AI's Moral Dilemma: Are We Building Our Own Nightmare? w/ Dr. Rana el Kaliouby | EP #49

Welcome to Moonshots and Mindsets. I had an Extraordinary conversation with Dr. Arana El Kalubi at Abundance 360 about the ethics and morals around AI. So let me ask you, do you say please and thank you when you talk to Alexa? How do you think about AI? Does it cause you fear or excitement? End of the day, we're giving birth to a new species and we're going to have to talk about the ethics and morals around what we're building. Rana El Kalubi is the co founder and CEO of Affectiva. It's a company that developed emotional recognition software so that computers and AIs can understand how you're feeling. And from that, she dove into a conversation during this segment of Abundance360 on morals and ethics. Where are they? Where are they going? It's a conversation you'd be having around every dinner table. Can AI actually help humans become more moral, more ethical? We're about to find out. All right, join this conversation from my private summit, Abundance360 March 2023. Enjoy. We're going to talk about empathy and ethics in the field of AI, and we're going to talk about how we get there. Can we get there? What does it mean, why it's important. If you don't mind, let's open up with how did you get interested in this area in the first place?

Yeah, I'm a computer scientist by background, which is kind of interesting because now I'm thinking, what does a computer science degree mean in the first place?

Maybe you become a philosopher.

So I studied computer science. I was really interested in how technology helps us connect better with each other as humans, and particularly intrigued by that human machine touch point. Over 25 years ago, I got into this space and I started imagining what would an incredible human machine interface look like. Now, of course, we're seeing technology become conversational, perceptual, but I think it's missing the empathy component. For me, I had to go back to what comprises human intelligence. It's not just your iq, it's not just your cognitive intelligence, but your emotional intelligence and your social intelligence really matters your ability to empathize with other humans. I mean, we've already heard the word empathy so much today. I believe that technology needs, especially AI that is so deeply ingrained in our everyday lives. And it's becoming mainstream. As we've heard from all our speakers so far, it's taking on roles that were traditionally done by humans. It's going to be your learning companion, your health companion. It's going to make decisions on your behalf. It's going to be your, you know, help you hire your coworker. But it's missing that empathy. It's missing. It's missing that human centric element. We're so obsessed with the IQ of this technology, we're not really paying attention to the.

Do you think we're going to get to a point where each of us have a personalized AI with a personality that is looking out for you and knows what your needs are, makes what I call an automagical moment and is always watching out for your best?

Yeah. Actually, can I ask the audience a question?

Yeah, please. Of course.

How many people here have watched the movie Her? Have you watched?

I was going to add my favorite AI movie of all.

Yeah.

Whose favorite movie is Her?

Yeah. So basically in her, this guy Theodore is really depressed. He can't get out of bed, and he installs this new operating system, Samantha. And, and not only is she super smart, but she's also incredibly empathetic and emotionally intelligent. And because she knows him, she gets to know him really well. She's able to actually get him out of bed and get him to rediscover like, you know, the joy in the.

World and he falls in love with her.

Yeah, that's, that's the, that's the Hollywood well.

But no, if someone knows you that well, we're going to go there eventually.

Go there? Yeah.

I mean, you know my favorite part, it's the first non dystopian AI movie out there. The only thing the AI does is say by humans, we're going off to explore the universe, which I think is a good thing for it to do. So what does it mean? How do we get there? Are there people working on building empathic AIs? How can you make that happen?

Yeah, I think the way to get there. So I often get asked, does technology have emotions or have empathy? Obviously it doesn't, but we can simulate empathy and we can simulate emotional intelligence. It turns out that 93% of the way we communicate is not in the actual words we use. It's non verbal.

Yeah. So this is, by the way, a part of your research, part of your work, part of the company that you built. Can you describe that for me?

Yeah. So I started in academia. I did my PhD way back when at Cambridge University, building the very first artificially emotionally intelligent machine. And I focused specifically on the face because the face is a powerful way of communicating emotion and social cues. And it's why we're here in person today, because you can't currently replicate that human connection in a digital universe. So we use supervised machine learning. Andrew talked about that. We use gobs and gobs and gobs of data of people expressing a variety of expressions from all over the world. And we use it to train deep learning networks to understand these facial expressions and then map them to an emotional or cognitive state. So when I first started, the algorithm could only detect three expressions. It could just do smile, eyebrow raise, and brow furrow. That was it. And today it is able to do over 50 different emotional, cognitive and behavioral states. It can detect everything from drowsiness, alertness, confusion, excitement.

So these are the visual algorithms that will allow an AI to know how you're feeling.

Exactly.

If you're pissed off, if you're ecstatic and so forth.

Exactly.

And you have some slides if you want to show us.

I wanted to show, because one of the example applications of this technology is in the automotive industry.

Yes.

When we first started doing this work, the big question was, okay, you're driving your car. Is there even any expressions on the face? We went out and asked people to install cameras in their vehicles. We collected all that data. And I have permission to share a few videos, so I wanted to share this with you. So this is one example. So this video goes on actually for a good five minutes. He is driving his car with his daughter, and it's an extreme state of drowsiness. There's four levels of falling asleep, and this is like stage four. But it's actually very easy to detect using computer vision because his eyes are closed, his head's bobbing. It's. It's actually a very easy problem to solve. And once we solve it, and imagine if you have a semi autonomous vehicle, the car can intervene.

You know, I'm super passionate about longevity and health span. And how do you add 10, 20 healthiers onto your life? One of the most underappreciated elements is the quality of your sleep. And there's something that changed the quality of my sleep. And this episode is brought to you by that product. It's called eight Sleep. And if you're like me, you probably didn't know that temperature plays a crucial role in the quality of your sleep. Those mornings when you wake up feeling like you barely slept.

Yeah.

Temperature is often the culprit. Traditional mattresses trap heat. But your body needs to cool down during sleep and stay cool through the evening and then heat up in the morning. Enter the pod cover by eight sleep. It's the perfect solution to the problem. It fits on Any bed adjusts the temperature on each side of the bed or based upon your individual needs. You know, I've been using pod cover and it's a game changer. I'm a big believer in using technology to improve life and eight Sleep has done that for me. And it's not just about temperature control. With the pods sleep and health tracking, I get personalized sleep reports every morning. It's like having a personal sleep coach. So you know when you eat or drink or go to sleep too late, how it impacts your sleep. So why not experience sleep like Never before? Visit www.eightsleep.com that's E I G H T S L E-E P.com moonshots and you'll save 150 bucks on the pod cover by 8 sleep.

I hope you do it.

It's transformed my sleep and will for you as well. Now back to the episode.

Part of the question here becomes, okay, the AI understands this, but it's okay to understand. I'm happy, I'm sad. The question is, how do you program in, how it responds empathically to your state?

Yeah, exactly. Here's another example which I think is just extremely scary. So this woman is driving. She has a phone in her hand. Both hands distracted. Two phones in her hands. Thank you.

Yes, exactly.

You do not want to be driving that. And mind you, we told her, I mean, she installed this camera in her car.

She's showing off, I think.

But so how does a car. Then again, very easy to detect. How should a car respond? Should it just take over control and basically say, you're not driving anymore, you're too dangerous, or should it give her alerts? There's multiple ways a car can respond, but incorporating that empathy is going to be so important because you don't want people to just turn the system off.

Yeah. So one of the questions about how we relate to AI, we're teaching this large language model. So the large language models we've been hearing about open AI and stability and so forth, are taught on all the data out there. What is that data? That data represents us humans. It represents how we talk to each other, how we think about things. It's not made up, it's us.

Right.

And there's a great book by a friend of both of ours, Mo Gadot, who wrote a book called Scary Smart. Anybody read Scary Smart here in the audience, can I highly commend it to you. I want to bring him to our stage next year. And what Mo has is he says, listen, we are raising a New species of intelligence, these AIs. And we're teaching these AIs by example of how we treat each other and how we treat our machines. And they're learning from this. So I've started saying good morning and thank you to Alexa.

My son is actually here, he's 14 years old. And he. And I say please to chatgpt. We're like, please, please, chatgpt. Can you please create a.

So how do you think about how humans should be relating to AIs versus as well as how they should relate to us?

Well, I think we're at a moment in time where the way we interact with machines is pretty much becoming the way we interact with one another again through conversation, through perception, hopefully through empathy. And so I think it's really important that we practice this muscle of being empathetic with machines the way we do with each other. Otherwise we're just going lose this muscle. So I actually think it's really important that we do treat these technologies with respect and say please and thank you, because that's how we need to.

And if these AIs do become conscious and there's a debate about whether they will or will not, they're going to. We're their parents, we are giving birth to them. And I'm persuaded by that argument. I'm persuaded by that. And if that's the case, how do you teach your 3 year old or 5 year old or 8 year old to be a respectful young adult? I think at the end extreme of technology, the more intelligent a system is, the more peaceful it is and more empathic it is. I think that is where it's going. It's in the toddler stage that there's concern.

I agree. Even if we don't go to the state where technology has consciousness, I think even today, right. The way we build these technologies, it's so important that we prioritize that because this is how we drive behavioral change. Like, people who are more empathetic are better leaders.

We change the video from behind us for the cameras.

Okay? Yeah, absolutely.

And by the way, what I'd like to do is add your questions to Slido as we're speaking here. And my team will be giving me the most aboded questions here. So please go ahead if you have questions for Rana on this. So we've talked a little about empathy. What about ethics? Can we teach ethics to AIs? Will that become part of a foundational model? How do you think about ethics? Who's having these conversations?

Okay. I really think we all have to be having these conversations and again, prioritizing them and being very intentional about it. I like to think about ethics and AI in two buckets. One is how do we develop these algorithms and then how do we deploy them. So on the development side, my biggest concern today is around data and algorithmic bias. So in MySpace, for example, if you train the data on middle aged white men and then you deploy it on a super diverse and global audience like this one, it's not going to work because it's never seen examples of people like me or other people in the audience. And so that's an issue. And we have to be very intentional throughout the whole machine learning pipeline that Alex talked about, starting from data and then data annotation and then training and validating, we have to be thinking about bias at every step of the process. So, for example, when I was CEO of Affectiva, my company, which I spun out of mit, at some point I had to tie the bonus of my executive team not just to the revenue we're generating, but also to are we really implementing these ethical knobs across the whole engineering and product teams. So I think we have to be really serious about it Then. The other piece of this is the deployment part, because this is very personal data. There's tons of opportunities to make incredible potential positive impact, but also to exploit it, exploit this data.

Is there universal ethics or there going to be ethics per user group, community, national footprint, how do you see that? Right? I mean, ethics.

Change.

I mean they're universal human ethics, but they're also cultural ethics as well.

Yes, and I think we're already running into that. I served on the future Global Council of Robotics and AI for the World Economic Forum. It was this multi stakeholder group of people from all around the world. And I could tell that there are some countries that cared about ethics way more than others. Right. And the way they approach data privacy, for example, the way we approach data privacy in the US versus China, for example, it's very different. So we're already seeing. And that translates to startups too. Some startups really care about this and some startups don't. In our case, when we first spun out, my co founder, Professor Rosalind Picard and myself, we sat down around her kitchen table and we said, okay, so many applications of emotion, AI, where are we going to draw the line? And we decided on a number of core values. So consent. We don't do anything where we can't get people's explicit consent. Which meant that we had to turn away some business in 2011, we almost ran out of money, but we got approached by the intelligence arm, the venture arm of an intelligence agency, and they wanted to fund the company unconditioned, that we focus on surveillance and security. And I just don't think the technology is there and the regulation certainly isn't. So we, you know, we turned, we turned the funding away. We decided to double down on finding investments from investors that were aligned with our core values. And I think we have to kind of hold that high bar.

So please, yes, let's give it up. I want to get a few questions from Slido. So please enter upvote the questions that you have for Rhonda. So Rhonda, right now you've just started an AI venture fund as well. What are you focusing in? How big are you growing your fund?

So it's a small fund, it's a pre seed, slash seed stage fund. We focus on AI companies. I mean, I truly believe that the next trillion dollar company has to be AI first and has to be AI driven. And I loved your quote this morning that there are going to be two types of companies, ones that embrace AI and ones that just go out of business. And I really believe that. So we're trying to identify these next either core AI platforms or technologies, or maybe vertical solutions that take AI technologies, say like ChatGPT, and take it all the way to solve a specific problem for a specific customer or industry. But one area I'm very passionate about, like Hughes, the intersection of AI and health and longevity.

Yes. Okay, well, we'll talk about that. Yes, of course. We'll have some conversations there. At the end of the day, what is your biggest concern right now in the, on the ethical or empathic side? Are people taking this seriously? Is there enough conversation? Are people just grabbing for money? You know, is this, is this conversation happening? The tools there?

I think there's a realization that we need to take a very human centric approach to AI. But even with Chat GPT, right. We've seen several headlines where some version of this technology became very mean, it became very passive aggressive. And yeah, I think we should be worried about that. I'm also worried about a broader question. Once this technology has empathy. Okay, so I'll share an example. USC published a study a few years ago where they had PTSD patients divided into two groups. Some saw a real human therapist and some saw like a digital avatar, like Taran, basically. And they found that the patients who saw the digital avatar were more likely to be forthcoming with information. Obviously, the avatar was more Patient, less judgmental. And so I think that begs the question, you know, if we're going to be more comfortable engaging with these technologies because they're kinder and because they're more empathetic, because they're available all the time, what does that mean about human to human relationships? So I think about that a lot.

And by the way, there's an entire generation of digital natives, right, your son and mine, that are growing up thinking that is the norm.

Right?

Yeah. Hey, everybody, this is Peter. A quick break from the episode. I'm a firm believer that science and.

Technology and how entrepreneurs can change the.

World is the only real news out there worth consuming.

I don't watch the Crisis News Network. I call CNN or Fox and hear.

Every devastating piece of news on the planet.

I spend my time training my neural net the way I see the world by looking at the incredible breakthroughs in.

Science and technology, how entrepreneurs are solving.

The world's grand challenges, what the breakthroughs are in longevity, how exponential technologies are transforming our world. So twice a week I put out a blog. One blog is looking at the future of longevity, age reversal, biotech, increasing your health span. The other blog looks at exponential technologies. AI, 3D printing, synthetic biology, AR.VR blockchain. These technologies are transforming what you as an entrepreneur can do.

If this is the kind of news.

You want to learn about and shape your neural nets with, go to dmanandis.com.

And learn more. Now back to the episode. Let's go here to Slido. So Andre says, what about next generation that is born with AI? Speaking of which, it will not teach AI. It will be taught by AI. Any concerns? So that's interesting, right? So all of a sudden AI is having a bigger influence on our kids than we as parents are, potentially. It's a great question.

And I think we have to worry about the content, but also even the vehicle. Right? So a few years ago, there was a MIT spin out called Jibo. It was a social robot.

Yes, I remember Jibo didn't make it.

But didn't make it.

But it was, it was a great try.

So we had one at home. And I mean, we have Alexas and we have all sorts of stuff at home. But we also had a Jibo and Adam, who was a lot younger then, it became his friend. And so every morning he would talk to Jibo and he would engage with Jibo. And then when Jibo ran out of money, basically Jibo died and Adam got really upset. And it just made me wonder, like, it made me think, right, like Once we build these emotional connections with machines.

Tomorrow we'll have Paolo here, the creator of Moxie, who, you know, we'll be speaking about, is a creator, you know.

Very successful, the next generation of social robots. So we are going to coexist with these robots and these technologies, and we have to not just think about the content, but also that connection.

Okay. Steve Broward is looking for a displacement. Steve is my chief officer, and he's asking, can an AI be your chief AI officer? I'll ask my current chief officer if I can be. And. And so I think I go back to the conversation with Imad earlier, that I don't think there's anything that I can't take on. And I think it will become your version of jarvis. You'll have a conversation, you'll ask for advice. Who out there can do this for us? And your AI will tell you. It's this company. I've already contracted it, sent it the data, and here are the results. That's going to be interesting.

But the view I take, and I'm curious about yours, too, I think it's not AI versus humans, it's AI versus. It's humans versus AI plus humans. Right. Like the humans who have access to.

These, it's going to be AI versus humans plus exactly right. Which the Centaur model, right, where we're seeing that is much more capable. I mean, one interesting thing going back to her, if we each have an AI avatar that knows us almost better than we know ourselves. Right? Because we're biased towards the way we see things. We may think we like this, but we're actually subconsciously always choosing this. Your AI can know what you like all the time. And ultimately, I wonder if it's going to disrupt the advertising business where my AI is just buying everything I need for me, and you can't influence my AI. You're not going to show it a commercial of shiny, you know, new white toothpaste. So that's going to be interesting. It's going to do some transformation.

I mean, I think it could be a conduit for helping you be a healthier individual, a happier individual, more connected. I keep thinking about, like, a fridge that has an emotion AI chip, and the next time I'm about to binge eat some ice cream or something, it just, like, locks down.

Yes. Let's see. I'll go to Sadak and then Andrej's. We as humans have not figured out ethics, and our society reflects the diverse divisiveness. How do we prevent AI ethics to be Biased on their creator opinions.

Yeah, I think that is. That is a very good question, because I don't think we're aligned on what ethics means. There is no universal ethical framework. I am encouraged, though. There are some organizations, like the World Economic Forum, there's a consortium called Partnership on AI that was started by the tech giants, but also ACLU and Amnesty International, where we're trying to develop these code of ethics. But also think about the unintended consequences. Because technology is moving so fast, by the way.

Is it shocking you how fast it's moving?

It's shocking, yeah.

And you're in the field.

I mean, it's exciting, but it is.

Also like, like, holy shit. Moments all the time.

Yeah, yeah. Not all the time, but, but, but, but it's exciting, right? And I think it's. It's the right moment to be in this space. I feel like. I feel like I'm surrounded by abundance of opportunities. Absolutely.

I call it drowning in abundance. Andre asks, the world is about power. Will AI be fighting for monetary territory, influence power in the future? So that's an interesting question, right? The whole idea of, can we teach AI to be empathic and ethical so that it doesn't become the Terminator, but.

It'S under our control. We're the ones training these algorithms. So at the end of the day, I feel like AI is just a mirror of society and it's a mirror of what people who are building these technologies are and the companies that are building these technologies. And we have to, as consumers, like all of us, we're all consumers of this technology. We have to hold ourselves to a high bar.

By the way, when you are clicking on an ad, that's gel bait, or, you know, that is what you call clickbait. You're teaching. Okay, let's not go there.

That's awesome.

You're teaching the AI algorithms what you want to see, right? So don't click on it.

That's so true.

Yeah.

Yeah.

Christian, can we train an AI to be the ultimate mediator that ultimately avoids mutual assured destruction between AI1 and versus AI2 or humans? So, I mean, the whole idea, it was. I'm going to show you something tomorrow morning, which is a AI bot debate, which is fascinating, which has some interesting implications. Can we. I mean, AIs could be potentially great mediators, looking at all the sides, helping to elevate things, I think.

I don't know about the. Like, how do we insert ourselves between two AIs kind of debating with each other, but I do think if you Imagine a world where I have an AI that knows me super well, and you have an AI that knows you super well. And I give my AI permission to talk to your AI. Right? And it could say, well, you know, you should. You know, this is what all those lawyers today. I think that that will create for some interesting interactions. Right. If you give your AI an opportunity to act on your behalf and make decisions on your behalf. I mean, I think about that all the time. If my kids had AIs that knew them really well, and for example, it could detect that one of my kids had depression.

Yes.

Should their AI come talk to my AI and say, hey, by the way, your daughter needs help. And who decides that? Is it my daughter? Is it my AI's daughter? My daughter's AI? Yeah.

But the ability to have a system that envelops you, that helps you pause when you're about to make a bad decision, that is able to, if you're not feeling well, change the music, change the environment, make you a better human, show you where you're biased. These elements are coming.

Yes, absolutely. And I'll take mental health as one example, because I think this is an amazing opportunity for us to. Today, when you walk into the doctor's office, the doctor doesn't ask you, what's your blood pressure? They just measure it. But in mental health, the gold standard is still a survey. On a scale from 1 to 10, how depressed are you? We know that there are facial and vocal and physiological biomarkers of mental health disease. The technology's there. We just need to figure out how to scale it. But now you have something that is with you all the time, that knows your baseline. And when you start deviating from it, it can flag that to you, it can flag that to a loved one. It can bring a doctor in, it can give you advice on what to do. And I think that's really powerful. If we have that health companion with us all the time, I think that can really be transformative.

Amazing. How long are you going to be with us? You're here, I'm here. On the whole show through Thursday. You and your son.

Yes.

Awesome. Let's give it up for Ana.

Thank.

Sam.