot of our secret sauce probably lives in the the cleaning processes and the stuff that we do to make the text usable. And we call it sentiment analysis, what we're doing at the end, but really, we just have a lot of more heuristic based sets of things that we're that we're looking at to get the most information out of that text. Rather than, you know, a lot of the open source models and stuff that comes out of the box, you plug in a sentence, and you might get a score between negative one and one. And we're finding that that's just not representative that the actual information that you want isn't embedded in that number. So we try to parse out more categorically how you can classify the text, you know, is it hyped? Is it informational? Is it someone trying to sell something? You know, does it look like it's written by a bot, we take a lot of that information and use that to build kind of a denser representation of each sentence that we're seeing and use that to figure out how interesting that company is? Or what, you know, whether that investors should look at that company or not.

Justin Grammens  6:37  

Gotcha. And so me as an investor, what are some of the knobs I can turn right? Am I am I am I looking at blue chip stocks? I just put in the word blue chip? How does that typically work? Yeah,

Sam Cartford  6:46  

so a lot of the output right now is stock based. So the ticker that you want to search to go in, and you'll get all the information that we've seen on that the good articles, how often bots are writing or tweeting about those articles. But we have a lot in the pipeline and looks a lot more like knobs that can be turned, we have, you know, want to add a feature that's like the Google trends of maybe stock news, you can plug in a word and see how often that's getting mentioned in what kind of contexts you know, positive or negative or more macro focused, more individual business, we have a lot of tooling that we're that we're going to be sending out, adding to people where it shouldn't be more like a blank canvas, where any question that you want answered, you should be able to find inside of the platform.

Justin Grammens  7:26  

Nice. Are you finding yourself? So I I've done a lot of since I've been kind of on the internet since the beginning, I guess. I've done a lot of like scraping and you know, basically text processing websites. And oftentimes, you'll just start getting blocked, right? Like sites sites will start blocking you. I mean, are you guys sort of like using Google News as your initial feed or just a mixture of all sorts of other things that maybe you find online that maybe you can get into? Are you running into some of these things? 

Sam Cartford  7:52  

Yeah, that's been an interesting problem, like the kind of deep down guts of getting our data if we pulled most of it from API's. So we can avoid any sort of copyright gray areas, we do do a little bit of scraping with people who we know are okay with that. And our current volume, we are doing 30,000 articles. It's been enough. The problem though, and kind of like you were talking about with with turning the knob, there are, you know, 10% of stocks that get 90% of the gains. And so, as people come onto our platform we've been seeing they go looking for stocks that are maybe outside of the top 100, and therefore might only get two articles a month. And so we go to places that are more dedicated to those specific stocks and stocks into those categories and get those plugged in. So we have no wider range of data and tickers that people can look at.

Justin Grammens  8:40  

Makes sense. I mean, your business model is a subscription business model, I'm guessing, right? So I pay a monthly subscription. And you guys go out and do all the heavy lifting for me, which saves me a lot of time from going to Google News and typing in some of these same tickers. Right, I guess that's the alternative as people are doing. Yeah. Okay, so now I want to invest, right? Like, do I open up an account with Charles Schwab? Like, would you guys do some ads through that way? Are you looking at something like on the backside with this as well?

Sam Cartford  9:05  

Yeah, honestly, we'd love to partner with a broker somewhere along the line, someone who has those accounts, and that we could just bundle a subscription to Babel inside have you know, their access to their brokerage. Otherwise, right now, we've just been kind of taking a bring your own broker approach to people who have accounts, and they've been, you know, we'll we'll add features to add to allow people to add their portfolio into their watch list. That'll be a good way to make it really easy to make sure they're getting the information they need.

Justin Grammens  9:31  

For sure, for sure. How do you guys deal with I guess, you know, bias in some way with regards to like, Jim Cramer is always posting something right. So when when some of his stuff, kind of start feeding the algorithm more towards some of these people that are always putting a lot of content out there and some ways you guys trying to mitigate that?

Sam Cartford  9:48  

Well, this is probably the most interesting piece of what we get to work on the kind of source reputation and also like the the value of the information that they're putting out. We have just kind of as a step back, we have have some people who want to maybe trade alongside the herd, as stuff is happening, they want to be in on the wave, and we have some people who want to get out of the way. What we'll know is how influential that person's information is, on the tickers they're talking about. If we see that every time that Jim Cramer tweet mentions the stock positively, you know, you see an 8% bump or something over the next month, we'll let people know that, Hey, Jim Cramer just mentioned this, you know, they'll get a higher weight maybe and how we're looking at projecting that stocks, future performance. And then information will, you know, give us a metric of how often folks are right, that set the platform standpoint, like Twitter all the way down to the individual users. And the sectors that they're talking about? The flip side of that is we down weight information that comes back and we know isn't novel, right? We can look back against every sentence that we've processed, and see on the spectrum of this has never been said before to everyone who's saying this, how novel this information is, and therefore, like how likely it is to be useful to the people who care about that stock.

Justin Grammens  11:01  

Cool. I mean, you're just providing the information so people can take it or leave it, right. I mean, it's just, it's just like, Hey, make your own decision. And at the end of the day, this is sort of what we're what we're so yeah, I saw that you guys. Were free for a period. But now you're going into just a paid paid model only is that right?

Sam Cartford  11:16  

Yeah, yeah, we were free for the first quarter and 50 people, we shut that off last week. And so we'll slowly ramp up the price as we add new features, and larger and larger groups of people. It'll be $4 for the next 250. And then 10. After that, well, very good.

Justin Grammens  11:32  

Well, so I have liner notes, and all sorts of text transcripts of all these podcasts. So we'll make sure to put your website but it's it's babble dot Dev, is that right? Yeah.

Sam Cartford  11:41  

Yeah. And the web itself, the website itself is on app dot babble dot Dev. And we'll have a discount code for you. 

Justin Grammens  11:45  

All right. All right, cool. Yeah, whatever you can give me I can push out to the audience. You know, I obviously, enjoy talking about AI and machine learning and its applications. That's what this is all about. So you guys fit right in that. But also, I just I love touring and, you know, talking to startups and other founders that are doing some unique things, you know, they're just getting into this space. So it's really cool that you've taken the LEAP here. And I remember talking to you and your co founder, when you guys were doing a thing with Red Wing Ignite, I think maybe a year maybe a year or more ago. Yeah, it was.

Sam Cartford  12:15  

We've been to a few startup classes since then. That was our first the first time that we learned about how to talk to customers.

Justin Grammens  12:21  

How's that journey? Ben kind of walked me through the past couple years as you guys have been building your business?

Sam Cartford  12:24  

Yeah, yeah, we started our first cohort through IoT based out of St. Cloud Nick teats if you're listening. Thanks, again, for being a good teachers. Awesome. I loved it. Yeah, we learned a lot, you know, in those kind of early classes, because my co founder and I are both pretty technical, and not, you know, to expose the business side of things. And so you'll learn a lot of the methodologies and tools and tricks that you can kind of use to make sure that the thing that you're actually making will be used and is wanted and needed. You know, we kind of came into this with just a data pipeline, where we knew that we could get a lot of information out of stock news texts, and make sure that people didn't have to spend all their time reading, but we didn't exactly know. Yeah, how do you talk about that? How do you make sure when talking to potential users that it actually does solve the problem that they, you know, need to all?

Justin Grammens  13:11  

Awesome. And so now you guys are in this beta program? How much longer does that go?

Sam Cartford  13:15  

It goes through February. So we'll be in through the rest of the winter here. We ended up winning the showcase during Twin City Startup Week, which was awesome. It's been a lot of fun. It's obviously our classroom is five blocks from my house here in Northeast. So that's been ideal as well, the people that we've gotten to connect through that have been super, super helpful in helping us figure out the best ways to add people, how do we start thinking about doing PR? Yeah, how do we get a lot of people exposed to what we're working on and make sure that we can get as many of them as possible activated on the platform and save them time and help them make more money?

Justin Grammens  13:48  

Yeah, for sure. Can you continue to use that space after you're completed? Or it can be a resource for you the three blocks away? If you want to get out?

Sam Cartford  13:55  

Yeah, I do like getting out of the basement here. So that'll be something I keep pulling on, as long as they keep letting me in.

Justin Grammens  14:02  

It? Sure. Sure. So you're getting all the content, you're doing analysis on it? Are you guys looking at like GPT? Three, or some of those types of things into then to summarize and write it back? Maybe already doing it? You know, I don't know. I'm just Just curious. Yeah,

Sam Cartford  14:15  

we've been kind of using ensemble approach for how we turn a lot of information in something more digestible. The simplest method kind of on our spectrum is looking at sentences that are the most information dense, right? And we kind of classify that through looking at the numbers that are shared, you know, are they talking about percentages, talking about money? Are they talking about things that are in the future versus in past, you know, they're speculating reacting? And so those individual sentences get pulled out. And then we'll also do you can almost think about it like, it's almost like a convolutional neural network, CNN, where we're taking those kinds of convolutional steps to make it more dense summarization, you know, individual parts, to make sure that we're getting the most information that we can out of it.

Justin Grammens  14:57  

Well, cool. Do you read on AI a lot? What sort of other books do you like to read? Yeah, I

Sam Cartford  15:01  

still have all my textbooks from school here. Otherwise, I've been following a lot of really interesting researchers on Twitter, they, you know, will let you know, when the new important things are coming out. So, you know, I looked at stable diffusion right away and some of those image generation kind of network and structures. And that was, you know, incredible to see, I think it's Sebastian ruder writes in a really good newsletter on the state of Endo. And that will be a little bit of, you know, stuff with the large language models, you know, the ones that we all know about, like GPD, three, as well as some of the up and coming research papers that he's found at conferences.

Justin Grammens  15:37  

Yeah, I mean, what have you seen other projects using NLP out there that have maybe caught your attention at all? Or maybe it's maybe it's just the stuff that you're working on in that in that sphere? But I'm just curious, or even across all AI, to be honest, or is there anything that you're seeing in the news, you're like, Wow, that's a pretty cool application. I hadn't really thought about doing it that way, or it being used in pacity? Yeah, in the

Sam Cartford  15:59  

NLP world, the most interesting things to me have been the code generation, right, like, mostly the fact that teams that start using like GitHub co pilot, the second that it's gone, they really, really missed it. And they found that their engineering time was reduced by maybe 40%. Because they didn't have to write things that, you know, are pretty easily inferable from the function that they're writing or what have you. So that one I've been following really closely. And, and thinking about, you know, buying a subscription or adding it to our, to our GitHub license so that we could take a look at that.

Justin Grammens  16:31  

That's cool. Yeah, yeah. So I teach at the University of St. Thomas in their master's programs. And as an educator, I actually do get a free license to try it out. And it's been on my list, it's absolutely been on my list to be like, yeah, just let's just play around with it. And I didn't really have much data on it. So that's really interesting to hear that, like you said, 40% of the people that use it, actually miss it once you take it away. That's a that's an interesting metric. I'd heard about it. And I've actually seen some demos on it. But I didn't really know how much of an impact it was making. Real world.

Sam Cartford  17:00  

Yeah, teams, teams at enterprises have loved it. It is a little expensive. Let me know what you think of the free one. And then maybe I'll get one for me and our front end developer?

Justin Grammens  17:09  

Yeah, sure. How big is your team, I'm guessing it's to two co founders, and then you're just contracting and other people to do the work with you. Yeah, we

Sam Cartford  17:15  

have our front end developer, the second that we raised some money, he'll be getting a full time offer from us. He's finishing up his last semester at the University of Minnesota. He's been awesome. You know, like certain people when they come on board, this stuff that they can do, and the effort that they bring to the table. You know, it's it makes it a really easy decision to say that they should stay around. So we'll be making it worth his while shortly. And other than that, yeah, just me Ramsey.

Justin Grammens  17:37  

Gotcha. And you guys are focused on you guys are kind of developing the models and stuff, right, this parsing algorithm, all that stuff is stuff that you guys do a lot is it mainly in Python, it's all

Sam Cartford  17:47  

Python, the whole back end, or API server, I wish I knew more other languages, but pythons just been the most fun and easiest show.

Justin Grammens  17:55  

Yeah, for sure. Nothing wrong with that. I mean, why reinvent the wheel, if you got a lot of libraries out there that can help you BeautifulSoup or whatever it is, you know, to have to go through API's got super easy to use that.

Sam Cartford  18:06  

And the machine learning support inside of Python has been awesome. Like, I don't know, if you've ever heard of spacey, the text processing framework. Basically, they have this data pipeline feature where you can add your own custom features to the stuff that comes out of the machine learning models. So that when a piece of text runs through, you know, it's basically running on bare metal See, essentially, and you can add whatever you want to it. And that comes pre compiled with the machine learning, or the language models that will have the text embeddings. And some of that,

Justin Grammens  18:35  

that's cool. I've been actually working on kind of like a text summarization solution for myself, because I, I put out a curated newsletter every week or so. And it's really been focused on Internet of Things, or the AI IoT artificial intelligence of things. And so I find all these articles, and I curate them and sort of like, summarize them and write them up. And I've been been kind of dabbling. This has been I've been I've been putting out in the Artic newsletter out for four or five years now. So I've got hundreds of issues. But you know, it's very manual. And I realized early on, oh, geez, how could I have it essentially do some of the summarization for me, and you know, what, what I found was open source tools are good, but they only get you so far. What I've also found was there there will be like, A, there will be an ad in the middle of the text. And then that thing shows up into summarization, right as some of these things. So it's been it's been interesting. I was like, oh, that's I'll just run this through, you know, this network here. And it definitely generates text and it's definitely decent. But you can tell Yeah, you can tell it's like not legit. And there's other garbage that's put in there and even around like, like it doesn't capitalize AI for example, like there's things like that that of course it just that you need to sort of tune in. So I'm guessing you guys are running into a lot of that. Well, yeah. What's what you've been doing is probably doing a lot of that type of the value add work on top of some of these libraries.

Sam Cartford  19:52  

There's a few there's a few key things in that especially with text with text, I think more than than most other fields. Cleaning is one of the most important pieces, you know, if you don't have something that looks like this, if you actually want considered by the model being passed in, you'll get side effects that you don't want, stuff like that, when you're looking at the open source libraries, they do a good job out of the box on out of the box problems. But the second that you have something more niche, it'll be tough to get something that looks clean out of it. What I would recommend, though, is to look at hugging face, if you've seen any of those models, they're a website and also a pip library now, where you can host and have your models kind of benchmarked against all the other models in the same kind of problem set, then they're available for download and install on the command line and inside of your Python programs. And we use a decent bit of that, for some of the ensemble tasks. What do you think about though, like, the first machine learning model neural network that ever beat someone in chess? I don't remember it was deep blue, or what exactly. But they still use the same approach where it's part statistical and part neural network, right, they get the game state and like you get your text into something that the machine learning model can interpret. as well. It's possible, using some kind of heuristic and statistical methods up front.

Justin Grammens  21:16  

And it seems like ensemble methods are usually the best across the board and solving anything. I've been doing some data science problems, there isn't really one that's the best you kind of try and bring them all together, there's

Sam Cartford  21:25  

a paper ensemble is all you need.

Justin Grammens  21:30  

I'll look that up. If I can find that. I will add that to the notes for sure. Like people after that, if I'm just coming out of school today, and I want to get into this field, what do you suggest? And not even not even NLP? I guess in general, I mean, so rewind the clock back a couple years. Yeah, you know, if someone's interested in AI and ML, what do you think they should be learning or studying or going after,

Sam Cartford  21:49  

you know, with all the tools that are available out there, it's, you know, really easy to start, the best way I think, to learn is to start playing around with something that's interpretable. Right. So not going straight for some deep neural networks that, you know, do something crazy, like simple diffusion, for example, start with something like a text generation, statistical method, you know, find a way to implement it, if that's your cup of tea, if you want to get into the code, and kind of work through, okay, I'm going to start with this, I'll realize what the problems are with this approach. And then let me find the next better one. So that kind of self learning with something that you know, you actually have a model you can play with, at the end of the day, that's been the best way to learn. And then also, you have something for a portfolio, and it makes it a lot easier to find out what you're interested in, right? Like, if you find you want to do object recognition, you know, maybe you want to go work at a Tesla, AI, you can you can do a lot of that stuff that you there's so many domains, and so many new ones are coming online every day. It's more fun to just get started and and work through the math, maybe not until you need to. Yeah, for sure. For sure. Don't do any linear algebra right now.

Justin Grammens  22:54  

Because it's not really, those are sort of things that happen below the water lines. I don't really need to touch those at all. So yeah, it seems like in school, of course, you know, you're kind of learning how to learn something new. Some of the rigor of that is just becoming somebody who's curious and can pick up the things but but you're right. Yeah, a lot of stuff. You you learn, you might get into the depths just because you kind of need to know at least at a periphery. Yeah. side of it. What's what's working, but yeah, stand on the shoulders of giants, I guess in a lot of these a lot of these cases. Yeah,

Sam Cartford  23:23  

I love my classes and the projects that we did. But I also had to take four different classes on data structures and algorithms. And I can't tell you the last time I had to reverse a linked list.

Justin Grammens  23:34  

Right, right, right. Yeah. There's methods to do that right now already built into the Python. So how do people reach out and connect with you, Sam?

Sam Cartford  23:41  

Yeah, I'm on Twitter. My personal Twitter is still mostly jokes and a little bit of machine learning. But that's at SamCart. And then, for Babbl stuff, you can find me on LinkedIn post about what we're up to, you can also follow Babbl on LinkedIn. Also, we'd love for people to check out our free now a weekly newsletter, where we're posting stuff that we're finding in the in the stock market, in the news, what's interesting, we did some back testing on the calls that we've made inside of just our newsletter alone. And we found that we made a decent return so far, you know, kind of in any market. So that's been cool, proves that hopefully, there's something in the data that we're actually parsing out, I think, yeah, I would love to talk to some people, you can email me, you could add that to the notes. Maybe we can get as many people signed up as possible. If, if that's at all interesting to any of your listeners, for sure.

Justin Grammens  24:28  

And it's ba BB L, right?

Sam Cartford  24:32  

Yeah, we may have to change the name. I don't know what's gonna happen there.

Justin Grammens  24:35  

Honestly, that's a good problem to have, and it means you're actually getting noticed. You know, so if you just sort of fly around under under the radar, no one even cares what you're named. That's not really the best thing. So good. That's cool. Is there is there anything else that I maybe didn't didn't cover that you would want to talk about in the center?

Sam Cartford  24:52  

I think we got through a lot of the stuff that we that I wanted to talk about, you know, this, this field is endlessly interesting. To me, especially in NLP being in, it'll be the first field of artificial intelligence that really convinces people that, you know, AGI is possible, or something just due to that, that humanity of it. You know, it's been a really fun place to work. And hopefully people are also interested in reach out and see what we're up to and sort of some of their own stuff, too. I guess

Justin Grammens  25:19  

before we do go, like, where do you see this going in the next, you know, three to five to 10 years? And you mentioned AGI that was something that is always a hot topic for debate. But where do you guys see you guys being I guess, when would you like to be maybe five years for Babbl? Or Babbl? But also probably NLP? You know, and just what's what's going on in the space? Yeah, yeah, for Babbl.

Sam Cartford  25:42  

It would be amazing if we could prove through the fun that we run ourselves that there's something in the news. And then in text we process that we can use to forecast the future. But for you know, thinking about AGI which I would always put in quotations, I think we'll see. Problem solved, like we never would have imagined would have been able to be solved. Well, you know, quality of life increases. Thanks to that. As long as the technology is distributed, distributable enough. I think that we'll see that shortly. But I don't know that AGI is, is a possibility being that we don't have a good understanding of what intelligence is, you know, artificial or otherwise, we won't be able to slap a functional definition, I think on what does it mean to be sentient? What does it mean to be alive? And just due to that? AGI? I don't think we'll ever have a definitive yes or no,

Justin Grammens  26:33  

that's a good point for sure. I guess. Yeah, we haven't really defined it as, as an industry yet to make sure we know what we're going. And

Sam Cartford  26:39  

the Turing test was one of the worst benchmarks maybe. And Turing is obviously the genius and has done, you know, more for the field than maybe any other person. But the Turing test of can you convince a person that this thing is alive? I mean, you can do that by statistical methods alone?

Justin Grammens  26:54  

Yeah, I just am thinking that, you know, some of this artwork, some of this poetry and Officer stuff that's being generated, I don't I you could easily fool he's absolutely, with some of those things. Because it's just, it's so sort of, like, open ended. And so then, you know, it's interesting, I was on this panel some weeks ago, but then it does come back to then. Well, so what is good art? Right? And what is not right. And, and, and then what is art, too, right? And so it's in A, it's it's interpretive, you know, and it's interpreted based by based on your interpretation of it. So it's very difficult to prove. So definitely, with the map,

Sam Cartford  27:27  

right, a stable diffusion came out and was open source and available, a digital artists from Colorado actually won an art competition and didn't reveal at the end that it was aI generated art. And that was kind of one of the impetus sins for some of the really spirited debate around, you know, are these people artists? Or does it matter? That it was made by person? You know, more than what I think of what was generated?

Justin Grammens  27:52  

For sure? Yeah, I do remember hearing that they were, for lack of a better term, I guess. How do you have a AI win in our contest? So Well, cool. Sam, I appreciate the time. It's always always fun to talk to people that are in the space, always fun to talk to entrepreneurs, and I wish you nothing but the best here, it's it sounds like a really, really cool problem you guys are trying to tackle here. And it looks like you know, you guys are continuing to sort of win awards, work your work your way up, and also, you know, gain subscribers. That's that's really what it's all about. So congratulations on crossing that that first threshold, and I know you guys will be continuing to add more in the future and more than happy to help you get the word out. So exciting times for you guys.

Sam Cartford  28:32  

Thanks, Justin. You know, I'll call it a win the first time someone emails us and tells us Thank you. Your information has allowed me to make so much money that I can retire and my whole family is going on vacation. So any day now hopefully.

Justin Grammens  28:45  

Maybe that person will be you one day, you know, I'm sure you use the tool like you said so you could benefit as well.

AI Announcer  28:52  

You've listened to another episode of the conversations on applied AI podcast. We hope you are eager to learn more about applying artificial intelligence and deep learning within your organization. You can visit us at applied ai.mn To keep up to date on our events and connect with our amazing community. Please don't hesitate to reach out to Justin at applied ai.mn If you are interested in participating in a future episode. Thank you for listening