Richard Yonck saw the emotional machines coming.
In “Heart of the Machine,” published in 2017, the futurist argued that computers would learn to read human feeling and respond in kind. Less than a decade later, millions of people are confiding in chatbots that answer with something that passes for empathy.
Yonck tells Independence Avenue Media that this fluency is precisely what makes the technology dangerous. Today’s systems, he says, can approximate what psychologists call “theory of mind” — the human knack for modeling what someone else is thinking and feeling.
“AI can’t do this, but it can create a better picture for itself very rapidly on the fly, that emulates this to a degree, and that makes it much more powerful and potentially much more dangerous,” says Yonck.
Emotion-aware AI now reads voice, text, face and gesture at once, he says. Paired with realistic avatars, it can, for instance, convince company employees they are talking to their own bosses.
Tools that could flag an AI interaction in real time, he cautions, are “still a little way off.” Asked whether governments can regulate any of it effectively, his answer is blunt: “I’d say currently no.” He credits the European Union for trying and calls the U.S. approach “a little too laissez-faire.”
“We’re not here to protect the technology,” he says. “We’re here to protect people.”
In the wide-ranging interview, Yonck also explains why the industry’s “bigger is better” race to scale large language models “will hit a dead end,” why he believes the next chapter belongs to what he calls experiential AI — systems that learn from the world the way a child does — and why, despite it all, he remains an optimist.
“There will always be problems in the world. There will always be problems with the future. But I like to remember that there is a great deal that is right with the future,” he says.
This interview was recorded on July 28, 2026, and has been edited for length and clarity.
Glenn Kates, Independence Avenue Media: This year you wrote a 10-year audit online of your own book, “Heart of the Machine” [written in 2016 and published in 2017]. That book, which came out before the AI revolution, explored the ability of machines to replicate human emotion. And as anyone who has used ChatGPT as a therapist knows, this is not science fiction. What did you get right about artificial emotional intelligence in 2016 and what did you miss?
Richard Yonck, futurist: In 2016 and the middle of the last decade, there was still a lot of excitement on the part of various players around what emotion AI could do. Artificial emotional intelligence later became known much more as emotion AI. It had a lot of different potential uses and applications in a range of spaces — from education to healthcare to human resources. In the interim, a number of things have changed. Some of that has a lot to do with our changing attitudes in society around privacy. I think people have become much more concerned and aware about what is involved in being monitored in our world by sensors, cameras and microphones.
And so we’re seeing more and more effort at creating safeguards around this space that really wasn’t happening at the time. In terms of the developments, a lot of things have developed, but understand that in my writing — in looking at the world as a futurist — I was looking a lot further out than just simply 10 years. Ten years, in terms of technologies, is a fairly short time frame.
It takes literally decades for different technologies to go from conception, very early stages to various forms of development, to finally being adopted into the world. This is speeding up most certainly, but you can only do so much around the space of how much we can leapfrog certain technologies and be able to build it to essentially meet our timelines.
IAM: You’ve expressed concern about the growing ability of state actors to do things like manipulate public sentiment and interfere and influence elections. You wrote about something you called real-time persuasion loops and autonomous social engineering. Concretely, what does emotion-aware AI add to a foreign state actor’s toolkit that they wouldn’t have had several years ago?
Yonck: Where emotion AI was 10, 15 years ago was very much a single-channel concept: Could we read a face and label a particular expression as meaning a particular emotion and so forth?
[This] led eventually as everything developed to a much more multichannel approach. You’re listening to sentiment in the voice and the text, and the words. You’re looking at faces, you’re looking at posture and gesture and [you’re] able to create more of a picture of how that person may be responding or feeling. This is important because one of the shortfalls of emotion AI early on is that it really doesn’t understand. It really doesn’t recognize, oh you’re upset, even though you’re smiling. What we do as human beings is utilize something that psychologists refer to as “theory of mind.”
Essentially we are projecting and modeling what we believe is going on in the mind of the person that we’re interacting with. And AI can’t do this, but it can create a better picture for itself very rapidly on the fly, that emulates this to a degree, and that makes it much more powerful and potentially much more dangerous.
MORE: Drones Reshaped the War in Ukraine. Robots Are Doing It Again
So if you’re interacting as an avatar — perhaps a very, very realistic avatar — on a webcam interview or what have you, you might create an interaction that is more detrimental to that individual. There have been situations where members of companies have been convinced that they’re talking to the CEO or a board and told to issue certain funds to somebody or some place and lo and behold they’re out of millions of dollars. So yeah it’s real and the question becomes how far can it go and how can we protect ourselves.
IAM: Yes, I’d like to talk about how you actually defend against something like that. What’s being done to protect people from these nefarious uses, particularly of emotion in AI?
Yonck: Emotion AI is only one piece of that puzzle, but it is an important one. It’s that little extra few [percentage points] of nudge that something can benefit by a form — let’s call it an advanced form — of social engineering.
As far as what can we do, we talk about education and informing the public — that’s useful and great but can only go so far. We evolved as a species to recognize other creatures that are similar to us that talk like us. And we’re in a period where we really can very quickly fall into a hole of thinking we’re talking to something that’s real. Even if we start off [having been] informed that it is artificial — an actor or an avatar — we quickly forget we’re interacting with them.
This goes all the way back to something like [Joseph] Weizenbaum’s Eliza chatbot from [1966] at MIT. You had people who knew darn well it was a chatbot and yet they would talk to it about their deepest, darkest secrets. So, we’re vulnerable. There needs to be ways that we can alert the user — have technology that is the equivalent of our current AI detection for text. We can read certain passages and say, this was AI generated, this is not. To be able to do something like that on the fly with real-time interactions and then be able to flag the user in that way — that’s still a little way off. I would certainly say there are a lot of challenges to it. It’ll probably, like everything else, lead to a little bit of an arms race around the systems finding ways around that and this or that detection having to get better, and so forth. We’ve seen this with everything from antivirus to other types of technology responses in the world.
IAM: What about government regulation? You’ve written in the past about the EU AI Act. It’s something that in Washington there’s been significant skepticism about. Can governments effectively regulate AI?
Yonck: I’d say currently no. In the case of the government regulating, there’s always the worry on the problem of overreach. When we talk about recognizing a problem in a new technology, we routinely have legislation being a slower process by design. It tends to fall behind and then by the time that actually catches up and deals with a problem, there’s often some level of regulatory overreach before it hones in on somewhere that’s a little bit more reasonable.
I personally believe that what they’re doing in the EU [creating a comprehensive legal framework for AI] is actually beneficial because we have to recognize that these standards have to serve people. We’re not here to protect the technology. We’re here to protect people.
So there has to be some recognition of this. In the case of how things are currently being done in the U.S., I consider it a little too laissez-faire. We need to not be looking only to support the businesses that are creating these, but also the end users — the people who ultimately can potentially become very vulnerable.
I like the idea of having certain levels of checks along the way and yet this technology is developing so rapidly that it’s really a challenge to figure out how this is going to be done.
IAM: Right. You’re a futurist. I have to ask you a question about the future. You wrote in a recent post that what you called the “bigger is better mindset” of some of the well-known actors, including OpenAI, Google and Nvidia, could potentially be a dead end. You pointed instead to [French-American computer scientist] Yann LeCun and others who are betting on something completely different.
Yonck: So to begin, yes, a lot of people consider AI to be something fairly new. A lot of the public, for instance, only became really aware of it in the past 10 years and probably less than that.
The reality is it’s been around for close to a century, depending on where it starts from. Technology takes longer than people think it does. And then on top of that, it doesn’t recognize or respect your quarterly report. It really needs to be done in the time it takes. Now, in terms of the model, we’re currently in the paradigm of LLMs, generative AI, and so forth. Unfortunately, in my mind, it will hit a dead end. It is something that is based on language — not just language, but it’s a big part of it. But our intelligence is so much more than just language and communication. There are so many other things that involve our experience of the world.
But in terms of continuing to scale being the answer to how we make LLMs better, I think that there’s a real diminishing point of return coming here. And I really don’t see that that is going to get us where I think a lot of people think it will. And by a lot of people it’s being driven by a lot of hype out of a number of these companies.
In terms of where we go next, I don’t necessarily believe it’s the end chapter. I think similarly the next paradigm is a next step, but not necessarily the last one. But you referenced spatial AI. Let’s call it a subset of something much larger, which I will call experiential AI. This is a direction that people like Yann LeCun are moving toward.
Instead of building their knowledge set from text and not having a true understanding and ability to understand causality, these models are seeking to learn from experience — much like a young child does, much like all of us did early in our lives. Now this is still very early days. This is not going to come about and be more than proof of concept in five years. But it is moving in a direction that I think gets us closer to how we think — how human beings and other animals actually exist and learn in the world. I think that’s really where the next chapter is likely to go.
IAM: I just want to dig on that a little bit more. Help me draw a picture of what an AI learning through something tantamount to human experience would be in practical terms. Is there an example of something like that that’s happening right now, for instance?
Yonck: Sure. So you can talk about it in terms of simulations taking place in a computer where you have interactions that are intended to build from very, very basic components. That’s one version of this. But think about a robot, or some form of embodied AI interacting with a coworker. [It’s interacting] in an environment where it’s learning the ground and how it falls when certain things happen. This is essentially building a model of physics that is much closer to how a child might learn from its world. These are different approaches and they can eventually overlap and inform each other. They can ultimately help the system build a model that is based in causality, based in context, that the current models simply don’t have. They can pretend, they can make you think that they have an understanding of what they’re referencing and talking about, but I can promise you that ChatGPT, Claude, any of the others, have no concept of what a rose smells like or what it feels like to touch someone’s hand or to look at a sunset. These are all experiences and this is something very far from where these technologies are today.
IAM: You have a newsletter called “What’s Right with the Future?” We’ve just discussed some things today that are not particularly right and I just want to ask you, what do you think is right with the future?
Yonck: I wanted to emphasize that for all the things that people say are wrong with how the world is developing, the fact is that we live in an amazing world that is the result of our co-evolution with technology for hundreds of thousands, actually 4 million years. And this has gone on from the point we were an unassuming primate on the savannah as Australopithecus, all the way to now when we’re essentially a globe-spanning species. In evolutionary terms, that is a blink of an eye. What has allowed that to happen is our ability to build and create technologies that have accelerated and augmented our intelligence. [That has allowed us] to transcend time and space to share knowledge across the world, and across generations. And this is huge in terms of how it has allowed us to build as a society, as a community — and it really has overall created a world in which we’re safer, more prosperous and certainly healthier.
This is incredible power that this relationship has allowed us to build. So from my standpoint, we’re living in a very, very good time. There will always be problems in the world. There will always be problems with the future. But I like to remember that there is a great deal that is right with the future.
IAM: Just to respond to that point, how would you respond to people who say, it definitely doesn’t feel “right” right now?
Yonck: Nothing in this world has developed, happened, been improved without people feeling that gee, it could be better, it’s not the way I want it. This is really what’s driven progress. If everything was hunky-dory from early on in history, we probably would have just sat around and accomplished nothing. The fact is that it is need, it’s want, it’s the desire for something better that has created this loop, this cycle in which we create, we invent and we build better institutions. This is for ourselves and for future generations.
