What AI needs to focus on right now, with Reid Hoffman and Rana el Kaliouby

0:00Starting a business comes with its share of ups and downs, which is why staying true to your vision is essential. a non-negotiable for Romeo and Milka Bregali, Capital One business customers and co-owners of Ra's plant-based restaurant in New York. Romeo and Milka took a leap of faith when starting their own restaurant, gutting an empty space and building it from the ground up. Every pipe, every wall, every detail. But building from scratch came with a heavy financial burden, which is when they turned to their Capital One business card. With the flexibility of the card's no preset spending limit, they were able to spend more and earn more rewards while bringing their vision to life.

0:36Today, Raz's success is proof that with passion and the right support, it's possible to make your dreams a reality. Learn more at CapitalOne.com slash business cards. Pioneers of AI is made possible with support from Inflection AI. It's not just enterprise AI. It's your enterprise AI.

1:02Hi there, and welcome to Pioneers of AI. I'm Rana El-Khalyubi, an AI scientist, entrepreneur, investor, and now podcast host. Think of Pioneers of AI as your guide to all things artificial intelligence. We'll be here every Wednesday, bringing you conversations with some of the brightest makers and thinkers in the AI space. The episode you're about to hear is a conversation I had with Reid Hoffman, host of the podcast Masters of Scale. He's the co-founder of LinkedIn and Inflection AI, and also a dear friend and mentor of mine. You'll hear about my backstory and my take on why we need human-centric AI.

1:45Plus, you'll get a taste of what's in store for pioneers of AI. Alrighty, on to the show.

2:02Welcome, Rana El-Kaloubi, to Masters of Scale. I've been looking forward to this for a while. Me too. Hi, Reed. Let's do a little bit of the origin story. So say a little bit about, like, your path to becoming an AI academic, like where you grew up and how you chose your PhD and, you know, that kind of stuff? Well, let's see. So I was born in Egypt, as you know, grew up around the Middle East. Both my parents were technologists, so we were always exposed to like a lot of technology early on. I wanted to be an academic, so I decided to study computer science as an undergraduate. Got really fascinated, I still am, by this human-machine touchpoint, the human-machine interfaces and how it not only affects our communication with technology, but our relationship with one another too.

2:51So that became my area of focus. And I left Egypt to go do my PhD at Cambridge University in computer vision and machine learning over 25 years ago. So people think AI is like this new thing. I'm like, some of us have been doing this for a while and pioneered an area or a category within AI called emotion recognition or artificial emotional intelligence. Towards the end of my PhD, I met this MIT professor. She was visiting Cambridge University to give a talk and we clicked and she invited me to join her as a postdoc at MIT. And that's how I ended up in the US. What, you know, like one of the things that I think is also for people who want to go in deeper into your story is the memoir Girl Decoded, obviously a great, you know, clever title in lots of different ways.

3:44People way under-describe the EQ issues within AI. How much of that understanding, in addition to the kind of the path of your life, would people find from the memoir? Yeah, the memoir is this juxtaposition of my personal journey with the journey of building the technology because it's so intertwined. So I talk about like, okay, why do emotions matter, right? They influence every aspect of our lives, everything from our little decisions every day to big decisions we make, how we learn, how we connect with one another. And I'm sure our listeners will recognize that it's not just your IQ that matters, right?

4:24Your cognitive intelligence, but your emotional intelligence, your ability to tap into other people's emotional experiences. And that's how we motivate behavior change, influence people, persuade people to do things. So people who have higher EQs tend to be just better leaders and better partners. I do believe that's true for technology as well. So what's your view? What's your view on that? Well, I mean, one of the things that I think is funny is the people who say, emotions don't affect how I make decisions, or I think actually, in fact, it affects them a lot and they just don't have visibility.

4:59The question around, like, you know, what words you choose or how different words, you know, kind of affect you in different ways, even if they have largely similar semantic meaning, the choice of it. For example, you know, very classically, you might say, like, I think I might have this point of view versus I have this point of view. Well, very different emotional impacts. But that's not what people typically focus on. Rana believes that AI models need to be designed to pick up on such nuances, models that can understand the difference between a user saying, I think versus I know. She sees this as fundamental to AI's development.

5:41And while others have ignored emotion in AI, she's dived in headfirst. The company she co-founded, Affectiva, revolutionized emotion AI. Rana has always been a trailblazer. From academia to entrepreneurship, she's plowed her way in a field where she was often the only woman and the only Muslim in the room. You know, I used to wear the hijab. I wore it for 12 years. And I put it on willingly and I took it off when it just didn't feel like me anymore. I also moved to the U.S. with, you know, as a single mom with two young kids. They were 10 and 4 at the time, and now they're 21 and 15. And so I think I'd like to believe that I've had this entrepreneurial spirit in me.

6:27And when I have conviction, I just go for it. And I think one of my core values is having courage to chart a path that maybe hasn't been charted before. So when it came to Affectiva, so here I am at MIT. You know, I was funded by the National Science Foundation and Ross Picard's lab. But we started to get a lot of commercial interest in our technology, which is this facial expression recognition technology. And all these Fortune 500 companies wanted to use it for a variety of applications. And so I remember we walked into the Media Lab director's office at the time. It was Frank Moss. And we said, Frank, we need more dollars to, like, hire more PhD students.

7:06And he thought about it for a second. And he said, you know, this is not research anymore. There's a commercialization opportunity here. And my initial reaction was like, whoa, wait, I'm about to apply for faculty, you know, don't mess with my academic plans. But then I realized, and that was the tipping point for me, and I think that's my message to academics who are listening to this. entrepreneurship is an amazing path to actually bring your technology to scale, right? Like you take it from this academic setting where you're publishing papers and maybe 10 people are using it to like scaling it worldwide.

7:42And that to me is what's so magical about starting a business. That sense of taking insight and thought and research that you get from, you know, the academic side, then applying it to how do you build scale companies, scale technologies, other kinds of things? You know, I have to say it wasn't easy, by the way. I want to share a story. Please. Like it's more of a confession, really. You know, I'm a first-time founder. When I started Affectiva, it was my first time doing this. And I think I was really, you know, there was a lot of stuff I didn't know. So for example, we had one of the investors who was courting us at the time send us an email and he said, one line, send me your BS.

8:22And I was like, I have no idea what this guy's talking about. Turns out he was asking for our balance sheet, right? Although, very early startups, BS also means the other thing that people think. Right, exactly. Yeah, I was like, the only BS I know of, I can't really attach in an email. But then I was also like, our very first product roadmap, we were like, Q1, we're going to penetrate the advertising research market. Q2, we're going to do education. Q3, we'll be health. It was so naive and we had so underestimated how hard it would be. But I actually think that's okay because that gives you that conviction and that kind of belief that you can go for it is what keeps you going, I think.

9:08I don't know. Do you agree? I mean, I think a lot of the entrepreneurship stuff, you cannot fully prepare. There's no I prepare everything. You can prepare some, but you just have to have that chutzpah and grit and, you know, kind of very broad learning capabilities to do that kind of thing. Yeah, that's right. So go into a little bit more depth about both your research and Affectiva and what the shape of that has been. So the core technology we developed at Affectiva was essentially an emotion recognition system that tracked and identified your various facial expressions and then mapped those to any number of emotional states.

9:54And when we first started, it could only recognize a smile, an eyebrow raise, a brow furrow. But over time, it now has a repertoire of over 40 different emotional states. And the range of applications is numerous. We service about a third of the Fortune 500 companies globally to understand the emotional connection their consumers have with their products and videos and content. And then a number of years ago, we pivoted to the automotive market where we now build driver monitoring systems and cabin monitoring systems to really understand what's happening in the car and with the drivers. Is the driver attentive?

10:28Are they falling asleep? And think of it as being a co-pilot to both driver and also passengers in the vehicle. In 2021, Affectiva was acquired by a company called SmartEye. But the closing of one door meant the opening of the next. Rana has since launched her career as an investor, focusing on companies revolutionizing the field of AI. So I've been investing for the last three years. I started a fund with a friend of mine, Rob May. We've made 40 investments. And then I'm now partnering with Gabby Zeiderveld, who was part of my team at Affectiva. And we are launching Blue Tulip Ventures to invest in human-centric AI.

11:09So entrepreneurship again, but in a different form. I just love being in innovation kind of spaces. And I also love supporting founders on their journeys. I found from my experience that some of my investors were so instrumental to our journey at Affectiva, being thought partners, being strategic advisors, being connectors and door openers. And I want to do the same for founders of early stage companies. I also think it's an amazing time in AI. And actually, you're one of my inspirations and mentors when it comes to that. It's an amazing time to be investing in AI. It's still early days. I think the next generation AI companies and transformative companies of our generation are being born now.

11:56And how cool would it be to play a part in these companies' journeys? Now, part of what we're doing in this podcast is we're going to be introducing another podcast. Because in addition to your hats as academic, inventor, entrepreneur, CEO, investor, you're also becoming a podcaster. and part of what you're going to be doing is Pioneers of AI. So say a little bit about what the podcast is, why now? What chart, what path, what journey will the podcast be heading out on? Yeah, it's another kind of entrepreneurial journey, I would say. So Pioneers of AI is a guide for all things AI to people who are both experts in the AI space who kind of just struggle with keeping up with all the innovations and what's happening in AI.

12:52But ideally, one of our goals is to also expand the audience for AI and make AI more inclusive so that you can come to the podcast to understand AI, embrace AI, anticipate what's coming next with AI. We definitely want to take a balanced view on all the potential good that can come out of AI, but we also want to be underscoring and highlighting where things can go wrong so that hopefully as a society, we can kind of guard against that. We will be featuring AI builders, of course, so those in the kitchen kind of building the next generation of AI technologies and use cases, but also AI thinkers who are particularly thinking about the implications of AI on culture, on society, on our arts, on science, on education, on parenting, right?

13:36So we want to be tackling some of these like everyday questions as it relates to, okay, how is AI going to impact my life on a personal and professional level? I also am very passionate about platforming voices that we don't typically hear from, so that's going to be one of our goals as well. Pioneers of AI is a podcast that focuses on all the big questions around AI and the latest trends. Many technologists will say that the biggest trend in the forecast is agentic AI. That is, AI that can perform tasks for us without any assistance. It's an important topic and one that Rana is addressing head on in her podcast.

14:19The best AI agent, in my opinion, would be one that knows you really well. And one way to know you really well is, of course, to get access to all of your information and maybe whatever health device you're wearing on your bank account so that it can act on your behalf and whatnot. But another way it gets to really know you is through your emotional experiences. And what's your emotional state? Are you stressed? Are you happy? Are you anxious? so that it can really customize and personalize its interactions just the way an awesome friend would do. And I think that's still missing from where AI is at.

14:53And memory plays a big part of that. How do you build actual models of memory in getting to learn a person? Yeah, because part of how trust is established is that you believe that the person knows you, cares about you, understands what's kind of good for you and interests and so forth and will act in those ways. And memory, of course, is essential to all of that. Right. Right. And if it's not there, right, then the interaction suffers. Yeah. You know this as deeply as anyone in the world, but like obviously one of the things about this agentic future is going to be kind of establishing trust. So what do you think is going to be important, not just for gaining trust, but actually, in fact, being a faithful holder of trust?

15:45I would say really knowing what's happening to your data. First of all, understanding who's behind these agentic AIs, the company that's building these technologies, I think that needs to be trustworthy. Knowing where and how your data is being used, that's so important. So I think a certain level of transparency around that is important. And also, control over that. The relationship between a person and their AI agents is so important. Like how much autonomy do you want to give your AI agents? And does that build over time as you trust them more and more? So there's a lot of questions that we ought to be asking kind of as design decisions when we're building these AI agents.

16:30I'm so looking forward to Rana's podcast conversations on pioneers of AI. She's just getting started. And when we come back from a break, Rana takes the mic and asks me a few of her questions about AI. Stay with us.

16:56If you've spent any time building AI products or leading technical teams, you know this. Transformation doesn't fail because of ideas. It fails because teams can't move together. Enter Atlassian's teamwork collection. It has planning in Jira, documentation in Confluence, video updates in Loom, and now AI agents in Rovo, which connects the dots across your work so nothing gets lost. It's one AI-powered teamwork platform designed for how modern teams actually build. Learn more at Atlassian.com slash TeamChanger. That's A-T-L-A-S-S-I-A-N dot com slash TeamChanger.

17:46We're back with a special Masters of Scale introduction for Pioneers of AI with AI scientist, entrepreneur, and investor Rana El-Kalubi. You can find this interview on the Masters of Scale YouTube channel and more about the new show at PioneersOf.ai.

18:09We're going to switch hosts. And you're now the host for the second half of this podcast. So I am handing the mic over to you. So exciting. And I will start by saying you also have a new podcast joining our network. Share a bit about Possible. What's it about? So one of the things that I think is a great disservice to humanity that's happening across media, across government discussions and so forth, is a fear and dystopia approach to technology first versus a possibility and a creation. And it doesn't mean that it needs to be blind optimism or blind utopic because navigating how you get the good is really important.

19:03Like, for example, when you're driving somewhere, you don't get somewhere by going, I'm first going to figure out all the things that could go wrong before I get in the car. Like I'm going to go, well, I could spin out of control. I'm going to make sure I could hit a tree. You've got to start driving towards a location. And you have to have some place you're going. right you can't just go well if i eliminate all the bad then the good will happen that's not actually in fact how a journey and a creation and invention happens it's a going towards something so the real question is what should you shape towards and part of the reason i decided to add possible into the the pantheon so we have master scale for kind of scale companies and entrepreneurs and and journey and impact and then possible is okay what could possibly go right yeah of course But like, what should we be building towards?

19:54It's definitely positive, but it's positive in a, you know, how do you chart the journey? How do you navigate? And that's why I added it in to the Pantheon of podcasts. I love that because sometimes you need to create the possibility, like literally for people, and then you plant that idea or you create that image or picture of the future and then people can work towards it. But right now there is so much fear. So you talk a lot about AI unlocking human potential. What do you mean by that? So last year, I published the first book on AI, co-written with AI. My co-writer was GPD4. And what I wanted to do is I wanted to show where I understand why, you know, John McCarthy et al., you know, called it artificial intelligence, especially back then.

20:42But, you know, it actually now has a little bit of a misnomer. And so I called it, you know, amplification intelligence because it's an amplifier of human capabilities. And that doesn't mean that just like, you know, kind of other technologies, sometimes parts of it so outstrip human that the technology just does it itself. I mean, like, for example, we don't have humans plowing farm fields anymore. So it's like any bit of technology kind of moves this and it changes that landscape. But in the change in the landscape, part of, of course, the discourse has been kind of almost like the movie Idiocracy.

21:20Like, is AI going to be doing everything? And are we going to be sitting on the couches, you know, drinking Gatorade and watching numbing Aldous Huxley Brave New World, you know, becoming just kind of sensate, you know, bundles of emphatic, lymphatic response. And I just think that it's just not very likely. You have all these discussions about like, well, I could just, you know, fire my marketing department and use GBD-4. It's like, well, no, you can't. Right, exactly. Marketing is a competition between multiple organizations. And, of course, if you're not using GBD4 and other AI tools in your marketing today, you're falling behind.

22:07You're not using key tool sets. But it's human beings doing it and being amplified in various ways. And you should be looking at what are the ways in this human competition, just like, for example, I use really good running shoes or I use flippers or I use scuba tanks or I use da-da-da. what's the way that I can be deploying this technology to be furthering, you know, my human objectives, my humanist objectives, my business objectives, et cetera. And so that's what the unlocking human potential. And by the way, AI and the amplification it brings will have some of those challenges for people, full stop, and we need to help and be more sympathetic.

22:47But ultimately, better for your children, better for society, better for industry. And if you keep a learning and evolving mindset, it can be better for you too. So for some of our listeners who are listening to this conversation and they're still kind of weary of AI, what are some of the applications of AI that you're most excited about? Well, part of when I'm talking to government folks who are like, you know, press is telling me that my primary job is to limit big tech. And part of the problem, of course, is a lot of the driver of AI of this new industrial, this cognitive industrial evolution we're getting is emerging from big tech.

23:29And I say, well, actually, in fact, while this is what everyone's kind of singing the song they should be doing, I have an alternative vision, which is we have a line of sight. We could construct in a small number of months a medical assistant that is as capable as kind of call it, you know, an average GP or better, that could run for small number of dollars per hour. How do you get that to be made and available to everybody? And then, you know, everybody who has access to a smartphone. And then you have a medical assistant for the billions of people in the world who do not have access to a doctor.

24:11And by the way, even here in In the U.S., there are millions of people who do not really have access to a doctor other than an ER room, and they may not even have access to an ER room. So that's one, and that will require some modifications of liability laws, some incentives, and other things. The second one, which I also think is super important, is a tutor on every skill, every capability, every subject for every age. Infinitely patient. Again, runnable on a smartphone. Wealthy people have always been able to afford tutors for themselves or for their kids. Now it can be everybody. Yeah. I love that you've highlighted those two applications in particular.

25:00I'm passionate about both of them. It's this idea of democratizing access to quality health and quality education around the world. It's very possible, like right back to your new podcast, it is so possible, like the data's there. You can create these like co-pilots for health and career coaches, right? That can be kind of learning companions and career companions. It's so exciting. What are some of the things that concern you when it comes to AI? Look, it is definitely possible that bad things can happen. Now, one of the reasons why I kind of advocate very strongly against the existential risk, the robots are coming, Terminator movie, et cetera, is because most of the real bad things to focus on are AI in the hands of kind of bad human beings.

25:50Whether it's criminals, cybercrime, terrorists, a variety of things. Rogue states, you know, interference with elections, you know, all of this kind of stuff in the hands of human beings with bad intents, because we have those too, are the top concerns that I have. Now, you know, again, a little bit of the tech lash is people tend to say, well, you know, but what about like, for example, things like hallucination or, you know, perpetuating a racial stereotype? and shouldn't we not launch until there was zero of that? And I was like, well, I tend to have the same view that I have on most things, which is you can never get all these things in these complicated systems to zero.

26:37So the better thing is make sure that you've got baseline good and then be improving constantly on those. So you launch anyway. One of the things that I've helped set up with various 501c3s is hosted conversations across all the major tech developers to talk about, you know, what kind of things could go wrong and what are safety cases and what should be on the test safety thing. So, like, people who are obviously don't have visibility naturally get some, you know, kind of— Sensationalized, right? Yes. Right. And look, having low visibility, a fear response is not irrational, right? Because, like, what are you guys doing?

27:17Are you just trying to build an AI that's just kind of trying to manipulate me to buy things I don't know, I don't want? or get really angry so I spend a bunch of time on social media? Anyway, so those are all definitely in the concerns, but there's the concerns that you go, let's steer right now before we launch. And then what are the ones that are dynamic and iterating? And the list of dynamic and iterating is very long. And the list of prevent before it could ever launch is compact and fixed. I'm hearing you also say that there just needs to be a lot of intentionality about how we're building these technologies and kind of thinking ahead of time about what are the unintended consequences, like what could potentially go wrong, but not letting us stop our innovation for that.