On the first episode of the AI For All Podcast, Martin Cordsmeier and Max Weidemann from millionways join Ryan Chacon and Neil Sahota to discuss emotionally-intelligent AI, artificial empathy, and AI ethics. They talk about the applications of emotionally-intelligent AI, whether AI is more empathetic than the average person, AI in mental health, using AI to understand one's self, and whether AI should be able to make decisions autonomously.

About Martin Cordsmeier

Martin Cordsmeier is the co-founder and CEO of millionways, which began as a non-profit organization. As millionways grew, it pivoted to for-profit, and Martin moved to the US with his co-founder. Martin published a book about his belief that mental health starts with understanding yourself and your unconscious traits and patterns.

Interested in connecting with Martin? Reach out on LinkedIn!

About Max Weidemann

Max Weidemann is the co-founder and CTO of millionways. After graduating from Goethe University in Frankfurt, Germany with a Master's degree in Mathematics, he became a self-taught full-stack developer primarily interested in Applied Machine Learning and AI. After co-founding the start-up, millionways, with Martin in Germany, he moved to the US to further develop, scale, and market their technology.

Interested in connecting with Max? Reach out on LinkedIn!

About millionways

millionways brings emotional intelligence to AI. millionways' technology can be used to give any platform the ability to learn its users' unique personality traits, hopes, fears, and ambitions and to ultimately create a deeply personalized user experience, with explainable outputs and more nuanced and empathetic than the typical AI black box.

Key Questions and Topics from This Episode:

(00:44) Introduction to Martin, Max, and millionways
(01:04) What is emotionally-intelligent AI?
(05:05) Artificial empathy and psychological AI
(08:17) Applications of emotionally-intelligent AI
(09:44) Is AI more empathetic than humans?
(13:56) How is AI helping with mental health?
(16:09) The evolution of emotionally-intelligent AI
(18:35) Domain expertise
(23:18) Using AI to understand yourself
(25:36) Will humans ever have to empathize with AI?
(28:16) Danger of AI that has only one goal
(29:28) Do humans need to believe the AI understands them?
(31:49) Ethics of personal AI
(34:29) Should AI make decisions?
(40:24) AI ethics, politics, and negotiation
(42:37) Learn more about millionways

Transcript

- [Ryan] Welcome everyone to the AI For All Podcast. I'm Ryan Chacon. I'm here with my co-host Neil Sahota and our producer Nikolai.

- [Neil] Hey, how's everybody doing?

- [Nikolai] Hello.

- [Ryan] Alright, on today's episode, we are focused on emotionally intelligent AI. To discuss this, we brought on two fantastic experts, Martin and Max, the CEO and CTO of millionways.

Their company is focused on building the world's first emotionally intelligent AI. Martin and Max, welcome to the show.

- [Martin] Thanks, Ryan. Great to be here. Thanks for having us.

- [Ryan] Excited to have you both. So let's start this off by having- let's start our conversation off by talking about what it means for AI to be emotionally intelligent. So Martin and Max, why don't you all dive in and tell us what does that mean for the layman, for the person out there that's trying to understand how can AI be emotionally intelligent?

- [Martin] Yeah, I can usually start from the high level perspective. I'm the non-tech guy here. Max is our CTO, so I always have the more personal perspective because I work on this topic, like understanding people in a holistic way basically all my life and before AI popped up, it was really like a, yeah, like a psychological topic.

People had to read books or go to whatever to figure out who they really are and hopefully figure that out in the end, which is still a hard thing. For me, this topic is one of the most important topics in the world actually because if you understand yourself better, it eventually leads to mental health, and that's what it's all about.

So this is my high level perspective and in the current AI hype that we are seeing, it's mainly about generative AI. Max can talk more about the differences, but generative AI means it's generative, it generates things like videos, texts, whatever, but it doesn't understand anything.

And nobody would actually- if you understand AI, then everybody knows it, but the masses obviously can't know that, so it looks like it's very, very dangerous or something. But in reality, for example, ChatGPT just creates words in a very, very good way, obviously, but that's it. And doesn't understand the meaning behind it and especially the personality behind it.

And we developed this actually long before the hype started a couple of years. Scientific background. I moved from Germany, as you probably hear. We had a research phase there and 25 scientists working on that. So really, really solid work on the scientific basis foundation and this is what we brought to the US now to scale, but I can hand over to Max because he has a tech perspective.

- [Max] Yeah. So basically the question of what it means for an AI to become emotionally intelligent. We have to go back a little bit in time. NLP, national language processing has been around for a long time now. It recently got more attention with all those generative models of course, but we've always been familiar with auto correct, sentiment analysis, speech recognition. So all of these models have been around for a long time. And we used NLP the first time to actually classify text based on psychological patterns. So what that means is we took a holistic model, a holistic personality theory that's been researched for decades in psychology and validated by science, and we basically taught the NLP model to classify text based on these psychological patterns. And this is how you can actually get to emotional intelligence because the main difference, ChatGPT obviously understands your prompt because it will use that prompt as an input to generate the next best word, and then it will do the recurrent thing where it puts the new word back into the model and generate the next best word based on that. But the intermediary thing of really parsing that text and understanding the underlying emotions or sentiment or psychological patterns, this is what's still missing.

And we believe that this might be the next very big holy grail in AI because this is actually understanding human intent, emotion, and personality, rather than just generating words based on the user's prompts.

- [Neil] Almost 10 years ago, this whole field of like artificial empathy emerged in AI. It really started saying could AI detect emotion in people. Could it figure out that someone's sad or happy or angry and they're looking at maybe body language, they're looking at word choice, like neural linguistics.

The work that you're doing at millionways, like where does it like fit into that space? Like where does emotional intelligence now come into the equation?

- [Max] Our approach was always psychology based. Empathy and psychology are closely related. So what we have done is we have identified linguistic patterns. So how a person speaks, what words they choose, what language they choose in their everyday life, and how they basically talk in their everyday life, is very highly correlated to one's personality. This has been figured out even before AI already in the field of psychology. And we use these insights to digitize it and use modern AI technology to basically make this scalable. Before that you had to do assessments with mentors or basically take personality tests verified by psychologists, but this now enables people to identify their personality only based on their language.

There is probably more to the equation than just language. We're also exploring audio, facial expressions at some point. But the reason why we started with language was because there was already all of this scientifically validated research on correlating language to personality, and this is where we got started.

But in the long run, I believe we will have more parameters that lead to identifying empathy even better.

- [Martin] One difference, what Neil just said about the and when it popped up in AI, usually they use big data even for the voice analysis. It's just big data models and you have some kind of statistics, probability calculations, but it's usually not based on real science. Even if you use some kind of MBTI or something like that, this is not really science. I don't want to be negative, but it's just a model to classify people with four letters, which is fun, but it's not really helpful if you really want to understand yourself better because you are obviously not either I or E, for example, you are much more, yeah, in between. And this is something that actually doesn't really exist. There are some people working on that obviously, but we actually worked on that for a very long time, and it's based on science from day one. This is a difference. It's not only better, it was also like others are much faster in development, but the good thing is we did it one time and now it's there and it's up and running and we don't need to do it again because the AI actually learned the psychological model that we are using.

- [Ryan] Let me ask you from a general consumer standpoint, obviously you mentioned the four letters like Myers-Briggs, but how would this- how do you envision this being- this model, this information, this ability for AI to be emotionally intelligent, to understand people, to be used? Like where- right now a lot of people are interacting with ChatGPT, and we've talked about that, but where are these- where is the application of emotionally intelligent AI most applicable, or where will that come in the most to the daily lives of people listening to this?

- [Martin] From my perspective, it's actually the fundamental thing. Like if you are in school, if you're in college, you always have to understand yourself. If you do dating, if you want to find the right friends or partners for your company, whatever it is, you always need to understand your especially unconscious patterns and traits and wishes and goals and dreams and whatever it is.

Some people are better with self-reflection. Some people struggle with that. But even if you are very good, I have self-reflected a lot in my life, and I still didn't figure it out for a long time. So this is actually the holy grail to understand yourself and what you really want.

I think it's- this is a fundamental thing actually. It seems like a nice to have, but it's not, from my perspective it's very, very fundamental, and people have to understand themselves better to be happy.

- [Neil] Martin, I gotta jump in here and I'm gonna ask the big question, and I'll answer it myself afterwards, but do you think that AI is better at emotional intelligence and empathy than the average person?

- [Martin] Yeah I definitely want to give that to Max, but just a quick answer. It's different. If we claim that we have real empathy, that's not a legit claim because the human brain obviously works differently. So we shouldn't say that it is empathy, but it can simulate empathy, or it can do the same but in a different way.

So I wouldn't say it's better or worse. It's just different. It's an additional objective layer to our human empathy. This is my answer, but Max, what do you think?

- [Max] Yeah, in our case, I can add just in our specific case, we have tested it, and it has been more accurate than a single psychologist who would've labeled something different or would've made more mistakes. Eventually our model can also become and outperform a group of psychologists, so it can really learn this theory and become better and more accurate and obviously faster, that's the main point, than trained psychologists. But still it is based on this holistic theory which has something to do with empathy because understanding personality patterns and understanding the psychology behind someone leads to empathy. But as- I would agree with Martin that the claim that we have empathy itself, human empathy, is way too early to make, but it can replicate empathy in a psychological sense.

- [Ryan] Alright, Neil. Your turn. You gotta answer it now.

- [Neil] I promised I would. So my own experience is, and we benchmarked this, is that I think, martin and Max put it really well, that the AI tends to be more accurate. We've rationalized that the reason it does that is because it has that laser like attention on you. Because even if you're a therapist or psychologist or somebody else and you're talking to somebody, you might be thinking about other things, right?

What are the kids up to? What do I do for dinner? What- who's that right now over there? Whereas the AI has this- all it is is totally focused on you, what you're saying, what you're doing, and we don't have that kind of attention, and so it can pick up on those more subtle details better than we can.

The add-on to that, and this is the surprising thing we found out with the chatbots in the early days, is people are more honest talking to an AI system. Like they'll tell things to AI that they're hesitant to tell like their doctor of 20 years or a police officer when they wanna file a police report because they feel like the AI is actually not judging them. So the- sometimes it's like the AI is actually getting better data as a result.

- [Martin] It's actually interesting, right? And also people post differently on social media than in their personal diaries. Why is that? There is a reason for that, and it's probably good, but still I agree. We also have the we- when we started with that, we actually do voice recordings for the initial test. We can also do chat, but we like the voice recordings because people are more honest, exactly as you said. And they are sitting in the room, and we were really thinking do people like sitting in a room talking to a phone about their most intimate thoughts? Is that really a thing?

I didn't know that. And then we tried it and exactly as you said, they're actually very comfortable. At the beginning, sometimes it feels like a little strange. But you get used to it within minutes and then people are very honest. Yeah. And we were actually touched on to see that because there's a big need.

We all know that we just want to be ourselves. It's just- it sounds very simple, but it's obviously not and this is why an AI can help because it's bias free, I know Max will say it's not 100%, it can never be 100% bias free, but you feel less biased because it doesn't judge you exactly as you said, Neil. Yeah, I totally agree.

- [Neil] You guys have alluded a lot to like mental health, so like how is this actually being applied? How is AI helping with mental health issues?

- [Martin] So for me it's definitely the first thing you need for being healthy, like mentally healthy, is understanding yourself and your needs and your wishes and everything. And once you have that, you are on the way to mental health. This is my personal experience. And I'm working on that since I was a teenager because I saw so many people struggling with that and including myself obviously.

Everybody knows somebody. And I think it's more about understanding yourself.

- [Max] That was also what our main focus in our research was. Identifying the two steps basically. We identified two major principles in how you- we can lead to better mental health in general. First of which is what Martin already said. Self-reflection, so understanding yourself better, your passion, your motivation, your interests, what you're good at, what you really want to do in life. And then secondly, and this is also where our AI model comes in as an additional helpful resource, secondly, you need to be connected to the right people. We obviously- we can emulate a lot of things right now with all this generative AI, but the human connection and the human interaction is still very, very substantial in leading to good mental health.

And you will always need a good mentor to reach your next career goals, or you will always need a good friend to fill some of your gaps in your personal life, or you will need a good partner that fits you. So all of these things are also very crucial and our model also helps there because we not only identify personality, but we also- we have some matching theories on which personalities work well together, be it in private life or in the job context. And this is also something where we believe AI can be very, very helpful going forward in bringing people together again.

- [Ryan] It's really interesting to think about this application and the value and benefits it has long term for people. Because obviously just having that outlet, having that ability to communicate with something that is- learns more about you, learns who you are, helps you understand who you are, is something that I think has really been shown to provide value for people who do struggle when it comes to the mental health side of things. And I'm curious to hear your thoughts about what does the kind of evolution of this look like when it comes to the application of it for people? So if you think about, and again, this is just, this is from a movie I saw a little while ago, if you've ever seen the movie Megan, where they build this AI robot to be able to connect and interact with the daughter of one of the scientists.

And it had me thinking that if you have something like this you're able to interact with on a daily basis, you're able to build relationships. I know it's with- it's not with a human, but you're able to build relationships, have an outlet, provide- get feedback, have something that you can engage with as opposed to feeling completely siloed and alone and having to maybe do that on your own, which is not technically super easy.

So how do you see this evolving from where it is now into the daily lives of the people or of people as this continues to get more sophisticated, and it gets proven out even further?

- [Martin] Yeah, that's a great question. It's always funny. If we do some predictions now, then we will be one of these people in 20 years where everybody's laughing about, yeah, they- but right now I think the very next steps are not making an AI as empathetic as us, but adding, again it's a different kind of empathy and understanding human behavior because it's not a human.

And this is good. And Megan is a good example. I love this movie, by the way, so this is- it got a little out of hand in this movie maybe, yeah. But this is- it proves how AIs are different because of the primary goal of Megan in this case is to protect or help or assist this girl, then she will do everything, no matter what, to do that. And this is not very human, obviously. But it's different. So this is why I think it should be used differently and not- we shouldn't- why should- why do we want another human in AI? We already have humans, so I don't see a big point in that. We- I think it should be different. That's my perspective.

- [Neil] I'm gonna run with the Megan thing for a moment because I think it's actually a great example of some of the current challenges we're facing that you have a lot of really smart technologists working on, like AI tools, but they don't really understand the domain. And spoiler alert, for those of you who haven't watched the movie, they have some smart technologists, but none of them are actually parents, let alone understand the science of child rearing, right? And that became a sticking point in the movie about some of the things that happened. I won't ruin that part. I think it goes back to a point Max you made earlier about that you still need the domain experts, you still need that expertise. How do you actually balance that out? Then how do you find that right integration between the two?

- [Max] We've also always worked closely with humans, not only psychologists, but also humans who understand empathy and who use their empathy in interviews. So I think if we teach these principles, not only the science, but also the intuitive approach to empathy that we as humans have, we can really get somewhere. But I also- I totally agree with Martin on this. I think it doesn't make sense to build another human entity or human-like entity. It's more about having kind of an objective solution where you're fully aware that this is only a bot and it's only technology, but it knows your personality objectively, and it can give you objective feedback. Because I also see myself sometimes, you're always obviously gathering your feedback from the people close- you're close with, but they're all biased. So once you have this- another entity, which is not really a close human, or you don't have a close human-like relationship with, but you know that it knows you, and it understands your patterns and can give you feedback on an ongoing basis, you might ask some personal questions just to give- just to get another third party perspective based on this expertise that you trust this entity has.

And then I think you can use this information to either change something about your behavior or again, get connected to other people because at the end, we should always focus on the human to human interaction and not build deep relationships with chatbots. That's my take.

- [Martin] I think I have a good example maybe because when in our research phase we talked to a lot of people about what they want to do in their lives. And one example is people want to open a restaurant and then the question is, why do you want to do that? Because maybe you know that you want to open a restaurant, but our AI can help you figuring out why and what exactly it is that makes you want that.

And then if you figure that out, maybe you don't want to open a restaurant anymore, but something else because you could have this wish because you love cooking, you could have this wish because you want to have- do some interior design because you just want to create the restaurant.

Or you love hosting guests. Very different motivations. And you don't exactly- usually people don't exactly know why they have a certain wish and then they are on a track and they only focus on that, but they could also get happy with something else much faster maybe by figuring out what the real intention is.

And this is a good example for how AI can help because it's objective and it's not a human, it can adjust, analyzes, for example, how you feel when you talk about hosting guests or how you feel when you talk about creating a- like the interior design. This is very different. And if we- if we have an assistant like that, and this is what we basically do, I don't want to just talk about our product, but I obviously have the most experience with that. So this is what we can do, and AI in the future will probably do more of that. And another example would be CarynAI, this virtual girlfriend. This is an example of people trying to replicate humans. And do you honestly want to have a relationship with this AI? I don't think so. I guess most people want to experiment with that. Want to have some troll sheds. And that's basically it. And of course you have a hype because it's a cool story.

But our AI was, for example, was able to identify the differences between the real Caryn and the AI. It's very different. Max could talk more about that if that's interesting, but it's not the same personality. It's just not because it's an AI.

- [Neil] Maybe that's more of a known quantity. I don't know. Maybe people are more comfortable with that.

- [Ryan] One of the things I was thinking about as you were talking about the the restaurant and being able to use it to understand if you really wanna open a restaurant or not. There are a lot of people out there that I think as they get out of high school, getting out of college, they're trying to figure out what they need to do- what they wanna do with their lives.

And I think a lot- it's a big dilemma for a lot of younger people, and a lot of the decisions they make when they're younger put them on a path that may or may not truly align with who they are or what they're looking to do, and I feel like there's applications here that could apply to helping people more quickly figure out based on past interactions and data that's collected by those interactions as to not just who you are, but what is it that you're best suited for, most likely going to find enjoyment in doing in for something like a career.

Has that been something you all thought about?

- [Martin] I could talk about that for hours because I actually published a book about that in Germany, and this was my topic before we were even an AI company because we had a research NGO, like a foundation nonprofit, and we wanted to do exactly that, like figuring out what people want and then help them get there.

And I can- I will not do it, don't worry about that, but I could talk for hours about that. So this is exactly what we can do. So for example, even in schools, we worked in a lot of schools, even with children like eight to 12 years old. But also with teenagers, especially teenagers, but even children. If you ask an eight year old or nine, 10 year old what he or she wants to do then you get some interesting ideas about what this person is.

And it, frankly- one great example that I always use is one guy that wanted to build, he was actually nine, and he wanted to build ships like with s h. I don't know my pronunciation. This is like not c h. So he wanted to build ships in his life, and we connected him with a former shipyard employee.

And seven years later they actually built ships together because he was 15 and did his internship in this shipyard. So this is a stereotype, like a nice story. It will not happen all the time, but it's a great story because we just connected two people. That's all we did because our AI figured out that they will probably get along well.

And this boy told his wish. That's it. We didn't do more. It's not magic. We just connected two people. Very easy. Nobody else does it.

- [Neil] I'm gonna flip this around for a second because we're talking a lot about how we can use this as a tool to help ourselves, and I'll lay a little context behind this question I'm gonna ask. When the use of ChatGPT started starbursting a few months ago, there was a reporter from Wired, I was talking to her, and she asked ChatGPT, "what emotions do you feel?" And it came back with things like info grade.

And she's like, "my God, is it sentient?" That kind of stuff, and I was explaining to her that's not really, it's trying to apply the concept of like jealousy to how an AI might view that, but do you ever envision the day where maybe we as humans would have to show great emotional intelligence in understanding an AI system?

- [Max] That's a fascinating question. I still think that right now, even with our approach, we will always teach AI to replicate certain human patterns or human behavioral patterns. And eventually also, yeah, replicate or simulate human emotions. But I always think that it will not lead to or probably not lead to sentience in the way that it can get to emotions from its own, and then you will have to understand as a human how to empathize with the machine. I think we're not there yet, and I also doubt that we will get to this full-blown sentience where they get completely individual in their decision making and their emotions.

But it's a great question. We also sometimes in the Caryn example, we already started looking into what is actually the personality of ChatGPT? Is it always the personality of the person that it's interacting with? Is it more of a holistic personality? Does it depend on what you're talking to ChatGPT about?

So it's definitely something that we are also looking into, and it's a very fascinating and important topic to also, yeah, supervise the personalities of those, of all of those chatbots and how they evolve and change.

- [Neil] That's fascinating because the holy grail is always like this digital assistant that knows yourself as well as you know yourself. And so if they have a personality, do they adopt your personality? Kinda like some pets do, right? They adopt some of the mannerisms of their human owners.

- [Martin] I think this is one of the things. We don't like to talk about dystopia or about utopia, but when we read this letter from Elon and the others like this is maybe one of the real dangers because if an AI, as I said with Megan, has only one focus, everything else is secondary, and this is a completely different perspective, and this could be dangerous. So there's an awesome German book. I just checked if it exists in English, but unfortunately it doesn't because it's really special. I never saw anything else about AI like that. But there is no English version unfortunately.

But this book is about an AI taking over, like this famous AI takeover, and they ask this AI what is the purpose of life or whatever. What is your purpose? And it says, my- the purpose is to exist and if an AI is at this point, that the main purpose is to exist, it will do everything to exist.

And what does this mean? So there are some dystopian scenarios, but again, I personally love to focus on the positive, but it is dangerous because they are different, because they are not the same.

- [Nikolai] Something I'm curious about is- what are your thoughts on people needing an AI to- like knowing that the AI actually understands them. So like how would- do you think that people care if the AI actually feels like they feel, because they know it's a machine. They know it doesn't feel anything, but is that a factor?

Do they- do people need to feel like it actually is empathizing or relating to them? Or is it just about their own experience. It doesn't matter what the machine is experiencing. It's about their experience of feeling understood. And that's what people really want. They want to feel like they're understood. What are your thoughts on that?

- [Martin] Exactly, yeah. This is an awesome question. I love it. This is like because I talked really to thousands of people in my life, and I don't want to emphasize that all the time, but the only recurring topic is always being yourself. And this is so- again, it sounds so simple, but it's so hard, especially in this world of opinions and noise and whatever and overwhelming stuff.

How do you hear the right voice, and you need some kind of reflection. And if you reflect with yourself, it doesn't really help. It's not really a reflection because it's yourself. If you reflect with friends, they are subjective. They are not objective, obviously. And they also struggle usually. Psychologists, they're also subjective. They- if you're lucky, you have a good one, and maybe it helps, maybe not. It's very- it's based on luck, right? And an AI could actually fill this gap. So I think- I really like this question because I think in the end they don't want to talk to an AI.

Of course not. It doesn't lead in somewhere. But what should it do? The only thing is that it could do is giving you insights about something you haven't seen before. But the dialogue itself shouldn't be the thing. This is a waste of time, right? If you talk to an AI, the only purpose is to figure something out or to get somewhere, and I hope that it's like that.

Otherwise, I personally don't really like that, but this is just my take.

- [Ryan] Alright, I'll go ahead and jump in here. I wanted ask you guys something because we've been talking a lot about a variety of different great topics and, but I think there's an underlying question a lot of people are gonna have when it comes to collecting that- personal things, right? It's the ethics. It's the ethical side of all this. And interacting with something on this kind of more personal level, how do you- how does that- how is that- this- how's that handled? How- where do the ethics come into this? What- how are they- to determine where the line is drawn and what can be done, what can't be done. What to do with this information, this data? Because let's say, for instance, a company is using a tool to understand their employees better or hire and talk to lots of people from a resource perspective.

Where do you- how do you balance the growth of and movement of the technology forward with that ethical line that is out there?

- [Max] Yeah. So I think the need for regulations is very imminent. We saw that with Sam Altman and Elon Musk obviously crying for this. So ethics should be a huge topic. The way I see it is always I, for example, I love using GitHub Copilot myself in my everyday life. It just helps you code more faster and more efficiently, and it auto completes your code.

But I definitely like this this name because in my opinion, AI should always be a copilot and nothing more. So it should not be the entity that ultimately makes the decision or that ultimately is responsible for a judgment or a text or even a diagnosis, a mental health diagnosis, which we are also separating ourselves clearly from because otherwise this is very dangerous.

Not only ethics wise, but also legally wise, who's responsible then for the ramifications of the diagnosis. So I believe it should always be a tool, a helpful tool in your toolkit to have this objective, opinion based on data and science and the model, but ultimately the decision should lie, should rest with a human.

You can use it as a copilot for your hiring decisions, making better hiring decisions or making mental health advices, but it should definitely stop at that point. And we have to make sure with the correct regulations in place that we don't exceed this point of authority for AI.

- [Neil] Do do you feel, Max, that it should be a copilot in every situation, or are there instances where we'd want an AI to make a decision?

- [Max] Yeah. People are already starting to make- to give the full responsibility to ChatGPT in some cases, for example, for content production or image generation, video generation, everything that's generated. But ultimately I think if we if we think about judgements that have a real impact, such as diagnosis or maybe even legal judgments or hiring judgments,

I think ultimately because AI will never be 100% accurate and also in some cases we have to add in human intuition to make the right decision ultimately. I think in those use cases, we should always keep it at the copilot level from my point of view.

- [Neil] So do you feel there are instances where an AI should make a decision, or it should be enabled or empowered to make a decision?

- [Max] I think in those situations where it counts, from my point of view, it should not be able to make any ultimate decision. But I'm curious to hear maybe other thoughts where it might make sense.

- [Neil] If you're monitoring a dam and you're having flash floods, hurricane level winds, or something, all that IoT data is for machine consumption, not human consumption. So you have an AI system monitoring the potential strength of that dam and it- the AI's thinking, hey, this thing could break any second now, do you want it to go and alert somebody and have some human take some action or should it be able to take action on its own to prevent the dam from breaking?

- [Max] I think in those use cases, we have to do some trial and error probably, but as soon as it makes some wrong decisions and it leads to casualties, we have the issue again, but in some of these cases it might make sense, but I was more referring to all those decisions that actually impact a human directly, like if they're hired or not, or-

- [Neil] I can put it this way. You're damned if you do, you're damned if you don't. Right? And you're right, everyone expects the AI to be perfect. That's never gonna happen. But if it makes a wrong decision about a diagnosis or a legal decision, people are like, that's not gonna work. But like in the dam case, if it's like well it alerted a human that took so long, the dam broke, people are like why didn't the AI stop it in the first place Right?

- [Martin] Humans obviously also make mistakes, and I read a lot about these medical devices. Neil, exactly what you said, this is a dam. Other example is surgeries even. AI will be better in- at surgeries in the future. So why shouldn't we use it? And if an AI makes a mistake, every doctor makes a mistake as well.

So the question is- also with a self-driving car, like Tesla, I always have these discussions, right? If a Tesla kills somebody, but on the same day, 50 people did the same, I don't- this is the ethical decision, and this is really hard to do because, and who is guilty then?

I- this is a hard thing. Is it Tesla? Probably not. Is it the person in the car? Probably not. So it's nobody, and maybe this is the ethical question, but also even with more easy topics like writing essays, right? If you- if everybody will write everything for school and college with ChatGPT in the future, then the whole thing will change.

Obviously the system has to change then because it doesn't make sense anymore. So this is- it will solve itself. This is what I wanted to say. This is not because if you continue like that forever, everybody writes with ChatGPT, the whole system doesn't change. So there- and sense anymore.

So there will be- there will be changes, and I think it's good, and I get it. I'm more positive than Max in this case, but it's just an opinion. I think it has to be like that because if AI is just better statistically, why should we still do it? There's no objective reason for it, right?

- [Ryan] I think the tricky thing comes in, to Neil's question, is it's such a situational thing that who's responsible for making the call of when the machine, when the computer, when the system can make the judgment call based on the data versus when a human needs to come in and approve or deny that thing from happening because obviously then it changes or then it brings in the question like you mentioned with Tesla, who's liable for what happens? And I think that's a big question that is gonna have to be thought about across many different industries, many different use cases, many different situations, and it's an interesting challenge to have to really think through because you could probably make the argument in many cases for either side.

- [Nikolai] Much of the legal system is based on responsibility. So it's very possible that with more and more powerful AI systems that this will have to change. You'll have to rethink the entire foundation of the legal system, which is based on the notion that you're accountable and responsible for pretty much everything because of freedom of will. But with an automated society, you'll almost- that's a society that's based more on determinism than free will. So I find that kind of interesting.

- [Neil] Well, we're not gonna resolve this on this podcast.

- [Ryan] You- wait, you're telling us, Neil, we can't solve this with the five of us on this call in 45 minutes.

- [Neil] You know, I hate to break it to you, Ryan, but we might need a slightly little more time.

- [Ryan] That'll be part two of our discussion with the- with Martin and Max. We'll cut- this will be posing the question, the next one will be here's the solution that's gonna solve everyone's problem.

- [Nikolai] I think one- one last point for me is that- because I brought up understanding before, and I think it's interesting that- so AI ethics, I think it's hard to sometimes distinguish between AI ethics and AI politics. So politics meaning are we having a philosophical conversation or are a bunch of sort of interests and parties just trying to get what they want out of the AI system.

And from there, it's just a matter again of everybody wants to be understood. They want to make sure the AI system is accounting for their interests. So, I'm wondering if maybe millionways can be part of the solution to this kind of problem of- well, okay, we'll just give everyone their own AI that understands you, and we're not all in the courtroom yelling at the AI system to represent them. It's like you get your own AI, and you're understood.

- [Martin] And everybody uses it differently, right? I think I like this idea. It is philosophical, as you said, even if we talk about the surgery example. Is it better to have a person to blame if a surgery failed? Is it better for the mental health? This is very philosophical, right? Because this is the only difference if you have an AI which is objectively better than, in the future, than a group of surgeons, maybe, I don't know. I'm just- it's just an example. Is it better for the mental health to have somebody to blame or not? Is this philosophical? It's very philosophical, and I agree about the personal AIs. Everybody should use it differently, and it will happen anyway.

So this is always- this is what I don't like in Europe, actually, I- we moved from Europe and these regulations is important as Max said, but you can also overplay it because it'll happen anyway.

- [Ryan] Fantastic. Gentlemen, thank you so much for taking the time. This has been an incredible conversation for one of our first AI For All podcasts. So appreciate all the participation and all the ideas and that- this'll be very thought provoking, I'm sure, for our audience. And before we wrap up, I just wanted to give a chance for you all to tell our audience where they can potentially learn more, follow up and engage with you all if they have any questions or stuff about this topic with regarding millionways and what you all are doing.

- [Martin] Yeah. Thanks so much, Ryan. Thanks again for having us and yeah, we are on millionways.me slash millionways.ai for the chatbot. That was a little confusing, so millionways.me is our main website and millionways.ai is our chatbot. It's basically ChatGPT with an additional layer of emotional intelligence.

Very interesting to see, and everybody can figure out their personality and compare yourself with others. But yeah, this is what I'm personally very passionate about, so thanks for having us.

- [Ryan] Absolutely. Yeah. Passion was very clearly seen through all the conversation. Neil, anything last- any last words from your end?

- [Neil] I think fantastic conversation. For all those of you, I would definitely some time play around with a chatbot and see how you can tap into emotional intelligence for your business.

- [Ryan] Martin and Max, thanks again. Excited to get this out to our audience. And then Neil and Nikolai, obviously always a pleasure to have you on here.

- [Max] Thank you so much, Ryan.