Welcome to Midwest 2023. I'm Rob Richardson and I'm your host. It's an honor to be here. Honored to also work with BSL Group who is teaming up with us to do this live podcast. As many of you know, if you watch the podcast Disruption, now we have conversations that talk about disrupting common areas and constructs that really challenge how we think about things. And there is not going to be any different With me as my guest is Javier Viana.
If you believe we can change the narrative, if you believe we can change our communities, if you believe we can change the outcomes, then we can change the world. I'm Rob Richardson. Welcome to Disruption Now. But yeah is no different from from the data that we have. It's just a matter of advancing it in a certain way and providing right outputs for that for a particular task. Right? But it would be, if I can explain it, this is a very common, humble, right. It's like a box.
It thinks inside, process them and it and then I'll just get an output. Okay. Out of it was cooking. Look who's cooking. That's good. That's the question. Yes. So it's usually well, you can host your A.I. in cloud, you can have it in your local piece. It's usually a computer. Write something, a program that an executable program, an algorithm. Right. That is that is running through all the different steps that you have there.
But it's just following the recipe, like all these little things that you do inside that that final rate. And it's what you put your passion is the pursuit of knowledge. Make sure it sounds like knowledge is shared emanated among humankind, which I share that and I believe we share the passion for innovation hearing of that knowledge. So it's I know for myself and for you, there's a lot of excitement about this new intrigue, about artificial intelligence, but there's also some concern.
So I'd like to talk to you about what you are doing with reasons. What problem is reason trying to solve. Yeah, So before I get to that, perhaps I can just give a brief overview. Yeah. What is reason or even before reason I would say what, what is the current state of it. Yeah. What are we facing. Because that would down locate a little bit for like right now we are, we are, we're facing a global emergency. I said this is a global emergency.
Yes. In terms of a yeah I think we are we are reaching a point where we are abusing of black box A.I.. We we call it it's opaque A.I. that we don't have logs. Yeah, it's like that. Definitely. It's not transparent. Like, it's it's basically something that we don't understand the way a certain like is becoming a bigger problem because the models we are using, the example you used earlier cooking we don't know would be what's cooking or what the materials are used to cook. Exactly.
And like how these ingredients are combined. It's a great dish at the end and it's beautiful. It tastes as you want and everything, but the way you reach that output, slightly concerning because there is a lot of mathematical operations behind the scenes and there's not a clear path to that.
And humans, we are having a hard time understanding or, you know, verify in these outputs because when you start implementing A.I. into important capabilities and you start deciding whether what what is the amount of oxygen you are administering to a certain patient in a hospital, in a in an intensive care unit using on a Yeah. And that patient dies for whatever reason. Are you you want to know why? You want to know why we're the inputs that triggered this result and how right and those things.
I think we are not taking care of them right now. Everybody's focused on performance, everyone's focus on how can you get the greatest, the most, the most output in the quickest manner. Yes, the most efficient manner. Yes. Not the most ethical manner. Not the most transparent manner. That's true. Right. And why do people I'm with. You mean interrupt you? No, no worries. How do we get people to care about that? Right. We we are in a we're in a society.
And I'm not disagreeing with you that it's important to I I'm with you. Right. We're having this conversation. But of course, we're in a society that believes in fast results, quick returns and more returns. Yeah. So how are we making the case that this is this is something that you should pay attention to? Like, how are we making that case given we know that balance, we know we know the place. What you said is absolutely true. Yeah. It's essential to understand what's going on.
But one way of well, when when you are using A.I. and you're starting to do inference with the A.I., you can you can have biased decision making a lot and not just decision making, but like the inference itself could be biased in a certain way. And what I'm trying to say is that if we have explainability in a system, we can detect bias. For example, on things like that, right? If we are not aware of what's happening and there's implicit,
you know, bad things going on. Yes, those are going to happen. You know, the the end user is going to suffer them. And we might not see that suffering right now in the short term. But as we implement it more and more and we started integrating the AI in every single day to the task, I see we're going to see ourselves.
We were starting to see it, you know, in and some time, you know, social media was about to say, I think we've see I mean to you I was getting right to my you must be must be on my way. Oh here life we have seen it like it's not we we have seen the destructive nature of what's an algorithm that we don't understand, that we don't know. Yeah. Can affect us. You know what? What's obviously new is that it's now exponentially able to produce things at a level.
So, you know, with social media alone, I've said this on my past podcast, it literally exacerbates the worst conditions of human nature sometime, right? Yeah, Yeah. The vanity of ourselves, the the the ability to be jealous, to divide amongst one another. Because what the algorithm is doing is very simple. With social media, their goal is not their goal isn't enlightenment. Their goal isn't to encourage us to get together in society, is not to inform is it's it's it is to get you engaged.
And so the goal is to get clicks and the goal is to trigger you. And the goal is to keep you coming. And they're going to repeat what you want to see. Which is it necessarily which was doesn't necessarily align with the truth, doesn't necessarily align with what's even best for you, but it keeps you entertained. Yeah. And people like that until they don't. And we've seen what what's happened with elections and things like that. So you're right.
So what other problems can you foresee not having transparent explain away. I well it's many of them right but especially when you start connecting together and like do you allow them to operate based on the outputs of the previews and keep building this chain? Right. This is called multipolarity. And you can you can phase an output that has been processed in a lot of different ways, but you have no absolutely no idea how these output like.
It's literally like a being we can think of a as a as a type of living creature that is is not it doesn't need oxygen like us. Right. It's like it's something that does a lot of researchers that believes that there's that level of consciousness already achieved in many models. But but the problem is that what I'm back up with. Back up. You said a lot there if I can. Yeah. And make sure we can explain this to a way.
The people that I need to understand like you've you've because I know you're deep into this research and I want make sure the audience understand. So I like to use analogies to make sure I understand. You talked about multiple morality, and I think that how I'm understanding it is basically you have a lot it's like raising a child, let's say, and you put a lot into that child.
And you we don't know if there's you're teaching that child bad habits and it's compounding on itself to create a new entity that can cause a lot of damage in the world. Right. And it seems like that can be a good analogy for what you're saying. Like we don't know how the data what's happening, and that's going to create this entity that's going to create more problems for us because we didn't look look at it, look at it on the front end to be transparent. Yes. Yes, that's correct.
It kind of Yeah, yeah, yeah. I I'm sure it's not a perfect analogy. No, no, no. But it explains it very well. So. So no, no. That we, we, we can agree that there's this big issue. Right. And we can see it more and more every day. Now, how do we solve it? That's a question, right? That's the question. How can we you ask the question. I mean, I'm interested there's not clear not a clear answer. And the reasons are various. One of them is like, you know, it's a very complex mathematical problem.
Sometimes the air is advancing very fast. You always need to keep up to date with the current technologies. But one way to solve it and the one we are working on at reason is trying to study what's going on inside a model of, you know, you have your air. We have think of it as the box. It has been trained. There's other people that are trying to create new boxes that are transparent from the ground up. But that's difficult to scale because everybody is using already.
Neural networks are very complex. So there's the one, there's the one solution. So it's like it's so important to understand the problem. You're trying to solve the problem. You're trying to solve is how do you make a more transparent? And you say the one way is people are trying to create new models with new models that are crazy. Yeah, not crazy, but whatever. Very challenging. It's challenge given that the infrastructure is already there, it's like trying to build a new nation.
That's what I did for my page. For example, a lot of Academy Academy people are working on that end as well, but it takes a lot. And industry hasn't adopted any of these books, right? Sometimes it takes a whole year to develop something. Yeah, it's like building it. I tell people like, could I do a blockchain? But as I said, there's similar stuff in that if you want to build a new blockchain, you got to build a whole new highway or a whole new nation. And that's that's a lot to start from.
But you're saying integrating with the current infrastructure. I think whatever it's already Pre-Trained Yeah, okay. And try to make it more transparent. Your technology would take current technology that's out there in A.I. and make it transparent. So let's say everyone's talking to that. Getty Of course, as if that's the only generative model there ever was.
It's that's, that's this way It is, because that was the first to get the consciousness of the public and made it in a user friendly way that theoretically would would reason be able to pierce into that the algorithm, if you will, to be able to tell the public this is actually what's going on behind the black box theoretically. Is that is that is that what you're looking at? That's the that's the ideal scenario.
We're going towards that where we work right now is on the smaller problems, the classical neural networks that have been used in a variety of applications that, you know, the medical use case that I was telling you about, prediction of oxygen, for example, and trying to explain what were the inputs that triggered this result and how each of the inputs contributed. And from there, you can tell a lot. You can tell whether, you know, one of the inputs was contributed too much.
Perhaps you were biasing the entire reasoning. You can tell whether these combination of inputs in this particular way was leading to a wrong outcome or a less confident outcome if that you can trust more the the the output if you understand that logic. Right. And and we well we go towards that direction of applying it to do better and a general way. But right now we are working on like more classical problems.
You have a big neural network, huge, sometimes millions of parameters that's doing a great job predicting for whatever you want right. But but you don't understand why those predictions are being made like that. And and you want to have that level of transparency. And that's becoming now a requirement in many in many places in Europe, for example. Well, the GDPR and the ACT are explain what that is. General data protection regulation.
It's a very popular regulatory framework in Europe for for practitioners because it's kind of making us design a Yeah, in a different way if that makes it more transparent. Whereas the US is like, do whatever you want, right? It is still it's taken a while, but I think there's going to be more regulations coming into play over the next year and we're going to see them.
I think they are necessary, right, Especially when we start, you know, using A.I. for for taking decisions that may affect human lives. That's absolutely right. So so yes. So that's the reason. Just hold that thought, because I want to just Georgia, I think when you think about why this is so important, this is not just entertainment. There's also reasons to not want to do false entertainment and things like that, because entertainment can lead to things.
But you're talking we're talking serious decisions. You know, we're we're going to use I know we we're going to have someone from Tallahassee on to talk about that. We use A.I. within the battlefield. We already do. And how are the how are we evaluating those decisions that A.I. is making? Is it just about what's the most efficient? It's not necessarily what's the most humane. Yeah. So these are things we have to grapple with. You're talking about medical uses we already use.
People don't know we are. We're already using we're using A.I.. A lot of people. We already have been much more than just be deciding whether you have credit or you get credit or not. For example, who gets credit or not? I'm based on what were the inputs, you know. Yes. How those inputs played. Like you said, sometimes they're biased. They are they are bias off the right. And I want to make a point here, if you don't mind. Sorry for interject. Oh, there's several types of bias.
You can have bias in the data, right? Like the data is biased in many ways in the world. And you can have great software engineers, future engineers working hard to make sure that it's clean and balanced and like perfectly leveled so that when you put the AI in place and you train the algorithm, it's the best quality data that you can have. You can do so much always, but then the learning of the itself because the learns, you have to train it.
At the very beginning, it's everything initialized randomly, but you trained all the time at the adult learning stage. You can implicitly extract bias or even if you have done a great job in the future engineering under cleaning, you can still generate biased reasonings because the learning happens very randomly and right now there there's really no way for us to understand whether your inference is being biased or not.
You have done a great job in the future engineering your data is clean, you know that. But now when you train your AI and you're making inference, you might not actually get it. You might get to a bias results. All right. Let me you're saying because I think often I think the case is that the data is actually is biased because people that enter the end or the data are biased. Right. So there's you know, one of the well, the hardware as well, the sensors.
I mean, I'm not talking about human bias, but that's what I'm trying to understand. So that's my question. I'm getting to You're saying you're assuming that the input has the people that they're actually entering the data and playing the data have taken precautions to make sure that it's not biased and and done things intentionally that way? You're saying even then after that's happened, you can still come up with a bias decision? Yes. Yes, exactly.
After the taking of the data, after the processing of the data and having a clean, you can still get bias in the AI. And that's because the learning, it's very random, very stochastic. If you train another A.I. later on the same data, you're going to have a different AI and it's going to have some difference. That difference is bias, you know, I mean, not directly, but but in a way you can you can start getting different outputs in different ways.
And one, I might be considering some things more another one, some other things differently, that different process is just like you and me. We are understanding whatever problem we are talking about in different ways. Sure, we can both give a similar answer. It's not going to be the same. We are taking into account different things in our brains. So that difference, that's the one I'm talking about. There's no way right now, no way for us to really see what side reasoning and that's what we do.
Reason So how do you how do you it's first of all, it's admirable work and I appreciate it. We are trying I don't know if we are. Yeah, well, I mean, there's not a lot of people working on this. No, there's not. People. We have to educate the people. That's what I say, right? So it's not a market like. Yeah, we were predicting, you know, this or that not we are trying to do that, but nobody is now it's it's important work.
But this is what I'm getting to right. It's you're going up against the current the current trajectory of how to commercialize this one on hormones like let's worry about let's worry about ethics. Let's worry about transparency after we made $10 billion. Right. Yeah. How do you balance against that? Like, really like what is your message against that?
Yeah, I think, you know, it's it's all part of what we want to leave to the next generation of people because when they are facing huge infrastructures that are already in big companies functioning very well and then you have the problem of transparency, bias or whatever and really replacing all that and like leaving them with that burden. I think it's a it's a big issue.
So, you know, yeah, it's going to stay for I mean, yeah, we're allies here like it and it's going to be in every single thing you can think of that has an electronic sort of like even labels these, these room is going to be controlled. The light intensity is going to be controlled by the conversation and measuring how well the conversation is going to maybe go lower when you know those kind of things. We're going to see it everywhere. It's just a matter of time.
So like being able to really trace that reasoning is essential. It can make a difference, like in because the way I like to think about it is if we are taking decisions based on an output that we don't understand how that processing happened, who's really making the decision, the human or the machine? That's so a couple of things on this. Like one, I've talked about this in the past, talked about it two years ago. I think it's important for us to continue to innovate.
But I think your your your your best case that you put forward in terms of people that are focused on the return is this if we don't build trust within AI, you can get a reaction in the opposite direction. And we're right where people there's a mass just backlash against innovation in AI that really puts us behind because we we live in a democracy.
And at the end of the day, if people decide that they want to elect leaders that say no at all, because we don't trust it, that can cause greater damage. So I would say building trust right now is is paramount to make sure that we can continue to align. And sometimes it's hard to tell people that you're building could be on fire until it gets on fire, which is the hardest challenge, right?
People, people like is not sexy to put out put out fires, but it's it's better people only only focus on putting out fires, as you say, versus actually preventing them. Yeah. And what you're trying to do is prevent it. And I guess we got a lot of people that really understand how big of a issue this could be. So I have another question with this. What does happen when the algorithms know us better than we know ourselves? Is that a good thing or is that a bad thing?
Well, I mean, it's it's just like any other technology, I guess it depends. If I may clarify a little more. The question, are you asking whether the algorithm is learning information about us that we don't know or or what what do you what do you question? I would just say if the algorithms actually know us better than we know ourselves, like like telling us like, you should go for a weekend in the I don't know, in a lake? Sure. Or like, advising us to do something.
We never thought of that maybe it's what we want, but have you seen. I'll give you examples. So let me give you a concrete example since we're going to go into a weird alternative Black Mirror universe. Okay, So there was this show on Netflix X and it was called I think it's called the One. I'm nervous. Okay, it's fine.
I was last on a weekend doing something I don't know, somehow found extra time that I never had and decided to go into what it was about is using essentially artificial intelligence and some type of data algorithm to say to help you find who your actual quote unquote mathematical one was. Right. And so you have people that are already with significant others in other things. And they got curious, they got on this and then they got them to leave the stuff. That's a wild thing.
But some of them were content before that. But when they went into this kind of like social media, people looked at like, Oh, I can look on here. And, you know, I look there's now there's now 10,000 options. It's not really 10,000 options, but you feels that way. That's what I'm saying. Is that a good thing for us, that the algorithm might know what motivates you, what triggers you, and it might do what social media done in an amplified way. Yeah, I think I understand the question.
Okay. So sorry. I was alone. No, no, no. It was. It was. It was right on point. But yeah, I'd like to clarify because I have a different name for this one. And so you're a recent searcher, I guess. Yeah. You figure out a way to go. Okay. No, seriously, I think it's, you know, it helps to have some sort of, like, automated decision making in your life, for sure. But I'm a very, very classical person, so I like to have some, you know, uncertainties every now and then.
And there are some things that you I don't know, I don't like automated, but but yeah, I guess it really depends on where you want to use the. Yeah, but I guess we are more concerned about A.I. being used for our own things, but it really is is in many other like processes that doesn't directly impact our lives maybe right away, like maybe on the guidance, navigation and control of a nerve. I don't know, like an airplane perhaps, you know, but.
But like, you have a lot of like, like maybe remaining just for life prediction of like different components that you have in your aircraft or like there's software out there, All the systems are starting to integrate AI And really the ones that you're talking about, I guess it's it's more like on, on, on our they might change over the course of our life I guess, right. Yeah I think they could. I mean small business. Yeah.
They are small ones now you know I say that dramatic example is, you know, they're like fining or whatever because I worry less about, you know, terminate people like a Terminator and stuff and Matrix. I don't think per se that's what it's going to turn into every time I have a conversation to explain why it always it's, yeah, of course it always does that. But I know better than to go there. I'm not going to have I'm not going to have a Terminator conversation.
I promise. Sorry. It's a I don't think that's the likely future. What is possible, though, is that because I'm seeing how social media has already got people used to it. I'm not anti social media. I'm not I'm pro social media, but I'm pro what I like to say I'm pro informed consent. What I believe is happens with a lot of social media is that we did not have informed consent, at least not for the part. No, not for the we do not know if that algorithm was which information. And Yeah.
So that that's, that's what we were trying to. Exactly. That's what I was trying to get to. Yes. I hopefully I mean that's also perhaps one of the main reasons why we are doing this is to raise awareness about like these technology and there's other people that are building other technologies that I believe can help advance transparency in.
A Yeah, but, but really what we are building on reason, I think is it's a step forward on that, like really giving the power to the user, to the final user, onto the, you know, and customer that they can understand how certain inference or predictions are being made. Right? Yeah. And I think that's very valuable that that's the value of Right. At the end of the day, you want to make sure how everything happened.
It's about, again, going back to informed consent, like I believe when we got on social media and everything else, it was originally to connect with our friends and to know what's going on with our friends and others. And it evolved into I mean, I guess it evolved in the media. I like to say, you know, even though I'm a media person, I try not to be a quick bait person. Maybe it's why I don't have 10 million followers. But it is it is what it is.
But I'll say that often media is to the mind what sugar is to the body. Oftentimes, particularly news. Right. And how we consume in news is to the mind what sugar is to the body. It's you get it, you get a hit form and it you know, it feels good emotionally for a second, but then it then it really drags you down. And that's what social media a lot has done with people and they don't realize it. Right. And and I think the challenge is how do we have informed consent?
I'm not saying get rid of social media. I'm not saying get rid of I we're saying together, how do we create a more transparent world where people understand the decisions that they're making and are truly given the autonomy? Because I feel as if you're saying, if we don't do this, we're going to have less autonomy and less understanding of the world that's happening around us. That is correct. I think. I think so, too. And it's it's a social right. We'll see where it goes.
I mean, with with all these integration of generative AI, awareness has has increased significantly and I think it's going to be going in that direction. Right. Everybody's going to request it. I guess a couple of rapid fire questions. Are you ready to wrap up? Yeah. What does the legacy of Harvey look like? Oh, my legacy. Oh, my. Yes. I don't know. I mean, I'm I mean, my late twenties. I don't know yet. My legacy. Well, that's why I want you to stop asking him,
I guess. I mean, my my biggest academic life has been focused on explainable A.I. working with Dr. Kelly Cohen and other people as well. But if if I can do anything that is meaningful, I believe it might be within the field of explainable. That's also why I'm trying to do reason and and all this. It just makes sense. Yeah. Okay. All right. Next question. If you had a theme, I don't know if I answered very well. No, it's fine. It's good. You've answered honestly.
If you had a theme that was the it's the story of your life or the saying, what would that scene or theme be? Do you have one that I really like? It's hard from it's of mine, but the world is like it's like a book. And those who do not travel only read one page. Oh, that's good. Yeah, that's good. So it's like a book. And those who don't travel is like reading one page. Yeah. Yeah, that's awesome.
I think it's really important to travel and really to get to know the different cultures and understanding how they work, how they how they socialize, how they do everything. And from there you can learn so much and get new ideas. Every time I travel, I get inspired on many of my because ideas have been a consequence of of a great trip. Oh, I completely agree.
You know, I tell people I spend a lot of my twenties traveling and and if I could go back, I wouldn't I wouldn't have split it any differently. Like, there's nothing more valuable than experiences. Like if you if you have $10 billion and you haven't traveled, you don't have any wealth, You just in fact, you are a prison of what you of your of your resources, because traveling is the most transformative thing. I agree. It is connecting with cultures. It's really what is why we do disrupt our two.
And it's because we want to connect with creators all across the world. Is this to me understanding learning cultures is just interesting. But to your point, when you're in a foreign environment, it forces your brain into another. Yeah, the brain is like. It's like it's the spine. It absorbs everything. Yeah. And if you don't get used to that plasticity, it's it starts to be more rigid. That's exactly right.
And it's funny because the people that I've met, at least in my in my life, that are very, you know, not flexible usually they haven't traveled that much. Yeah. And what you have to guard against and, you know, as we advance in age that we don't become so rigid, you know I guess to my podcast Rodney Williams, that this is the most important skill to his success was learning how to learn.
And you know, learning is a constant journey and the challenge when people gather a little bit of knowledge because none of us have a ton, even the smartest of people have a very small corner of knowledge. But when you get when you have success, you then become hardened to actually understanding new perspectives, which is why going to a new country humbles you. Because you realize this is something you don't know any of this. Yeah, you don't know any of the culture.
You don't understand any of the language, the history for all of your knowledge. You are a baby here and it's that humbling reminder when you stand in front of the ocean of how small you really are. Yeah. And when people lose that perspective, they lose their way. Even, you know, Albert Einstein, one of my one of my favorite people in terms learn from and obviously a genius. And he was humanitarian, but he didn't see the next evolution of physics. He actually rejected it. Right.
Even though he was the founder of it. He rejected that that was coming because, again, rigidity can go in the mind once you have something. So I think that's the challenge to everyone is to is to keep that perspective and and in a humble perspective when it comes to knowledge. All right. Yes. So like. All right. So a final question. Yeah. All right. You have three advisors for business in life. Who are these advisors and why? Paul McCartney, The first one. Who's that?
Paul McCartney. Okay. Why he is he's been so creative and really creativity to me. I think you build that ability to be creative. You can be, you know, born with, with more or less capabilities to be creative, but really to be humans are are excellent in creativity and it's really what's driving us and everything. So I think Paul has a huge amount of creativity and I, I, I really admire him for that. So yeah, I also like writing songs and music, so.
Oh, you do that. Okay, everybody music. But you know. All right, all right. We're going to we're going to do some music on this trip down to the platform. All right. We're going to have to one of your creative side. I'm going to get you to drop some music. All right? Now that I know this. Okay? I'll and yourself and everyone. But don't play it. So play the life of this. But. And then. I don't know. Yeah, my mother for sure. Okay. Yeah, My mother would be on my list, too.
Yeah. Yes. And well, well, well, that's three. That's three. You know, you don't have to. Doesn't mean you don't love the other people, you know? I mean, I'm sure it is. They're married. It's. It's their doctor call. It'd be on a list. There's lots of other people that have been helpful in and both of our lives like, yeah, if I started out as a harpist, you'll just you'll be there forever. I want you just to go with the with the top three.
Okay. Well, obviously, we all live in their apple, so you're good. You're good out here. It's been a pleasure. Pleasure is glad you come on. Thank you very much. Yeah, We're here at Midwest Con 2023. Rob Richardson with Disrupt Art You we're taping at the Digital futures building. It's been an honor to have Harvey Air Vienna his company is reason Make sure you check it out. Make sure you understand explain the way I and what it means. But until then, we'll see you next time. Thank you so much