How to inject AI into your core business

Table of Contents:
- Every business should be asking: How is AI going to affect your core product?
- Rana el Kaliouby on the evolution of generative AI
- How Smart Eye experiments with different AI tools
- Why diversity of thought around AI is key
- The consequences of moving too fast with AI
- Why AI will need to consider our emotional states before giving advice
- Weighing the advantages and disadvantages of being an early adopter
Transcript:
How to inject AI into your core business
RANA el KALIOUBY: Even if your business is not AI-driven, thinking about how AI can accelerate or change or help you innovate on your core product is really key.
To me, the mindset is an experimentation mindset, exploration mindset, knowing that you’ll try things that may or may not work, knowing that you are an early adopter and there’s advantages to that, but there’s also disadvantages. I’m always asked this question like, when is AI gonna become mainstream? Like, should we wait? Is it like in three years? I’m like, no, it’s already mainstream. You gotta be trying it now.
BOB SAFIAN: That’s Rana el Kaliouby, Deputy CEO of Smart Eye, an AI-based pioneer. Smart Eye has been re-assessing its own products and processes, as the new wave of generative AI has rolled through.
I’m Bob Safian, former editor of Fast Company, founder of the Flux Group, and host of Masters of Scale: Rapid Response.
I wanted to talk to Rana to get a peek inside what operating with AI every day looks like for a business today.
Rana is an established AI expert, as an operator and as a thought leader. She essentially created the subfield of “emotional AI,” co-founding Affectiva, which spun out of MIT. And she’s an early-stage investor in some 30 AI-related startups.
Rana shares key lessons about fanning AI excitement, while making sure that ethical boundaries aren’t crossed. She offers specifics on how to begin experimenting with AI, even for those anxious about its implications.
Masters of Scale will be emphasizing AI’s impacts on business in our episodes, as we look to advance the choices made and actions taken in this fast-evolving moment.
Rana’s insights in this episode span both the enthusiastic and the cautionary. She argues that we need to work with these new tools, and in the process, to what she calls “rehumanize” ourselves, to make the most of what we’re creating.
[THEME MUSIC]
SAFIAN: I’m Bob Safian, and I’m here with Rana el Kaliouby, the deputy CEO of Smart Eye, and a pioneer of emotional AI. She’s co-chair of the Fortune Brainstorm AI Conference, general partner at AI focused venture firm AI Operations, and was a guest previously on Masters of Scale in 2021, talking about how to guide new technology. Rana, thanks for joining us.
el KALIOUBY: Thank you for having me, Bob.
SAFIAN: So you have worked with AI for many years, positioning it as a tool for humanizing our relationship with technology. With the explosive adoption of ChatGPT, AI is suddenly available to individuals on a broad level. Is that the kind of development you’ve been looking forward to, you’ve been fearing, or something else?
el KALIOUBY: It’s definitely a super exciting time to be in this space, and I think we’re at a moment in time where actually this idea of the IQ and the emotional intelligence, the EQ of AI becomes super important. So I’m generally excited, but I do see how this could really go wrong.
I’ve always been a very staunch proponent of the ethical development of AI. Some of the issues we’re seeing today around AI safety, bias, even hallucination, these are things that existed ten years ago. It’s just the speed at which it’s happening. And the implications, because it’s so much more massively scalable.
Every business should be asking: How is AI going to affect your core product?
SAFIAN: So you co-founded a business, Affectiva, that utilized AI to identify human emotion by reading people’s non-verbal signals. And now at Smart Eye, you’ve got an expanded portfolio of AI supported insights from industries, from automotive to marketing. You’re already an AI-based operation. So how does this new wave of AI impact Smart Eye’s business and your plans? Or is it just like, oh yeah, we’ve been doing this the whole time?
el KALIOUBY: So I started my company Affectiva out of MIT and right from the get-go, we’ve been deploying machine learning approaches to understand people’s emotions and capture these non-verbal signals, which by the way, constitute over 90% of how humans connect and communicate with one another — be it facial expressions, or hand and body gestures, or vocal intonations. So we’ve always been a machine learning shop. So when we first started, we were using a lot of feature engineering, and then when deep learning came along, we migrated to be deep learning based.
And then Smart Eye, which is the company that acquired Affectiva a couple of years ago, is also machine-learning based. So what we’re doing today is we’re asking the question, okay, AI is at an inflection point with all its generative AI advancements. The question is, how does it affect our core products? And I think every business should be asking this question: how is AI going to affect your core product, whether it’s a threat or opportunity? And then how can you leverage these new AI tools across other functions in your business?
And that spans everything from finance to operations, human resources, sales and marketing, because that’s going to allow the business as a whole to move faster and be more efficient.
Rana el Kaliouby on the evolution of generative AI
SAFIAN: So you’re an AI shop, machine learning based, then you leaned more into deep learning. And now the question is about generative AI. Can you try for the non-techies here to explain what the evolution is of those different approaches?
el KALIOUBY: AI is this umbrella field that is all about building different forms of intelligence. The approaches by which you do that is what I encapsulate under machine learning. The holy grail of machine learning is unsupervised learning, where the algorithm is able to learn on its own, like it doesn’t need a human to teach it. We’re not quite there yet, but we’re definitely getting there, but let’s backtrack. With deep learning, which in my books is just an example of a machine learning approach, you needed gobs and gobs and gobs of data. And the data had to be labeled to basically understand the difference between say, a smile and a frown. So we give it hundreds and thousands of examples of smiles and frowns. It basically looks for all the things that are common. And it learns those differences. So that’s deep learning.
With generative AI, large language models, it’s self supervised. It starts hiding pieces of the data that it has, and then it tries to predict it. And if it gets it wrong, it gets dinged. If it gets it right, it’s like, ooh, I learned something new. And it iterates and iterates. It’s very computationally expensive, but it doesn’t particularly need a human in the loop. And that’s what’s so powerful about these models. It gets us an inch closer to this idea of unsupervised learning.
How Smart Eye experiments with different AI tools
SAFIAN: So you’re there at Smart Eye, you guys have a business model, a tech stack, a way that you’re operating… generative AI becomes more and more available, affordable, right? So what are the conversations that you’re having now about those two areas that you talked about: your core business and then your sort of functions?
el KALIOUBY: So we have an AI task force, and again, I think this is something every business should be doing. We have an AI task force that’s cross-functional, and we meet on a weekly cadence and have these subgroups that are off exploring new tools, and they report back every week. We also disseminate all this information to the entire company. There’s about 350 of us.
So, as an example, we spend a lot of money and a lot of time collecting data of individuals driving around vehicles. Because that serves us training data for our algorithms. But now, we can leverage synthetic data. We can use tools, eventually like DALL-E and Mid Journey, to create synthetic people driving in a vehicle. And we could say, well make them look tired or ensure that they are ethnically diverse or like, gender diverse or age diverse. And you can do this in like seconds, right? So the cost of producing high quality data is totally reduced. So that becomes a competitive advantage too, because you can train these algorithms cheaper and faster time to market.
SAFIAN: You’re testing and experimenting in these areas or are you like switching over wholesale? Or you do like one thing at a time?
el KALIOUBY: Well, we’re experimenting with all these different tools and we’re reporting back. So, sometimes we have interns that are tasked with like looking into these tools. But increasingly we’re asking our main machine learning engineers to go off and, you know, really figure out these tools. And actually sometimes, like with synthetic data, we have our own synthetic data tool as well. So it’s a decision of build, partner or buy.
SAFIAN: For a lot of our listeners, AI is not already a part of their business, and they may find that sort of experimentation daunting to try to dip into. Is AI’s impact on a core product something every leader should consider, or sometimes should the focus first be on a company’s functions?
el KALIOUBY: I’m a member of the Young President’s organization. This global network of 30,000 CEOs around the world, and they’re all asking the same question. Even if your business is not AI-driven, thinking about how AI can accelerate or change or help you innovate on your core product is really key. Especially if you have unique data about your customers. If you have unique data that you’re collecting through your product, I think that would be worth asking some of these questions. If we look at other functions across the business … content generation is just a low hanging opportunity, right? So you can use tools like ChatGPT to create blogs, to even like brainstorm blog ideas. I don’t think the idea is to take exactly what it spits out and using that as the end product. You still need a human supervisor or human editor in the loop, but it’s definitely a helpful tool.
SAFIAN: Rana’s message here, about looking for opportunity without giving up control, is central in a generative AI context. As she told me at one point…
el KALIOUBY: There’s potential intellectual property issues, there’s potential confidentiality issues.
SAFIAN: After a quick break, we’ll dig in further on AI’s implications for businesses, and for each of our jobs, and Rana will give some examples of the kinds of folks at her firm who are finding new ways to thrive. Stay tuned!
[AD BREAK]
SAFIAN: We’re back! Before the break, we heard AI expert Rana el Kaliouby, the deputy CEO of SmartEye, talk about the difference between AI’s impact on a business’s core products and on its specific functions.
Now she talks about how AI will change the way all of us work, and gives examples from within her own firm on the kinds of people best positioned to thrive. Plus, she shares insights on AI’s evolving emotional intelligence, how we need to ‘re-humanize’ ourselves to get the most out of AI, and why this new tech could have an unprecedented impact on mental health.
Please note that the next segment includes references to suicide that some listeners may find distressing. If you or someone you know is facing mental health challenges, help is available.
Why diversity of thought around AI is key
Do you find the folks within Smart Eye are anxious at all about what the implications of this is gonna be for their jobs?
el KALIOUBY: I think AI is going to change the way we all do our jobs. Like for every job, every type of industry, your job’s gonna change with AI. And I really believe that the folks who figure out how to partner with AI to be more productive or more creative or more connected or more empathetic, are gonna do better and be more competitive and more marketable, and more kind of in demand than people who don’t. And I believe that at the individual level, and I believe that at the organizational level. So in general, I’ve been kind of encouraging my team to really lean into this.
It almost depends on the personality of certain individuals. So when we started this AI task force, I basically put it out there to the whole company. I was like, Hey, “I think we need to dive into this. There’s gonna be threats, there’s gonna be opportunities, who’s in?”
And it’s the people who want to take the time to learn about this. They’re the ones who raised their hands, and now they’re on this task force doing this. Some other people are extremely skeptical — they go to the negative first, like, okay, but look at all the ways this is gonna go wrong, which is important. The diversity of the thought and the perspective is really key. It really depends on the personality.
The consequences of moving too fast with AI
SAFIAN: You mentioned that you invest in early stage start-ups. There’s so much enthusiasm, I guess would be the nice word. I saw a news report of a company in France that raised over a hundred million dollars, and they’ve only been around for like less than two weeks. In that kind of environment, do you get worried about the point we were talking about initially about sort of the ethics and what the long-term implications are gonna be when folks are running so fast?
el KALIOUBY: So the good news is, yes, there’s a plethora of start-up companies springing up everywhere. But at the same time, what are we breaking while we’re moving so fast? And the story that comes to mind, this was not ChatGPT, but it was a similar kind of chatbot interface in the EU. This particular guy was Belgian.
He was suicidal, and he asked the bot to give him some ideas on how to commit suicide. And the first response was like, ah, you really shouldn’t do that. You should seek help. Then he asked the question in a slightly different way, and that tricked the bot. And the bot basically gave it five ways this guy could end his life. And he did. This whole text conversation with the bot is in the public domain now. And it’s just really concerning. I’m not concerned about the existential threat of AI and it’s gonna take over our jobs and, you know, it’ll be the end of humanity.
But I’m really, really concerned that we’re deploying these technologies as mental health coaches and life coaches and, you know, assistants. We need these tools to be a lot safer and we need to do that fast, because we’re deploying these technologies exponentially, but there’s very little guardrails.
Why AI will need to consider our emotional states before giving advice
SAFIAN: You’ve always been focused on how to humanize technology before it dehumanizes us. And the idea of emotional intelligence, like tech doesn’t have emotional intelligence. Will it, can it, is that something we shouldn’t even be trying to create, like, how is our relationship with technology gonna change as it gets into all of these areas where it hasn’t been before?
el KALIOUBY: For the last, you know, x number of years, we’ve been interfacing with technology on its own terms. And we’re getting to the point where actually the way we interface with technology is very similar to the way we interface with one another: through language, through natural conversation, through voice, through perception. And I think the final frontier is through empathy. These technologies — they’re deeply ingrained in every aspect of our lives. And actually, they’re taking on roles that are traditionally done by humans. And so we now have to think about the emotional and social aspects of these technologies.
It’s not just about the smarts, right? It’s not about giving a smart answer back. It has to be an empathetic answer. It has to be a thoughtful answer. It has to consider your emotional state before it gives you advice.
SAFIAN: In that sort of future world, my AI, my assistant or companion will respond to me differently if I’m in a good mood versus in a bad mood. It’ll read that, it’ll feel that?
el KALIOUBY: Yeah, I mean, imagine you’re driving your car, and if it can tell that, okay, there’s a traffic jam, and you look really stressed, and your meeting’s about to start in 10 minutes, it might say, you know, Bob, I’ve noticed you must be super stressed right now. I’m so sorry. Do you want me to like, send a text to your meeting and say you might be running a few minutes late, right? Or if you are in a chatty mode, you’re not in a rush… maybe it can start a conversation with you: what are you thinking for the summer?
And it can start kind of co-planning the trip with you. If you’re falling asleep at night while you’re driving, it can basically say, oh, I noticed that you’re very tired. Like, I’m gonna like, have a conversation with you here to try and wake you up. Do you want me to roll down the windows a bit? It can act as a companion. So we are definitely thinking about those applications in the context of a car or a mobility experience, but you can take that anywhere. You can think about that for any application.
SAFIAN: You told me that this AI moment forces us to rethink what makes us human. Can you explain that a little more?
el KALIOUBY: If these technologies have a lot of IQ, but they’re also exhibiting signs of being very creative and potentially empathetic, what makes us uniquely human? My theory at the moment is that it’s gonna force us to go back and put more value on authentic human connections and putting a lot of emphasis on our emotional experiences. One of my favorite studies from a few years ago now: they had PTSD patients divided into two groups and one group conversed with an actual human psychiatrist and the other conversed with sensei, which was a digital avatar, and they found that the patients were more forthcoming with the bot, because it was way more patient. It was available 24/7, you could talk to it any time, and it was non-judgmental. And I sit with that, I’m like, wow, what does that say about us as humans, right?
SAFIAN: Yeah, well, maybe we have to become better as humans so that people want to spend time with us as opposed to with their machines.
el KALIOUBY: Totally. I call it “re-human.” We need to rediscover, rediscover our humanity.
Weighing the advantages and disadvantages of being an early adopter
SAFIAN: The Fortune Brainstorm AI conference that you chair takes place in the fall. In previous years you’d be building the agenda by now, but with so much shifting, you’re waiting. If we don’t know what’s gonna be top of mind about AI just a few months out, for the business people who are listening to this, like, how do you go about making decisions? Is there a different mindset that we have to lean into when things are so unsettled, opportunistic, and scary all at the same time?
el KALIOUBY: To me, the mindset is like an innovation mindset, an experimentation mindset, exploration mindset, knowing that you’ll try things that may or may not work, knowing that you are an early adopter and, you know, and there’s advantages to that, but there’s also disadvantages. And that’s okay — knowing that you might make mistakes, right? I’m always asked this question like, when is AI gonna become mainstream? Like, should we wait? Is it like in three years? I’m like, no, it’s already mainstream. You gotta be trying it now.
SAFIAN: I do think there are a lot of people who are hoping they can wait and have other people’s mistakes be things that they can learn from. But I don’t need to be the first mover here. And so let me just wait for everything to shake out. And if I’m hearing you right, your encouragement is to try not to fall into that because then you won’t be in a position to take advantage when things do become clearer.
el KALIOUBY: Exactly. I think if you’re experimenting and you’re at the forefront of trying these technologies, that will put you at an advantage. I actually love Peter Diamandis’ quote. So he says, “there will be two kinds of companies at the end of this decade: those who are fully utilizing AI and those who will be out of business.”
SAFIAN: Those are pretty stark choices for all of us, right?
el KALIOUBY: Exactly. But it’s also, like, super exciting. It’s a time of discovery. My internal algorithm is definitely all for embracing this, but in a thoughtful way, and holding ourselves to higher standards in terms of the ethics of it. And also, my big thing is, if you look around, unfortunately, a lot of these AI tools are developed by the same types of people. I believe AI can elevate all of humanity, but for that to be true, we have to really ensure that not just the consumers of AI, but the builders of AI are diverse human beings, in their experiences and their points of views and perspectives. I’m thinking about how to do that, how to do that as an investor, how to do that as a thought leader, and just make sure we’re changing the face of AI.
SAFIAN: Well, Rana, this has been great. Thank you so much for doing this.
el KALIOUBY: Thank you for having me.
SAFIAN: Listening to Rana, I want to double-click on what she calls our “internal algorithm”
when it comes to AI.
There are biases that we’re introducing into AI models, but there are also biases we bring in assessing AI’s potential, both positive and negative. And we need to be mindful of both.
A key takeaway for me is how the most impactful AI tools, for businesses and for society,
will engage us emotionally. And we need to be thoughtful about our own emotions, and our own humanity, in building these tools and in using them.
I’m Bob Safian. Thanks for listening.