Announcer: 0:00

Entrepreneur, author and financial consultant, Marc Bernstein helps high-performing entrepreneurial business owners create a vision for the future and follow through on their goals and intentions. Ang Onorato is a business growth strategist to blend psychology and business together to create conscious leaders and business owners who impact the world. Founders Forum is a radio show podcast sharing the real stories behind entrepreneurship as founders discover more about themselves, while providing valuable lessons and some fun and entertainment for you.

Marc Bernstein: 0:32

Now here's Marc and Ang, good morning America, good morning world, good morning universe, because it's expanding before us, before our eyes, or at least our awareness of the universe, is expanding. All kinds of space exploration going on again, and we're we are. The world is moving fast, that's my thought for the day, and it's almost hard to keep up with and I enjoy it. I think it's extremely exciting but at the same time, the awareness and trying to keep track, forget about the news, forget about the weather, forget about all that other stuff, but whatever, everything that's changing around, all of that, and certainly the news, is changing as a result of, you know, the intensity and quickness of media and accessibility to social media and all kinds of media by everybody, the democratization of media, all kinds of things. Ange, tell me about your experience with the pace of change in the world and our guest today, logan Dopp, who we're having him for part two. I'm sure we'll have some observations about that as well.

Ang Onorato: 1:39

Yeah, talk about the pace of change. And you know AI is I mean, I think it has such a small initial AI because it changes so quickly that you just can't even keep up. But you know, for me personally, you know I'm, I'm, I've slowed to adapt to change. In fact I get seasonal effect disorder and I don't even, I can't even believe the summer just blew by me in a in a blink and I wait all year for warm weather and long days. So you know, we're just got to keep up with it somehow.

Marc Bernstein: 2:05

And I call that a lunar on stat blue by you, yeah the blue by.

Ang Onorato: 2:09

exactly. Well, I don't know if AI could find a way to give me, you know, kind of stop time a little bit and let us enjoy the summer a little bit more. That would be great. But you know, here we are getting ready for early, early autumn it looks like right around the corner. So great to be here again and doing kind of part two with with Logan. I think this is going to be a super exciting and informative discussion and excited to kind of welcome him back to the show.

Marc Bernstein: 2:32

By the way, logan's in Florida, so he probably can't wait for the changes.

Ang Onorato: 2:36

He's a perpetual summer. Yeah, exactly.

Marc Bernstein: 2:38

I think he's anxious for some. If he's like most people there, he wants the summer to end, this summer to end anyway, right, logan?

Logan Dopp: 2:44

What do you think? It gets pretty hot.

Marc Bernstein: 2:46

Yeah, I'm sure, I'm sure. How are you, how are you today?

Logan Dopp: 2:51

Doing well. Yeah, today is not that hot, but it definitely in the winter seasons is where you really want to be in Florida and escape some of the cold.

Marc Bernstein: 2:59

So for you, this is a very loaded question, but how is the pace of change impacting your life?

Logan Dopp: 3:06

Oh man, I think it impacts everybody. And, Anjani, we're talking a couple of weeks ago about like a job posting for a prompt engineer for AI and it's like no one even knew what that was like six months ago and now they're trying to hire experts in it. I mean sorry. I mean it affects all of us and it's exciting to be in the AI industry, but also a little bit nerve-wracking because of the pace of change.

Marc Bernstein: 3:28

Logan and prompt to question that I'm thinking about as I'm talking to you. What do you? I mean, I know there's things that we're not even thinking about now that are going to exist three months from now and a year from now. What do you see as far as that goes, from your perspective as somebody who works in AI full time? What do you? What, what? What is going to? What is out there that we're not even imagining it right now? That's going to happen.

Logan Dopp: 3:53

Yeah, I think, if we just step back and talk about like the phases of artificial intelligence, that at space level we've had machine learning types of technologies for a long time.

Logan Dopp: 4:03

I mean that kind of technology has been around. And in the future, when we look at like AI that can bridge the gap between just identifying something or just generating a response and actually intuiting things, that moves out of specific AI, which is kind of the phase that we're in now of AI has a specific function, at different form, into more generalized AI where it can interpret and connect the dots between things. And we're still several years away from generalized AI. But once that becomes part of our mainstream, how do we as humans then use that as a tool and think about how we can make our lives faster? And I always am interested to hear the academia debate on do we let students use chat GPT, do we not? And kind of my philosophy is technology is here to stay and especially when we get into generalized AI that can intuit these things, we as users have to figure out how to apply it to what we do to make us better and more efficient, and not try and reject technology. But it's a scary space to be sometimes.

Marc Bernstein: 5:06

This gentleman I was talking to the other day said everybody first step ought to be they ought to be on chat GPT right now. They should be doing it right now because get their first taste for it. I did it by accident. My mother turned 90 recently and we were having a birthday party and everyone was gonna say some words about her and I wanted to write a song and I was stuck. So chat GPT and I wrote a song together and it came out pretty good. It was pretty interesting experience. We got all this stuff going on with the writer strike right now in Hollywood and they should be protesting because. But on the other hand, I think it goes back to if they have their arms around AI. I gotta think the human impact is still gonna be important in the arts, and particularly in regard to writing, and that the ones that know how to get their arms around it, I would think, are going to still have jobs and the ones that aren't, I'm afraid, will be out of jobs. Sounds like.

Ang Onorato: 6:09

And Logan, before you, kind of comment on that, for anyone that missed our first show, we should probably do a quick revisit of an intro to you. So again we're here with Logan Dopp of Torrent AI. This is part two of our conversation around not only his company but really everything AI that we know today and how this almost be kind of a public service kind of show to sort of help provide information, your insights and education, maybe help normalize some of the vocabulary around it. So with that, just to kind of let catch people up that maybe didn't hear the first show yet. Logan, I'd love to have you comment on Marc's last question or thought there, but maybe expand on that as well, and we talked offline about that. There are multiple different kinds of AI and you alluded to that as well. So maybe just give us your sort of like take on the industry. Really, what is AI? What are we familiar with that we didn't think of or don't typically put it in that category and just help our listeners kind of get up to speed with all that.

Logan Dopp: 7:10

Sure, yeah, as I've been out talking to users and our customers and attending conferences and things like, I think people would be surprised when they dig in at how many industries even in heavy manufacturing and things we don't necessarily think as tech forward they're using AI in a lot of ways to automate some of their processes, make their employees safer. And for our company, torn AI, we've been around about three years and we were originally founded out of a program called Hacking for Defense, which is where the DOD will give universities a problem to try and solve, and the problem for us at Georgetown was the Office of Naval Intelligence wanted to use cameras to go detect and identify boats out at sea. So our company uses computer vision, which is one kind of AI that takes images, video feed or still images, and lets the AI decide what it's looking at and give you a classification of something. And computer vision has been around for a while, but the pace of AI expanding is allowing computer vision to detect more and more types of objects and more and more nuanced things.

Logan Dopp: 8:17

And then chat GBT that you mentioned, Marc, is a large language model, so it's a text based AI that can generate new text and some of the exciting things is that it can help you write a song, but some of the maybe concerning things are it might invent case law that never existed. When it gives a brief to a lawyer. So you know, I'm a lawyer too.

Marc Bernstein: 8:40

I don't know if you knew that, but I am a lawyer by background, so I have thought about that. In fact, I'm glad you mentioned it, because someone I was talking to the other day another AI business that's more marketing oriented mentioned about. You know, if chat GBT doesn't know something, they'll make it up and they call that hallucinations. So I didn't want to take you off track because I want you to continue on the different types of AI, but when we get to it, let's talk about false AI as well, which I think is where you're going with that, right.

Logan Dopp: 9:10

Yeah, I mean, the hallucinations are fascinating and some of the things in false AI is you know, we hear from a lot of people that you know AI is Skynet from Terminator or Big Brother and it's going to do all of these things. And you know, perhaps the future, that kind of generalized AI might get to that point. But for today, we almost face the opposite problem that AI is quite specific in what it does and it needs to be monitored for accuracy and it needs to be implemented in the right way. And slapping the word AI onto whatever you're doing doesn't necessarily make it an AI and in fact the SEC has released some updated guidance on, you know, keep your AI claims in check. You can't be promising the world for something that's unproven.

Logan Dopp: 9:57

And with something like chat GBT, where each version of that model is being released to the public to kind of get it in test, we're seeing some really fascinating things and kind of, as you know, the billion user mark of chat GBT. We're actively involved in that process of seeing AI get better. But as it stands today, a lot of AI is quite specific and is not the end all be all, and some of the things that we are using, you know, may not be also called AI. At the end of the day, sometimes it's just an if, then statement.

Marc Bernstein: 10:32

You bring up an interesting point and I don't want to go back to what we were talking about before, but you know we're still trying to get our arms around social media and whether that should be regulated or not regulated, and there's free speech issues and all that. But now you're talking about you know that falsities that could be in falsehoods, I guess, is the word in AI and gaps and hallucinations. Should, in your opinion, ai be government regulated? Have you given much thought to that? I guess you have, because you work with the government to a large extent.

Logan Dopp: 11:04

Yeah, yeah, and I've got some great friends and classmates that actually work in the field of AI policy and have had long conversations with them about this. I think you know, as a technologist and a technology company, we would err on the side of not regulating too much, because that helps the technology become innovative and work out some of these issues. On the other hand, there are some implications to how we use AI, especially in the generative world, where you can create scenes and scenarios. But at the end of the day, ai is a tool and I think that there are some of the various things that can be generated with a tool, but there's also a world of good that can be generated with a tool like that and we should err on the side of letting those things innovate and work themselves out.

Marc Bernstein: 11:49

Also just the display of likenesses. I saw a presentation the other day on you know. It's got, like you know, biden and Trump as super superheroes, and Elvis and Beyonce and all these people, and in fact they've already come out, I think, with a program that's coming out where you can do karaoke and sing in Elvis's voice. How do you regulate that and where is the line drawn between stealing people's personas and voices and likenesses versus artistic freedom and creative freedom and intellectual and technological freedom?

Logan Dopp: 12:33

Totally. And that starts to touch on some of the privacy issues too, that you know, if you can do computer vision, like our company does, where you can detect and classify something facial recognition has been around a long time but you know, at what point do you have a reasonable level of privacy that you expect to not be facially recognized as you walk around? And we see a lot of individual states that are banning the use of facial recognition within corporations or unions that are having that as well, so that the you know, the union members, the employees, have like a reasonable level of privacy. So yeah, I think the tool of being able to use computer vision to detect and classify things has massive upside but also some privacy considerations that we take really seriously.

Marc Bernstein: 13:18

And, by the way, another thing in the news this week and we're talking right now over Zoom and Zoom has just you know you're now automatically signing an authorization when you record on Zoom. We're not recording FYI, but if we were, they can pretty much you're giving them the rights to use that information for whatever they want and financial services. I mean. That's an issue. We're having intimate conversations on Zoom. We don't generally record, but AI is a good reason to record because we could take blips of conversations, do summaries for clients, send it out. But then you've got that privacy issue.

Logan Dopp: 13:53

So that's you know, and data is the goldmine. So AI is only as good as its training data, because AI, at the end of the day, is a model or an algorithm that has had data given to it and then taught how to make a decision, and once you output that, you can use it all over the place, but it's only as good as the data that was put into it, and so there's ways that you know, with this generative world, you can create new training data. But, at the end of the day, real data on a real environment is gold for creating an accurate AI model, and so it makes sense why companies like Zoom would want to be able to harvest that data for their own training purposes or whatever they're using it for. But yeah, I mean, data is in social. Media brings it up as well. Data is the commodity.

Marc Bernstein: 14:41

Data is the commodity. Listen, I'm also. I've been a Tesla early adopter, for I just got my fourth Tesla, and Tesla everybody thinks the value is in the cars they sell or the you know, but obviously they're a technology company and I always say, when you look at the stock, their most valuable asset that nobody really completely understands what they have is the data they have, particularly when it comes to autonomous driving, which you know right. So which? Which? Just as a little aside as I'm talking about this, you know I grew up with the Jetsons and the 1964 World's Fair and you know the future and I remember the 1964 World's Fair.

Ang Onorato: 15:19

It's no longer the future, Marc, it is what we're living in.

Marc Bernstein: 15:21

Well, look, we're right now we're talking to you over our iPhones and on Zoom, and I remember when they had picture phones and all those years I was thinking we're still using dial phones. And then there was touch tone phones One of those picture phones coming, and they never came. Instead, we now carry the pictures in our pockets. Yeah, you know, and just the future is here. I saw also demonstration on robotics, because we've been taught where are those robots, you know? Like what was their name on the Jetsons? Rosie, rosie, that's right, rosie the robot.

Logan Dopp: 15:50

So, but all of a sudden Boston Dynamics robot dogs are getting pretty good.

Marc Bernstein: 15:56

Well, I understand Tesla's coming out with one I understand, very shortly that's undercutting the price of Boston Dynamics, I think, by which is what he does like 20,000 instead of $26,000 or something like that that robots will cook for you and do everything that you imagined.

Ang Onorato: 16:11

You know it's, it's here, so I think that's the part that Really kind of terrifies a lot of people. You know that did a cartoon coming round. We have a break coming up here in a minute, but when we come back, you know, logan, I'd like to you know, have you got a little bit deeper into sort of exactly how you're you know, when you talk about the type of work that Toran AI is doing, particularly manufacturing? I think that'll be really fascinating conversation for people to hear. But one other quick thing is you know just to footnote on what you guys were just referring to is that you know the privacy issue and all of that.

Ang Onorato: 16:42

But I think there's a second layer here, that's that's an issue too. Is that, you know? Is the fact checking part right? I mean, you can today, if you go to factcheckingorg or snoopscom or things like that to verify things that we hear in the media that we don't think are real. But now you're living in a space where you know even AI can generate a hallucination and a fact checking perspective. So how does anybody know what's real and what isn't real, right? I think that for me, I think, is a big concern, right, because there's been deep fake video and things going around for a long time where you can splice. You can see it in political ads. They're splicing video and making it seem like people say things they don't, and so you know. I'm just curious about your thoughts on that side, which is the, which is the you know. So that's a point we're going to hold there, let's do a quick break and when you come back I'd love to hear your thoughts on all that.

Announcement: 17:35

Establishing a self-directed Roth IRA or other self-directed retirement plan offers access to investment options usually only available to well-connected or wealthy individuals. Anyone can build wealth and achieve tax-free income for life and financial freedom without the inordinate risk of speculative investing, securities, trading exotic financial instruments or hoping for the best from the stock market. Since their inception in 2003, camiplan has come to be synonymous with expertise, flexibility, responsiveness and diversity of investment choices. They provide innovative investment opportunities for investors seeking direct control and more choices in how they deploy their investment capital. Camiplan doesn't just make it possible to invest in alternative assets. They also teach investors how to put their money to work. Take the road to financial freedom today. Go to camiplancom forward, slash IRA.

Marc Bernstein: 18:37

We are back at Founders Forum and, ange, I'll let you continue. Maybe just restate your question real briefly and we'll give it back to you, yeah.

Ang Onorato: 18:47

So, just before the break, where we were talking about or following up on the privacy issue and just sort of the unknown nature of what direction AI can go, we've brought up the fact that it might be difficult to have fact-checking of what's real and what isn't if a thing like a large language model can make up fake news or fake briefs and legal or whatever. So, logan, love to hear your thoughts on that and then maybe take us down the route of exactly what your company now is doing in manufacturing, because I think it's going to be really fascinating for people to understand what the cameras do, how you integrate all that. So, for the remainder of the show, love to hear your thoughts on the tech side of this.

Logan Dopp: 19:23

Yeah, it's a great question.

Logan Dopp: 19:24

I think it fundamentally starts with an understanding of how AI generates information or comes to its conclusions, and a challenge in the policy realm, in thinking about how we handle these deep fakes or ghosts or, you know, generated things that don't actually exist, is we don't have a great way to explain how the AI model arrived at its conclusion.

Logan Dopp: 19:50

We have an input and we have an output. We don't necessarily have a gravy to explain that, and it again gets back to the data that there may be biases in that model, because, you know, maybe the AI has never seen a case of a goldfish walking on the land, you know, and so if that happens, if it's never been trained on it, it can't generate or make a decision on something like that, and so the way you put data into a training and algorithm may create an inherent bias, and we don't have a great way to really understand what the biases are or how it arrived at those outputs. And that's an active conversation in the policy realm and in the tech space which is really exciting, and I think everyone being involved in understanding like what is AI, how does it do its job, is going to go a long way in helping us determine some of those things.

Marc Bernstein: 20:40

By the way, that's also in the ethical realm as well, I guess, because that's where AI can have a mind and even possibly a heart of its own, apparently, right.

Logan Dopp: 20:51

Yeah, I think in the biases they have, it can have a certain leaning that we may not be completely aware of until we start asking it the right questions and seeing where that bias and leaning actually is.

Logan Dopp: 21:02

But you know again it comes back to training data that if you give it other training data to other leanings, you'll be able to make more intuitive decisions.

Logan Dopp: 21:11

Just like humans are exposed to their own environments and develop some of their thoughts and opinions based on their environment as well as their own nature. And I think you know, for us at Tor and AI, where we're using video feeds, we want to make sure that the video coming from the customer is part of its own training data, so that we're not imposing a bias on what a hard hat looks like in someone else's job site. We're doing what is a hard hat or a forklift to look like on your job site and using that video feed from the customer any video feed they have as a training data helps to improve the AI and make it more accurate and more explainable in their particular environments. So not only are we using the video for detecting and classifying things like a forklift or proximity when there's a person too close to an exclusion area, but we can also then begin to train the AI and have it be more specific to our customers.

Marc Bernstein: 22:06

So you just gave a great lesson, I think, in how to pick an AI company, because you're setting up. Basically, you know moral boundaries, in a sense. You know to avoid bias and present data as data, which may or may not be the case with all the companies out there, I would imagine.

Logan Dopp: 22:27

Yeah, we lean very heavily into you know number one, that the data from our customers is helping to enhance the AI, so it makes it inherently customizable and tailored and tuned to you. And then also we present the information in, you know, without bias. Just this is what the AI saw and allow the users to make them what do I do about it decisions. And I think that's a really good handoff for how humans and machines work together in the automation space and now in the AI spaces. It has to start with how does a human do their job, how do they intuit and analyze their situation and how can we just best present them that data for them to then make decisions on? And AI can help speed all of that up. But we never see a scenario where AI is supposed to replace a human.

Ang Onorato: 23:13

Right. So Logan with that, give somebody you know. Give our listeners an idea. If they are at a company and the company you know works with Torrent AI, what's the day in the life like in the manufacturing plant for them? What is your technology do? What would that day in the life look like when it's implemented?

Logan Dopp: 23:31

So I mean at its core.

Logan Dopp: 23:32

Typically, video cameras are streaming the feed to some sort of DVR or location.

Logan Dopp: 23:38

You can view your video and it's probably being recorded for 90 days, and they can go back and look at things if they need to audit, or they can pay someone to look at the video feed full time and detect, you know, safety violations or slowdowns in the production line.

Logan Dopp: 23:52

What Torrent AI does is it allows all of those video feeds to have events and objects extracted out of it and put into a location where it becomes searchable. So then the users don't need to be watching the video feed 24-7. They can simply go and search for the types of objects or events that are of interest or set alerts that they can be made aware of those events in order to have the human go and make a decision on what to do, and that information and data then becomes retained, you know, indefinitely if they want. So you can always go back and look at how are things changing on the manufacturing floor? Am I seeing less potential safety violations month over month? Am I seeing product come out line at or above capacity? And let the AI automatically extract those things out, so that you don't have to go stare at video feeds anymore.

Ang Onorato: 24:43

Wow that's. I mean, it's amazing because I think you're perking the ears up. We have a lot of folks that we know listen and work in manufacturing, listen to the show and you know I think you're touching on such important parts of it actually also makes the case of why humans are actually so needed. They're just needed in perhaps a different capacity and more of that strategic critical thinking and maybe, you know, not as much always in sort of the manual piece of that. So it's about retooling and reskilling versus replacing you need the humans for that strategic part.

Logan Dopp: 25:14

And a lot of this comes from my background, in our team's background, that you know, when I was working in counterterrorism security like I was the dude that was staring at video feeds looking for threats, and I know the pain of having to try and pay attention to eight, nine, ten or fifty video feeds and I know how hard it is to go search through those things. And so we've made it so that it's easier to triage what's happening and we're not trying to say, oh, the AI is now replacing a watch person or a health and safety inspector or, you know, the production manager, but it's allowing them to do their job more efficiently and be more strategic about what they do.

Marc Bernstein: 25:52

Hey, one thing I wanted to make sure we covered and we only have about a minute and a half left. But we've talked a little bit about gaps in AI, we've talked about hallucinations, we've talked about false AI. How do you have any tips on how to spot false AI? You have mentioned a little bit about it, but any any sort of thoughts to wrap that up Sure, yeah, I mean I would look for customization.

Logan Dopp: 26:17

I would look for things that are actively learning. I think a huge challenge in AI is if you don't have a way to customize to your environment or update that training rapidly. Even if the AI is good out the box, it's going to become stale because it's not learning on the new data. So I would I would be keen to look for things that can be customized and retrained and updated and have that baked into what they do, rather than just trying to say, you know, oh, this is chat GPT, slapped onto a post-it note, which there's a lot of those. And also look for ways that the technology is applying the AI and some of the works for the humans and isn't trying to come in and say ignore what we're doing, let the AI make the decision for you. I'd be concerned with those kinds of things because like we said, there are some gaps.

Marc Bernstein: 27:06

Yeah, that's really, that's really good advice, so good that I wrote it down because we're, we're looking at that ourselves. Ange, did you have anything?

Ang Onorato: 27:17

Yeah, I mean, I think this is just, it's just such the tip of the iceberg, right. I mean there's so much and I think you know we'll love to probably even have you back Logan. Let's say, like three months from now it'll be like dog years, right, you probably have three or four more years to talk about.

Ang Onorato: 27:33

You know, and I think we're just be. We'd love to have you come back because I'm curious to hear you know some clients that you have what their experience is, let's call it, six to 12 months down the road, what's the feedback from clients. You know, I think that'll be really telling to kind of really give some some real facts back to the factual piece, to give you know real factual events, to say this is what AI has done in, in, you know, this industry or whatever. So we're excited to see that and hear that and really appreciate all your time today to record a couple shows for us. I think it's a public service to all our listeners as well. So thanks for being here.

Logan Dopp: 28:07

Yeah, I appreciate it as well. That's such an important conversation. So, you know, happy to be here and helping people understand, kind of what is AI, how might it be applied, and sort of opening people's horizons to you know, what could I use my cameras for beyond just security? You know what? What could I use with this data that there's so much becoming possible. It's exciting to see people like becoming educated on it and thinking for themselves of, like what do I want to do with AI?

Marc Bernstein: 28:34

It's been great. Logan, we are out of time, so thanks so much for being here, and thanks to all of you for being here with us today on Founders Forum, and we'll see you next week.