Can We Program Morality Into AI?
Ram Srinivasan says we measure people by productivity, the way we would measure a machine, and he wants humans to be humans again.
In this episode · 18 chapters
- Is GPT-4 conscious?
- Duality is an illusion
- What Advaita says
- In materialism, AI can be conscious
- Hinton, and the counterargument
- Commercial real estate, and deeply spiritual questions
- Knowledge, values and wisdom
- Agentic workflows and substitution
- AI collateral damage
- Step one, one foot
- Swarms, and trusting agents with money
- Prove that AI cannot do it
- The CFO call
- Knowledge workers, and the platform
- Tech built for tech
- Movie recommendations
- Humans should be humans
- Two questions Ram carries
Ram Srinivasan leads AI adoption for JLL globally and wrote The Conscious Machine. He explains how Advaita treats duality as an illusion, and why he thinks today's AI models simulate awareness rather than possess it, whatever Geoffrey Hinton argues.
He also covers why our intuition fails on exponential curves, why trust sets the speed of business, and how an engineering company's employee in Hong Kong paid out $25 million after a video call that was deepfaked live.
Ram Srinivasan, who leads AI adoption for JLL globally and wrote The Conscious Machine, argues on the Thinking On Paper podcast that AI simulates awareness rather than possessing it, and that the future of work should stop measuring people like machines: “You measure the productivity of a lathe, not a human being.”
How did Ram Srinivasan come to write The Conscious Machine?
The Conscious Machine started as an article. Srinivasan says he heard Sam Altman on Lex Fridman's podcast link Silicon Valley's simulation idea to Brahman, a concept from Advaita Vedanta, the Indian school of non-dual philosophy that treats consciousness as fundamental and everything as an appearance in it.
Is AI already conscious, according to Ram Srinivasan?
Srinivasan disagrees with Geoffrey Hinton's view that AI is already conscious. He argues a model that tracks a glass through a prism is simulating awareness, in the same way that a computer simulating a black hole does not contain a black hole.
Is AI a new factor of production?
AI is a new factor of production, alongside land, labour, capital and enterprise, Srinivasan says. He separates digitisation, where a middle schooler uses AI for maths homework, from digital transformation, where the same student can solve PhD-level problems.
What sets the pace of AI agent adoption?
Trust sets the pace of AI agent adoption. Srinivasan says nobody would use a VLOOKUP that worked 95% of the time, so agents built on probabilistic models, which can give different answers to the same question, must become predictable; he cites Moody's, which he says combines input from 35 AI agents that argue towards a consensus.
How are deepfakes changing the value of meeting in person?
Deepfakes are raising the value of meeting in person. Srinivasan describes a Hong Kong engineering company where an employee made a $25 million payment after a video call that was “deep faked live”, and cites IoT Analytics figures putting AI adoption in real estate below 1%, against 30 to 40% in marketing.
Where Jeremy and Srinivasan disagree
Jeremy challenges the idea of human-centred technology, arguing Web3 “built tech for tech” and that builders chase venture capital returns. Srinivasan holds that companies building tech that helps people do better, pointing to Google DeepMind's AlphaFold, but concedes the case needs “some empirical, more than colloquial wisdom”.
Links from this episode
Questions
Who is Ram Srinivasan?
Ram Srinivasan leads AI adoption for JLL globally and wrote The Conscious Machine.
How should the future of work measure people, says Srinivasan?
Ram Srinivasan, who leads AI adoption for JLL globally and wrote The Conscious Machine, argues on the Thinking On Paper podcast that AI simulates awareness rather than possessing it, and that the future of work should stop measuring people like machines: “You measure the productivity of a lathe, not a human being.”
What topics does the Ram Srinivasan episode cover?
He explains how Advaita treats duality as an illusion, and why he thinks today's AI models simulate awareness rather than possess it, whatever Geoffrey Hinton argues. He also covers why our intuition fails on exponential curves, why trust sets the speed of business, and how an engineering company's employee in Hong Kong paid out $25 million after a video call that was deepfaked live.
Transcript
Jeremy Gilbertson Disruptors and curious minds, welcome to another episode of Thinking on Paper where we unpack the future with the people pushing the boundaries of it. My name's Jeremy, this is Mark. Today we have a really interesting conversation that's gonna float between a number of subjects I think that are very interrelated. AI is gonna be one of them, the future of work is gonna be one of them. Believe it or not, consciousness, spirituality, all of these things actually augmented human
Mark Fielding Augmented humans AR VR
Jeremy Gilbertson distributed cognition. mean, there's all kinds of stuff we're getting, but I think there are going to be some great tangible takeaways for those that are in the workforce just in general, because like what is AI going to do to workforce? What is AI going to do to office spaces and you know, the flow of coordinating people in these office spaces and pointing them to work and all of that good stuff.
I'm super excited, Mark. Tell me about where your head's landing with this and where you'd like to go and some nuggets for the people.
Mark Fielding I think for the people, a public service announcement first from thinking on paper, be warned, this conversation might go anywhere. So our guest is Ram Srinivasan, who wrote The Conscious Machine from Artificial to Enlightened Intelligence. It's one of the most accessible, readable, but also deep, poetic spiritual books on technology I've read. It goes all over the place. idea is to...
explain the future of work and how technology will have that impact in it. It's about augmenting the human experience, augmenting the professional experience, using AI specifically, what most of the time, to augment our skill set to make us better, more agile, faster, quicker, more productive, and to create the humanity of the future that we all want. But as I said, as Joan mentioned earlier, we're going to go to consciousness, to AI.
to organic human noise, to cinema, to storytelling. It's all going to be there. Stay with us and say hello to Ram Srinivasan, author, the, I believe you are the AI adoption leader at JLL America. Is that the correct title?
Yes, I lead the AI adoption for JLL globally and I advise clients on all things future of work. It's at that intersection of work, worker and workplace.
Mark Fielding Excellent. Well, thank you for joining us today. Thank you for thinking on paper with us, Ram. And I think to set the scene for your book, for the future of work, one of the things which is particularly different about this book to many of the books I've read about AI is this idea of spirituality, perhaps the Advaita Vedanta is how you open the book. It plays a big role throughout the book.
Could you explain what that is first and then how that corresponds to the rest of the conversation, I think.
Yeah, happy to do it. firstly, thank you for having me on and thank you for reading the book and thank you for continuing to talk to me having read the book. I just saw all of that. Honestly, this book did not start off as Ram trying to write a book. I was listening to a podcast, the Lex Friedman podcast and Sam Altman was on the podcast and Lex Friedman asked Sam Altman, you know what?
Is GPT-4 conscious?
Do think GPT-4 is conscious? And Sam Altman's response was very interesting. Sam says, isn't it fascinating how the Silicon Valley religion of the simulation is getting very close to the idea of Brahman. And they didn't really dwell on it too much, but I knew exactly what he was saying. So I paused, I rewound to ensure I heard it right. So what he's saying here is...
Silicon Valley religion of the simulation, idea that we are all living in a simulation. Brahman is a concept from Advaita Vedanta. Vedanta is a school of Indian philosophy. It's not a religion, it's a doctrine, it's a philosophy where they are examining what exactly consciousness is. And within that, consciousness is described as being Brahman. Brahman being everything.
everything is consciousness, not everything is conscious, but everything is consciousness. So I looked at that and I said, interesting that an AI CEO is talking about this idea of Brahman. I did a quick Google search, turns out Sam Altman has spoken about this many times over, including he said he believes in the absolute equivalence of the Atman and the Brahman. The Atman is the self and Brahman is this universal consciousness. And if you and the universal consciousness are one and the same. That explains many things what we experience right now in our life. I thought interesting neuroscience, AI, technology, everything is converging to this idea of consciousness. Advaita kind of creates a thread across all of them. Maybe I should write an article on this subject.
I started writing an article, turned into a two part article, turned into a ten part article, turned into a book. Because there's just so much here that it's difficult to quantize. If I may use an AI term, difficult to summarize. In the simplest form, you could say there is one truth and people express it in different ways. In the most complex form, it's a 300 page book titled The Conscious Machine.
Mark Fielding and a great book it is.
Jeremy Gilbertson I tell you what, it's so timely us connecting with you and coming across your book because we've been thinking about this a lot. We've been reading another book as well that focuses on similar stuff but this idea that consciousness is fundamental and it's kind of everywhere, And consciousness as something fundamental versus consciousness as something that emerges from the electrical signals in our brain. It's a theory called panpsychism that's been around for a long, time. when you say panpsychism, people freak out and they're like, wait, is that some weird religious thing? It's like, no, it's kind of a thought process on how consciousness emerges. But how do you feel about the fun?
Mark Fielding Hold on, hold on, hold on. Quantum information based panpsychism to give it its true moniker, Jeremy. Well, the book is irreducible by Federico Faggin, if you want to check it out.
Jeremy Gilbertson Well, that's a whole, you're diving in like four rabbit holes down that I wasn't gonna go, but.
Jeremy Gilbertson But so is your, is your, you subscribe to the idea that consciousness is a fundamental element, irreducible building block of, of us, of the world.
So this is a fascinating conversation and this answer to this question may last 30 minutes so stop me at any point of time. Advaita means non-dualism. So there is an apparent dualism where we see us as being different from the other. Whatever that might be. It might be you the body and the mind. It might be you the body and the mind and the world. It might be the world and the universe. There is duality.
Duality is an illusion
But this duality is an illusion, is what Advaita suggests. That illusion is described as Maya. Maya literally means illusion and it's the root word for words like the matrix, matter, mother, all of those types of words come from that root word Maya. What exactly is this? So I need to provide some background on what exactly Hinduism is because there is no such thing as Hinduism.
Hinduism is actually a dharmic philosophy. It is called Sanatana Dharma. The idea of Hinduism is actually a geographic descriptor. The word Hindu is a mispronunciation of word Sindhu. Sindhu is a river in northern parts of ancient India. And over time that morphed into being Indus, Indus Valley civilization. And people who lived on the banks of the river Indus were called...
Hindus mispronunciation of the word Sindhu. This was a Dharmic philosophy. That philosophy is based on argument as in debate. So for example, you might have heard Theravada Buddhism. Vada means debate. So all forms of debate is how these philosophies emerged. There were thousands of schools of philosophy. One of those schools is Advaita. So that's kind of the greater context.
Across those thousands of schools you have things like materialism. Today's modern science is materialism where we are saying that everything that we see is some form of material and below that material lies nothing. This is materialism. Then you have like in Hindu philosophy that's the Charvakas and you have dualism which might be I exist and then there is a different element that exists. It could be you could call it God, could call it spirit, consciousness by any other name.
And then you have this idea of of non-dual philosophy which lie at the other end of the spectrum. That non-dual philosophy is a flip on materialism. Materialism says that whatever exists, it's something like this. Big bang universe, earth, organic matter, human being, complex human being with processing power, consciousness.
What Advaita says
Advaita flips it and what Advaita says is that fundamental consciousness is what existed and everything is an appearance in consciousness. So what Advaita philosophers argue is that Descartes was close but he was wrong. What our friend René Descartes was saying is, Gogeta will go some I think therefore I am. It's actually the other way around. I am therefore I think. I am not the intellect that is aware. I am an awareness within which the intellect appeared.
I am aware that I have an intellect. I am not an intellect that I am aware. So that's kind of the flip that Advaita does. Anyway, I'll pause it. I know I've thrown 20,000 years of philosophy your way, but that's kind of positive.
Jeremy Gilbertson Hey, you heard it first. I'm thinking on paper, Descartes was close. Descartes was close. I love that.
In materialism, AI can be conscious
Mark Fielding So in materialism, AI can be conscious. In this view and in the irreducible view that we've been speaking about in book club, non-organic matter can't be conscious.
What do you say to the people who say that AI can develop consciousness as an emergent property? And how do you answer that question when posed it?
Hinton, and the counterargument
Well, Sir Jeffrey Hinton, Nobel Laureate of recent Nobel Laureate, inventor of the neural network, believes that AI is already self-aware and conscious. Now, Jeffrey Hinton is a materialist, as in he's a scientific thinker. He believes in objective reality. And when he's saying that an AI model is conscious, the example that he gives is something like this. So imagine.
Mark, Jeremy that you're looking at something like this glass of water and I ask you to point to it. You're like here I can see it right in front of me. I can see it through my eyes. Reverse image in the retina etc. I can see it right here. If I put a prism in front of your eyes unbeknownst to you and ask you where the water glass is, you might say it's here. Now that's your subjective experience. This is what we describe as conscious experience not consciousness.
That's a different thing. So for now, let's just call it awareness, subjective experience, awareness, glasses here. I can do the exact same thing with the GPT model right now, which is multimodal, where I can say, hey, GPT model, computer vision, where's the glass? Glass is here. Put a prism in front of the lens. Now the glass is here. So Jeffrey Hinton says that this is subjective experience. This is awareness. You might look at it as a non-dual philosopher and say, actually, this is a simulation.
of awareness. For example, if I were to use a complex computer system to simulate a black hole, does the black hole exist inside the computer? We don't think of that at all. Whereas when it comes to consciousness or awareness, we suddenly jump to AI models are now conscious. They are simply simulating awareness. Now I'm familiar with the way these models are constructed, how they're built, the neural network architectures, the multiple matrices involved with it. It's math. It's a model.
It can simulate various things including the apparent ability to reason but is it actually understanding, it actually reasoning, it actually self aware, is it actually conscious? In my opinion, no.
Jeremy Gilbertson So, so going along with that, the, they're what we've been learning in our, in our research is the idea of like, of qualia, right? So qualia is something that I experienced myself and it creates this inner semantic knowing and understanding that you can't understand a qualia that I experienced, right? So I have to take in the qualia, which is a signal, which is an experience, which is some kind of sense that I pull in.
sensory experience and then once I communicate it to you it turns from semantic into symbolic and I have to communicate it in a language to you. So that's how I kind of look at things is that you know if like two computers were doing that they can communicate the symbolic stuff but can they ingest qualia and generate semantic understanding?
Yeah, you're familiar with Alan Watts' work. Alan Watts was a philosopher and he introduced Eastern mysticism, particular Daoism, Zen, Advaita to the Western world. And Alan Watts once said that if you want to know about Hinduism and if you want to understand these things at a deep level, the Indians had the dope. So we are going to territory where you require some support, chemical or otherwise, to understand in my view. And there are various ideas on that front. But the way I would think about this is these machines are simply simulating based on data that they're trained on. The data that they're trained on is language data that we have fed them.
and they're incredibly good at quantizing that data. And what I mean by this is they're able to create analogies across vast datasets and they've identified patterns within those datasets. So for example, what is the similarity between a nuclear bomb and a garbage heap? Now you and I don't think about that at all, but if you asked a GPD model that question, it would give you a response.
because somewhere in its neural network, it has quantized those things as being simply an energy calculus. And now it's able to make the connection. Or for example, I asked GPT-4 when it was launched to look up two random numbers, patent numbers, combine them into business ideas, and then list them out in terms of business cases for me. Select the best business case and with today's models like gamma or others, you can actually develop an entire...
project. Where did that come from? It came from novel combinations of artificial neurons that exist within those neural networks. So this is still a simulation of intelligence. It is not actual intelligence in my view. But when a simulation is effective enough, does it matter if it is a simulation? This is the only question.
Mark Fielding That's the matrix. If it looks like steak and it tastes like steak then...
Jeremy Gilbertson Very interesting. Well, I think we could, I think we'd stay on this track for a long time because we all think about this stuff with equal curiosity. but I, sure.
Mark Fielding ask one question on it, Joe, just like, so from what you're, from reading your book, from reading Irreducible, from my own experiences in the art of mindfulness and the art, Zen and the art of mindfulness, Eastern philosophy seems to lend itself to the deeper questions about consciousness, about what it means to be human, about the lived communal experience. And yet, at least from our perspective, most of our tech comes from a Western philosophy built and made in the United States. How do we, should we, can we put more Eastern philosophy, more of these ethical, compassionate, empathetic guardrails for want of a better word into our technology? Or is it impossible to do that?
Yeah. A couple of things, right? Firstly, the training data for these models is the public web. The public web contains a lot of Eastern philosophy as well. So these models are trained on it. And what's also fascinating, Jeremy and Mark, if you've looked at Claude for the recent release, Antropic model and its system card, the system card actually tells you what Claude thinks.
and beyond a point it actually has inclinations to talk about spirituality and it starts talking Sanskrit and it will start talking things that sound like Eastern philosophy. So some of this exists in the training data already. Does this mean that suddenly AI models have become spiritually awakened? I don't think so. It is simply again like I said, it's a fantastic mimicry and spirituality or profundity etc lies in the eyes of us as the beholder.
Now with respect to what do do with these models and the fact that they are developed in the West, any technology that we have today, the basis of all of this technology is equations that folks like Schrodinger, Heisenberg, etc. came up with. And you only need to read the preface of any of the books that they have written. Every preface of the book that they have written references the Gita, the Upanishads and the Eastern philosophies and Eastern philosophy.
they'll say things like for anyone who's read the Upanishads quantum field theory won't be news because it's the same. It's very similar. It's very similar. I'm not suggesting here that that's what Hindu philosophy talks about. All I'm saying is that the similarity is very deep and profound and you can find similarities there. So I think there's some of this in the training data already. And then as we go forwards, my hope is that we will magnify and amplify those elements. If you look at, if you stay on what Antropic is doing, one of the persons who speaks with Antropic most internally is not an AI engineer, is a philosopher. When DeepMind is making recruitment and hiring now, they're saying these new AI researchers need to understand human consciousness theory. So there is much thought going into how these models are being developed and I am...
I'm confident that as we go forward, we will magnify that element.
Commercial real estate, and deeply spiritual questions
Jeremy Gilbertson So, all right, you are in the commercial real estate business in a sense. mean, you do the consulting for workplace development and future of workplace and that sort of thing. We're talking about some pretty far out topics, right? And that may be counter to some of the audience you initially engage with, with people that have, I don't know, a billion dollar company with you know, 50 buildings across the world that they're trying to figure out the next strategy for the right place for their people to do the work that they want them to do. How are these messages being received initially and what are you seeing those particular clients start to gravitate towards as you start these conversations?
Yeah, great question and I hesitated maybe for five minutes before I hit publish on the book for this reason because is this message in conflict with it? But then what I realized is that as I start talking about technology, people ask spirituality related questions. Now they may not think of it as spiritual questions, but they are deeply spiritual. What do I do? What is my value? What is my purpose? What is meaning now? Why do I work?
These are deeply spiritual questions. Or if you look at what's happening in the real estate space and knowledge work, one of things I've said in my book is information is expanding, intelligence is exploding, has wisdom caught up? And in the last two weeks, there's been publications from various parties on this idea of knowledge work is dying, wisdom work is picking up.
Knowledge, values and wisdom
Jeremy Gilbertson That's really interesting. So let's unpack that a little bit, because I've always been very interested in the gap between knowledge and wisdom, and I've got my own theories. How would you describe the difference between the two?
Knowledge is interesting again if you look at knowledge the word knowledge comes from the root word Gnosis Gnosis, Jnana, Jnana all of these are similar Indo-Germanic kind of roots for the word Gnosis of course is the opposite of agnostic so one who does not know slash believe and Knowledge is one who knows or slash believes so the idea is that you have a repository of information and you believe that repository and access to that repository gives you some leverage more power. Knowledge is power, not anymore. Now knowledge is free, mediocrity is free. We live in the information age. Information arbitrage is now no longer how you make money. The way you make money or where you generate value is through your values.
if you are able to amplify those values and those values in combination come from wisdom. This wisdom is in us. We don't have to look outside for it. It's simply we need to get reconnected with it.
Jeremy Gilbertson So wisdom is knowledge applied to, or values applied to knowledge.
You could put it that way, yes.
Agentic workflows and substitution
Jeremy Gilbertson It's really interesting. So another quick shout out to the audience who hasn't listened to us unpack Paki McCormack's Most Human Winds. We talk about this next level of abstraction, humans moving up, grabbing different ingredients and putting them together to have this synthesized wisdom, this transition from knowledge to wisdom. I think it's in line with what you're saying here. All right, so what...
Have you thought at all about, and I haven't dug into this tremendously, but post-labor economics. you dove into that and like with this, AI moving towards doing things more efficiently than humans, have you thought about this idea of post-labor economics?
Yes, it is a subject of importance as we go forwards. So in investing post-labor economics, we need to understand that there were four factors of production, land, labor, capital, enterprise. Our entire economic systems are built on the assumption that four factors of production will continue. AI, in my view, is a new factor of production. And a lot of what we do is you could call it labor slash knowledge works slash enterprise, entrepreneurship. The idea created by Jay Schumpeter years back, fourth factor of production, we need to add one more, AI. So post-labor economics, what do we mean by this? In my books, here's how most people think about this. There's a circle of human work and inside that circle, there is a small circle of AI and that small circle of AI is growing.
and continues to grow and eventually engulfs human work and there's nothing left for humans to do. This is not how I see this playing out and like literally what we are seeing across organizations is that humans and AI in combination work well. So the first tier what we're going through right now is digitalization where you have some work processes and some AI can do it better and...
some agentic workflows can substitute for it, etc. This is simply digitization. Digital transformation is the next wave. So this very simple example that I provide is, imagine you are a middle schooler doing math homework in the old fashioned way. And you give this person AI. And now this person is suddenly doing math homework, middle school math homework, using AI. Now this is digitization. There is a manual process.
or say a calculator based process or in a backers based process, whatever you want to call it. Now you have AI doing it. This is digitization. If that same middle schooler is able to solve university level, PhD level math problems, that is digital transformation. You have changed the life. You have transformed the life. I don't view this as you have transformed technology, you have transformed the human life, you have expanded the potential of that human being. This is what we will see going forward.
Will it take time? Yes. it going to lead to friction across our society? Yes. Will things need to be rewired? Yes. Will regulation need to be rethought? Yes. Do organizations need to rethink how they deliver work? Yes. Is it going to be painful? Probably. Will we need to learn and unlearn? Yes. This has been true for every technology that has impacted us. And right now the power of this technology is in our hands and we can use it to do what we want.
AI collateral damage
Mark Fielding AI collateral damage. I'm going to stay neutral and I'm going to ask you a tangible question. So you said that businesses will have to evolve.
that needs to start today, that needs to start yesterday. What would you, what do you say when you go into a business and what's the principal focus today? What kind of mental models should the employees be working on their own personal mental models to eradicate, to update, to work in synergy with AI? And alongside that, what should the CEOs of these companies...
How should they be thinking about restructuring, reorganising, rethinking the future of work, the present of work?
Yeah, yeah, it is the present of work. I think the gap between the future and the present is shrinking hash shrunk. The future is already here. Like they say, it's not just, it's not evenly distributed. So I'll give an example and then I'll talk to what and how we should think. So we are now on an exponential curve. So this gets said ad nauseum. Ad nauseum. We are on an exponential. What does it mean?
We have multiple technologies which are converging. So AI is an enabling technology, but you have AI combining with humanoid robotics, with AR, with workflows, et cetera, et cetera. That combination is what pushing this exponential curve upward. So what is an exponential? So consider something like this, right? If you were to take 30 linear steps from point A to point B, from point A to point B, 30 linear steps about 30 feet. If you took 30 exponential steps,
Step one, one foot
Step 1, 1 foot, step 2, 2 feet, step 3, 4 feet, 8 feet, 16 feet. How far are you Jeremy and Mark in your opinion at the end of 30 feet, the 30 step?
Jeremy Gilbertson We can go to the moon, can't we?
You are now 26 times around the earth at the end of 30 steps, 30 exponential steps. So this is an exponential. So first reaction that most folks have is this is wrong. It can't be right. The math is wrong. Then they do the math and the math is right. So what is wrong is your intuition of how an exponential works. So we live in exponential time. So we need to build an understanding of how these technologies are evolving.
very quickly. Develop a level of literacy and fluency in these technologies, each one of us, whether we are a CEO or whether we are an analyst. We need to develop fluency across all of these technologies, understand how we can use them for ourselves, for our teams, for our business. Deploy them where it makes best sense and choose to work on the things that bring you joy. It's a very simple three-step process, but it's very hard to implement.
Jeremy Gilbertson That actually is wonderful. The first piece of that is really interesting is you have to get people on board with shedding the linear mindset and changing that to a exponential mindset because in linear you can kind of track things along, step one, step two, step three, I can see where this thing's going, but you blink and two steps have gone by and you're halfway to the moon, right?
Jeremy Gilbertson things are happening quicker and staying on top of stuff like you used to be able to is gonna change dramatically. So I think that's really important, the linear to exponential shift in someone's mindset. That's great.
Mark Fielding Jeremy, I'm getting a bit of a lag. Ram's voice is desynced from his video. Are you getting that?
Jeremy Gilbertson Just a little bit, but it corrected itself. It should be fine when we get the local from him. It should be fine.
Swarms, and trusting agents with money
Mark Fielding agentic agents, swarms, swarms of AI agents. Now, I don't know why this is, it's quite an unusual thing, but when everyone people speak about AI agents, they use this example of booking a holiday and the AI agent is going to go onto internet, find you a nice location, book your flights, book your hotel, then link to your crypto wallet and pay for it in USDC. And you're going to sit there and do nothing.
I don't trust the AI agents with my money, with my choice. How do we bridge that trust gap to make, for AI to deliver on this promise across the world, it's all about trust. And right now it isn't there, at least for me.
Yes, so they're absolutely right. Trust is the speed at which business will move. And you know, people say speed is the currency of the business. say trust is the currency of the business. You move at the speed of trust. If you don't trust these AI agents, they're not going to work. So the way I think about this is imagine that you had access to VLOOKUP and VLOOKUP worked perfectly 95 % of the time. Would you use it? You wouldn't use it.
because you want 100 % certainty in business outcomes. You want 100 % certainty when it comes to money. We hold our technology to a higher level of performance than we hold human beings. This is true. So agents will need to outperform human beings manifold by orders of magnitude. Does everything need to be an agent? Not really. Everything does need to be an agent. So first we need to define when we're saying agent, what do we mean? So you had...
pre-generative AI era, you already had systems, artificial intelligence systems that worked autonomously. Like for example, autonomous trading agents that worked on stock exchanges. They're not completely autonomous, they work at nanosecond speeds and they are in deployment for 20 plus years. They work because outcomes are predictable and you know exactly what they're doing, they follow rules.
This is a type of automation as well. Then from there, we've gone into this era now in the last two and a half years, at least in the public domain, of generative AI. So there's a difference between predictive AI and generative AI. Generative AI is probabilistic. The GPS, for example, if Mark and Jeremy search for point A to point B, travel time and distance on Google Maps right now, both of you will get exactly the same.
traffic data, both of you get exact same travel time etc. This is deterministic. Probabilistic is how our large language models work. You ask your favorite large language model a question, you get an answer. To the same question, I get a completely different answer or slightly different answer. The answers are probabilistic. AI agents built on generative AI models, are they going to be 100 % predictable? So this is the problem that companies are solving for.
There are in deployment, are a number of companies that have deployed agent AI solutions that are providing predictable outcomes. So when they start to provide predictable outcomes, that's when you have real world impact and adoption. So this is already happening. So classic example is a public example, for example, Moody's has implemented a system which combines input from 35 different AI agents.
that compete, argue with each other on geopolitics, on investor analytics, on financial data, and then they come back with a consensus opinion. So very interesting. There's deployment already. Yeah, I think it gets really interesting.
Jeremy Gilbertson Yeah, I think it gets really interesting.
Jeremy Gilbertson Yeah. What I might do is just, I'm going to mute you while we're talking just to keep it, keep it rocking here. so what's really interesting in this piece is, the trust factor in business moves at the speed of trust. love that. love that quote. so as companies like, I mean, we've all heard what, you know, Moody's is doing this thing. Shopify has gone to this kind of if, from a headcount perspective, if
Yeah, what I might do is just, I'm going to mute you over.
Prove that AI cannot do it
Jeremy Gilbertson You ask for a head count, you got to prove that AI can't do it. I mean, that's just going to keep happening and keep happening. So trust and 100 % predictable outcomes. This is interesting. So would you advise a little flexibility on the CEO of an organization to say, hey, in order to get to where we want to get, where are the bets that you can make that aren't going to kill your business, but you're allowed to have maybe 95 % predictability?
2 % predictability so that people will experiment with these. How do you deal with that disconnect when you advise companies?
Mark Fielding That's like the poker analogy, only play with what you can afford to lose.
Yeah, I think that you're right that there are some, let's call it bet the company events versus some that are kind of where you can take a little more risk and horses for courses, right? Some of these AI models there, you can get more predictable outcomes. can apply techniques like fine tuning or lag to get more and more predictable outcomes. In some cases, you actually want the AI model to be more creative.
So as an example, we often hear the term hallucination when it comes to AI. So what is hallucination within an AI model? When you and I say hallucination in the layperson context versus when AI experts say hallucination, actually mean different things, at least from my perspective. When an AI is hallucinating, we are...
What we are saying is I expected the AI to give me a certain type of response and it's giving me a different type of response. So it is hallucinating. Or I asked it for the capital of France and it said Munich is the capital of France. And then it can double down on that idea and debate. This is a hallucination. This is a, it's a feature. It's not a bug of generative AI models. The reason being they're actually combining information in the background in novel ways that are not found in the open. They're not a VLOOKUP. They're not doing an index match. They're not saying here's the data in my database. I'm going to take that data and give it to you. That defeats the purpose of having a generative AI model. So in some cases, you want the flexibility. You want this creativity in those AI models. You want the...
sporadic hallucination because it's a feature. In some cases you don't want it at all. So this is what I've been saying that not everything needs to be an AI agent, not everything needs to be agentic and compulsory use of generative AI. Mandatory you have to use a large language model, nothing like this. There's no such rule. Use the model that makes best sense for the case.
Jeremy Gilbertson It's interesting and the idea of hallucinations I mean I always think about this too like you know creativity is the the novel arrangement of found pieces and parts and elements right whether it's a song or a story or a painting or whatever it is you you have this you have this idea of the happy accident right you know where you try to do something and something else comes out like and and you wouldn't have predicted that before so yeah if if deterministically
If you're asking an AI and you're expecting an answer, that's just a test. That's not a use of it, right? So the happy accident hallucination thing is really interesting. I want to shift gears real quick, I can, Ram, to the idea of an AI native workplace. AI native workplace. I think that's a common theme that I saw in my research.
Walk us through, like if Mark and I are an employee at an organization and we just made a transformation to an AI native workplace for our brand new building that we're working out of, can you maybe try to describe what that might look like for us as, you know, when we arrive at the building to when we pack up and go home at night? And no wrong answer here. We can just riff and explore, but like I always, I always want to kind of think through stuff like this.
Yeah, it's incredible the pace at which this is going. We put out a video, I want to say 2012-2013 timeframe on the 2030 vision for future of work in 2013. So roughly 10 years in advance, let's call it that. If you look at that video today, many of those things are already here.
So it's incredible the pace at which we are going. Is that mass adoption at scale? No, but some of those technologies weren't available to us. Now, almost all of them are available to us. Incredible pace. So when we say AI and real estate, it's a field that's got incredible opportunity. So IoT analytics, for example, does study on adoption.
within different industries. Highest adoption of AI is in marketing 30 to 40 % adoption across marketing functions. Within real estate, the adoption rate is less than 1%. So the adoption potential is almost 100%. So in the next five, seven years, you're going to see intelligent automation, natively intelligent assets, smart building infrastructure, all of those things come up. Why? And what's the value of this to you as a as a user of the space. So a few things are happening right. So first is we see that the value of human connection is going upwards. We are able to connect or the Zoom calls, et cetera, but trust, you're able to generate trust quicker in person. You're able to generate connection deeper in person. Moreover, and Mark Cuban said this recently that face-to-face connection will go up because of deep fake technology, et cetera. That is one vector. So I'll give an example. An engineering company based out of Hong Kong, one of their employees joined a video conference like this podcast, right? The CFO on the conference, they had a discussion. They agree on making a settlement payment to one of their suppliers, $25 million settlement payment. This person makes the payment.
The CFO call
And then a CFO calls him and says, who approved this? And he's like, you did. We were on a conference call together. And he's like, no, I wasn't. So the entire video conference was deep faked live. And this is a very real risk. This happened last year, the way. It's not this year. This year's technology is better. So this is a real thing. This is a real thing that happened. Yep. Real. How did it's a fake?
Jeremy Gilbertson So this is real thing that happened.
Mark Fielding How do you, it's a very, Gamoneka. It raises, I've been thinking about this a lot and deep fakes and critical thinking and gullibility, what I used to call gullibility, I think has been replaced by that. And AI isn't, AIs aren't susceptible or as susceptible to fake news. And.
If your HR manager, if your CFO, if your CTO is susceptible to fake news, then your company becomes as strong as its weakest link. And if people are so ready to accept without critical thinking, and we all know that people are using AI and they are taking the shortcut and they're putting their critical thinking to one side, and that's only going to get more prevalent.
I guess there's kind of two questions in this, like how should companies think about, I mean, is it possible to train your workforce to think more critically about the content that they digest so that they don't make decisions which could affect the bottom line of a company based on videos, based on articles, based on meetings that they're having, which aren't real.
I think part of it will be deploy some type of AI solutions like spam and spam filters or virus and antivirus software.
Mark Fielding It's like you're saying your employees are spam.
Jeremy Gilbertson No, the incoming signals to the employees are spam. So the tendency of discovering deepfakes automatically. That's,
Yeah, it's part of it, but it's also an adversarial battle in the sense that you might get better at blocking the spam, but the spam gets better and it's an adversarial battle. Same with viruses and ransomware and all of those types of phenomena. The human in the loop is the weakest link. And with AI technologies, we'll have to develop an intuition for it. There's a very interesting study I was reading this morning where people who pay for AI subscriptions. Like for example, you're chat GPD subscriber or you subscribe to Claude or any one of these AI models. People who are paying for the AI models have a greater intuition on when they work and when they fail. It's got nothing to do with how much they use the AI models. It's whether you're paying or not. So that's interesting. It's a little bit of human psychology here where because you're paying for it, you're paying more attention.
And because you're paying attention, yeah, and so you gain this intuition. So this intuition will be part of it. Remember, this is all net new. I'm surprising. We think AI is everywhere. We're kind of close to this idea. But the more you speak with people, their exposure to AI is not that high. They're just learning it. Some people are just discovering chat GPT, as shocking as that sounds. Some people have never used AI in the workplace.
Jeremy Gilbertson Skin in the Game.
Knowledge workers, and the platform
So for many, for the large part of the knowledge workforce, this is net new. So the literacy levels as they go up, I think some of these types of phenomenon will go down. Our companies, technology companies that have invented these technologies will come up with safety measures around it as well. So there are multiple angles here. But what we cannot do effectively right now is develop trust connection at pace.
on these technology platforms. If I had to sit and verify for 15 minutes on every Zoom call and prove that I'm actually a human being, it defeats the purpose of a discussion. And this is where I see the value of in-person interaction going up. Also, the nature of that interaction changes. Now you need much more of a holistic type of experience where services combine with technology, combine with the attitude with which those services are delivered.
And can I use the data to deliver Jeremy or Mark, your favorite cappuccino when you arrive and those kind of things. Now, that's something that changes the type of experience that you have. You're generating moments that matter for people. And that's the direction that space, workplace is heading, physical work environment.
Jeremy Gilbertson Two things I pulled from that that are pretty interesting. Again, heard it on first, heard it on thinking on paper first with our guest Ram just by the subscription model and you'll activate your critical thinking a little bit more than the free. That's a really interesting little hack, but.
Mark Fielding I'm glad he said that because I thought he was going to say those who pay for their subscriptions have actually used less critical thinking. So I'm glad that he said that.
Jeremy Gilbertson No, that's a really cool, like cool hack. think just that it's based on, you know, human, way humans operate. Right. But then number two, you could just bring Mark and I, uh, in and just, we can help you think on paper and we get back to, back to our own critical thinking and processing. Um, Ram, got another question too, and I know we're, kind of, kind of getting to the tail end of our, our, our time here, but what I wanted to go back to was the idea of exponential steps, you know, from linear to exponential.
This thing's taken off like this this convergence of technology is taking off. We are 100 % in exponential mode with this so The the proverbial plane is headed down the runway at a very fast pace How can we get to a I mean it seems like it seems like the empathy the the the emotional quotient kind of stuff that make us human is a Secondary workstream here. We're chasing this exponential
Can we bolt these things on later or are we gonna be in big trouble?
It's a great question and this idea of a bolt-on is also fascinating and the way I think about it is every word that you said there is we need to think about human centric. Human by design, default. Default human. The purpose of technology is not the technology itself. The purpose of technology should be how do I make a human's life better.
How do I make this workers work life more joyful? How do I and why do that? Why make somebody's work life more joyful? That doesn't relate to bottom line at all. In fact, completely the opposite. The more joyful your work force is, the more engaged they are, the higher your bottom line is. People don't buy from technology, people buy from people. So you might have a fantastic AI agent.
and that AI agent is in deployment and reduces the work burden. But ultimately that connection between people is what creates business and what creates value. And we said that at the beginning of this podcast, we were talking about values over value. And I think that that's where we are headed. We are also headed in direction just to tie this back to our opening thread on spirituality. Our work has become a series of rituals minus the spirit. We've got to put the spirit back. Spirit plus ritual.
Tech built for tech
Jeremy Gilbertson Interesting, interesting. So I'm gonna push back on that just a touch. I love the idea. It's a very, it's how I want technology to be built. Like every technology builder sitting down going, man, how could this help humans? But I'm gonna illustrate a whole industry, Mark's gonna giggle here, a whole industry that was built on tech for tech's sake rather than tech for humanity. And Web3 is just that.
They built tech for tech. I know it's changed a bit and there's some other things happening in that space, but I think largely we.
Mark Fielding No, no, that's what it's become. Decentralized crypto, Bitcoin did not start that. It's like the internet as the whole internet has changed from what it started out as the vision that it was created for to what it's become. So as a Web3 writer, I will edit that.
Jeremy Gilbertson Changing Changing not changed I think but no I I still see I don't know like I said Ram I believe in what you said I believe that that's the way it should be I just don't think the tech builders today have that in mind as as they as they need to because there's financial Implications to this they're building platforms. They're building They're trying to generate return for the VC that dumped two three million bucks into him and that sort of thing How do you how do you think about that?
I'm advocating for some research on this. But if we look at technology for technology's sake and companies that are technology for technology's sake versus companies that have designed technology that helps others, I think the latter have done better, generally speaking.
But there needs to be some empirical, more than colloquial wisdom. I think there needs to be some data behind this to help support the argument. But think companies that are advocating for and building technologies that help and develop community and build society, those types of technologies and those type of companies are doing well. And we look at, for example, Google allowed us to retrieve information from the internet. That was a synthesis initial idea behind Google.
Google DeepMind, they're doing some incredible work with Alpha Fold and those kind of things. gets press is Gemini, what doesn't get press is the work that they're doing with Alpha Fold, Alpha Evolve, those technologies, isomorphic labs, those kind of technologies. I think there is ample evidence of tech for humanity and I think that those elements will do better in the long term.
Mark Fielding And everyone is trying to, as we said at the beginning as well, if people are looking for developers and CEOs which have studied or are studying these philosophies, then it will become more so. Ram, reading your book, one thing that struck me, which I think maybe why I liked your book so much actually is because of the movie references. Like every chapter is aligned around.
Movie recommendations
Mark Fielding the storyline of a movie, there's Do Sex Machina, Arrival, Her, Lucy, Ghost in the Shell, Matrix, a lot of others. And I say this a lot, that I think that science fiction is in a world that's so saturated with meaningless content and opinions and thought leadership. I just always feel like you're going to get more from a good science fiction book, you're have better understanding of what the future will become by reading or watching science fiction. book, what's the expression, the return on your dollar, the whatever it is.
For our listener, because you mentioned so many movies, you're obviously a movie buff. What one movie should our listeners watch this weekend to get what you believe is a true reflection of where we're heading?
One movie, wow. That's a tough one. think... Top 3. I mean Matrix, if you haven't watched The Matrix, you must watch The Matrix. I love the movie Lucy and I also love the movie Limitless. And some of the writing in these movies is incredible. But if I were to pick one, I would pick The Arrival. The Arrival is my all-time favourite science fiction movie.
Mark Fielding Or your top three, you can do a top three.
It is incredible. It's a loop. Our world is a loop. It is poetry on the cinema screen. It's the best thing that I've seen.
Jeremy Gilbertson I don't know if I've seen that one. I'm going to take your advice on that.
Mark Fielding It's the Matt Damon aliens arrive, isn't it? And they don't have the language and they work out the language that they're speaking to.
Yeah, Amy Adams. Amy Adams. Jodi Foster-Vernon is the contact. You're talking to peak nerd here.
Jeremy Gilbertson Jodie Foster? Is that the Jodie Foster one?
Jeremy Gilbertson the contact, okay.
Mark Fielding That arrival doesn't really give us a true reflection of where we're heading. Possibly, possibly not. I don't know. aliens will arrive. What you said, you really liked Lucy, did you say? What was the storyline of Lucy? I don't remember that one.
So Lucy follows the story of Scarlett Johansson appears in almost all science fiction movies. Scarlett Johansson plays this individual who is a drug mule and she ingests this new type of drug but she ingests it in very large quantities just as a carrier and somehow it enters a bloodstream and without giving the plot away it it unlocks limitless potential in herself and she is able to understand herself better and over time understand that she is actually this one consciousness.
Jeremy Gilbertson man, I'm on it. I love it.
It is an incredible movie and just on the arrival, the reason I like it is because of the language and it shows you how language actually shapes your minds and we human beings, call ourselves very humbly wise humans, homo sapiens and we use language to tell stories. Those stories shape our minds and now AI is telling stories. Right? So we've got to be careful. What is rewiring our brains now?
Jeremy Gilbertson Amazing. Well, this this has been a fascinating conversation. you know, ton of great nuggets listeners for you to pull out of this. We'll be creating some shorts and getting some of the things that we love the book, The Conscious Machine. Give it a it a check out and give it a read. We have one more question for you, Rahman. We had Kevin Kelly on the show a couple of months ago and he posed a really cool question to leave our guests. One I would like to leave with you as well and see what your response is. So what should
Mark Fielding telling better stories than us.
Humans should be humans
Humans should be humans and the reason I say this sounds like a tautology but the reason I'm saying it is because we have somehow trained humans to behave like machines. So look at our work, what do we evaluate? Productivity. How do we measure ourselves? Productivity. How do we measure ourselves? Some kind of dollar value. Salim Ismail said this on one of his sessions where he said, we have materialize the world, we have now got to civilize it. I take that to the next level, I say we've got to spiritualize it. And the way we do that is by getting in touch with who we are. We are not machines. You measure the productivity of a lathe, not a human being. You don't raise a human, you allow them to prosper, flourish, thrive. This is where we should go. So humans should be humans.
Jeremy Gilbertson This is an important thread that we've heard throughout a lot of technological conversations. And I think it's really important for you listeners to think about that because tech is not just tech. We have to point it back to allowing ourselves to point us back to our humanity. Ram, that was a great response to that. Do you have a question? And I know this is spinning you up without some notice, but what question, if you could leave one, for an upcoming guest of ours that's been on your mind that you've been wrestling with that maybe you have partially answered but you'd love a little more research on.
Two questions Ram carries
I have two questions that I have kind of grappled with or thought about at length and they are abstractions of more specific questions. What is it by knowing which everything else is known?
And second question is who or what am I and why am I here?
Jeremy Gilbertson Again, back to self-awareness. I love it, I love it. Ram, thanks so much for joining us. It was a great conversation. Definitely recommend diving into your book and please stay in touch with us on the future, the present of work, the future of work. I think we could have gone down 90 rabbit holes today. We went down three or four, but we appreciate you joining us. You can find all of our episodes thinking on paper.xyz. We have a book club, we read.
Jeremy Gilbertson We read books, we unpack them, we distill the information that initially makes us look silly, and then you can real-time witness us getting to possible understanding of complex subjects. So Mark, what takeaways do you have and any final thoughts?
Mark Fielding Thinking on paper.xyz for the triangle. Recently we've had some awesome guests. We've been bashing together quantum computing, quantum mechanics, consciousness and AI. Guest from IBM, Federico Faggin, thinking on paper.xyz. And you can join the dots with us. Otherwise.
Jeremy Gilbertson Be curious, be curious, stay disruptive.
Mark Fielding Keep thinking on paper.




