#477 The Singularity of Hope: Sam Sammane on Amplifying Humanity with AI

Episode 477

[00:00:00] 

Mehmet: Hello and welcome back to a new episode of the CTO show with Mehmet today. I'm very pleased joining me, Sam. Sammane, joining me from the us. Uh, and Sam is a bestseller author. He's a serial entrepreneur and very happy that he's joining me today. Sam, I like to [00:01:00] keep it to my guest series themselves. So tell us a bit about you, what you currently up to, and then we can start the conversation from there.

So the floor is yours. 

Sam: First of all, thank you for having me. I'm very happy to be with you. So, uh, I'm a serial entrepreneur. My education, early education is an applied physics and I did like a PhD in nanotechnology, which was more about, uh, theory, improving some simulation using, uh, machine. Today we call that ai.

Somehow, it's a kind of early ai. Today we have The generative ai. Then I went into business. I started several companies. Essentially. It wasn't the domain of life science, not telli, it was just business, but it was a very successful business. So we created the first group, a second group. Now I'm, uh, soon I be announcing my third group of labs at the same time.

I always had, uh, a passion for software. So we always built software for labs. More recently, now, we are also doing. Two kind of startup. One is in public [00:02:00] relation, AI based called Alexa, and the other one is a consumer branch for a first time in my life, which is em with somehow, uh, related to, uh, large language models and, uh, let's call it the reward of using AI to amplify our human intelligence at the same time.

Um, I have a passion for writing books. My first book was nonfiction. It was a distiller. Of Hope. And the second book that I was just released and become a quickly bestseller is a novel is a fiction called The Republic of Mars. 

Mehmet: Wow. Like really? It's amazing. Uh, you know, background and journey. Sam, out of curiosity, you mentioned multiple things, so fiction, nonfiction, books, and then technology.

What's the what, what, what, what is like, uh, the thing that attracted you to both worlds? 

Sam: In fact, I will tell you something, I have three girls, so I always define myself first of being a, an [00:03:00] engaged father. I have three beautiful girls. I always think about their future, how, how the technology will shape our society and will change our lives.

And I want it to happen in a way that, uh, improve our quality of life, uh, give us prosperity. And give us freedom. Um, this is the common thing in all what I'm doing. So between life science labs to, uh, technology companies, but essentially books is about thinking about the future. So I have nonfiction and fiction.

I. So the nonfiction is very hopeful. This is what I wish to happen. This is how I see things. And the fictional one, in fact is a little bit dark, is what I wish not to happen. It's the Republic of Mars. I don't wanna spoil the story, but it's somehow my fears. I. At the same time, my hopes a little bit. And um, I put some stuff there, uh, before it happens.

And yesterday it happens. I thought, oh my God, I'm not [00:04:00] foreseeing the future, but I wished for something, something is happening. So yesterday, an impressive piece, uh, meeting that I described in my book in a fictional way. I, I pronounced a meeting between Saudi Arabia and US and I put other countries, China, Russia, and I called that the golden Path, and that will happen and it changed the humanity face.

Uh, I wrote that two years ago. I published it, uh, two months ago, so I was surprised that it happened. I wish, really this is, uh, the great thing for humanity and wish that to be a, a start new beginning for, or human race, but especially the Middle East. We are done with war and trouble stuff. 

Mehmet: Absolutely. Uh, and by the way, the time we're recording, you know, this, this episode, I'm transparent with my audience, so this goes live maybe in two to three weeks.

So yeah, so big things are happening. I don't talk, talk politics much on the show, but yeah, so we, we. People here, we are [00:05:00] hopeful, um, that good things are gonna happen and, uh, you know, a lot of good things started to to to happen also as well. So big investments going both ways, uh, especially on the AI part.

Now, uh, Sam, you hold a PhD in nanotechnology, right? Mm-hmm. Uh, so I'm very curious. Maybe the audience knows me. I, I'm very curious. So when we talk about nanotechnology, we talk about like, you know, something very sophisticated. So how, and, you know, you need to, to really understand, you know, materials. You need to understand physics, you need to understand a lot of, of, of, of the, you know, these, uh, scientific concepts.

So how that, you know, influenced also your thinking about. AI in, in, in practical applications, for example? 

Sam: No, it's, it's a. It have somehow shaped me very early. And when I was young, I had this amazing, uh, uh, physics teacher when I was [00:06:00] in the high school and I decided to go and study physics. So when I went to the university, I studied applied physics, and I got a degree from, uh, France, Bru Noble.

It was, uh, one of the major physics, uh, engineering school, but that was not enough. I thought it's, it's so like. Somehow general. So I went into computer engineering at that time. We call it hardware circuit design in France, not computer engineering. And that's really how, how you build the check, how, how you do the computer deeply.

I went into nanotechnology and that means creating nano robots, but that was late nineties, early 2000. That means. Plus to early, early for that. And at that time, my subject or topic was really how we can make computer things, how we can do automatic theory proving, and we can do verify circuit using symbolic simulation.

So sophisticated and mathematical. I will be honest. At that time we didn't succeed. It was early [00:07:00] time. It was like 25 years ago. Uh, that was early, but we said at that time, we need 20 years. And what's magical? Nothing, we, we, we anticipate will happen happen, but I. Some other stuff happened, result of that early research.

So this early research give me like a structural way of thinking and also I understand a lot the technology. I tell people all the time, we have a lot of hide. That doesn't mean that AI is not great technology. It's amazing technology. Always we say done by ai, that's frustrate me and I correct people. No, it's not.

It's done by a human using AI tool. This give the machine what it deserve, but hey guys, human is still needed and I think for a long time and we are casting an erosion of intelligence sometimes that is over tested and people think really that the machine is somehow aware. No it's not. It's streaming.

It's result is the same concept of. [00:08:00] Google search, anticipating the next word, but now instead of anticipating a word, because of the how big the machine is, how good the algorithms are, it could cast you pages. It can anticipate pages, and that's amazing, of course, but still the real influence. I understand how these things works and yes, you will not, uh, cheat me.

Uh, I, I gave a small story in the beginning of the Singularity of Hope. I said there was a little boy and he was in the nineties. Uh, trying to impress his grandma. So I did this software and showed to the grandma and her, listen, the, the computer's talking to me. It's alive. It's a smart, it's a grandma. In the nineties, she said, okay, maybe I said, this boy has, is now not a boy anymore.

He's a computer scientist and he's not cheating his grammar. He's cheating the whole word. Now. 

Mehmet: It's, it's, yeah, I like it. Um. [00:09:00] Uh, just, you know, for me also when, when, when I was, uh, I think seven, eight years old. So, you know, we used to watch this, uh, I think Japanese, uh, they call them anime. We call them here cartoons.

Right. So, and always, you know, they, you, they used to show the, the robot, which is very smart, and you talk to it and you can ask it any question, it answers you. And it looks like now, you know. We are there almost. Right. And I'm happy, Sam, because you bring both practical experience plus academic experience and I'm happy you explained this about the ai that AI doesn't think by itself, right?

So, so it needs. Us to give input. So it provides output. Yes. It's fascinating and I think to your point, you know, the algorithms, you know, today, I, I always tell people like, it's, I think Google, but of course on, on a, on a steroid, like right, so, so you have all the knowledge of. Humanity [00:10:00] and it has the algorithm to just grab pieces and then put it together and all those people.

Okay. I know it's hard. I, I don't have a PhD degree, but I understand papers. I find it hard, like Sam and you, this is why I was happy. You explain to people this way. I tell them guys, there's something called the transformer model. Like you, you take multiple things and then you convert them to something else.

I try to make it simple. Now, back to the point about. You know, AI and the use by us, right? So debate AI will replace humans, will not replace humans. Uh, from your perspective, how you see, you know, a partnership, uh, could look like between us and, you know, if I might like little bit also make it loaded question I know with what.

People are talking about now, especially, you know, open AI and you know the others about a GI. So, so how do you envision such relationship between [00:11:00] humans and this super AI could look like? 

Sam: Oh my God. Uh, I wrote the whole book about it called The Singularity of Hope. So I talk about each of that in a full chapter.

And it's amazing how you put everything, uh, in one sentence, but, uh, let me start by one by one. And it should be the topics of all the discussion. So first of all, jobs I tell people, uh, it's not AI that become human-like it's we humans, we have behaving like, like machine. In the last 50 years, we are asking people to memorize things, to follow orders, to follow SOP, to follow instructions.

In fact that was the problem. Now we have machine that can execute that. All this kind of, let's call it, uh, matic jobs or machine jobs will go back to the machine. And I am hopeful. I tell people we have a chance to gain humanity back. But what's the humanity? What's the difference? I will tell you what's the difference?

We human, we learn with love. [00:12:00] What does that mean? You have a baby. And by the love of baby and parents, the baby loves. If there is no love, the baby will suffer. And we have a lot of studies about it, but even later in life, if you like something, you learn it quickly and we struggle to learn something we don't like.

That's, that's the truth is love likeness is, uh, in the, in the broad meaning is our essential mechanism, but we have other mechanism. We have what we call intuition. Intuition is a very complicated algorithm because without data we can judge something as good or bad or, or right or wrong. And most of the time we are right, like we call it, sometimes educated.

Guess when it's mixed with education, I. We have also our empathy. We can feel each other without communication. Sometimes it happens a lot. Like you see a wonderful announcement, uh, historical event, and you are in tears without exchanging data. That that's, that's [00:13:00] amazing. That's the human part. We have also what we call the soul.

And the soul I defined in my book is the mind talking to the self. It's something very complicated. Every human being has experienced this situation when your mind is talking to yourself, but it does not mean a lot for a machine. What does that even mean? So to go short these things, mix them with an ai, then that can access all the humanity knowledge in a fraction of a minute.

I'll not say second, but fraction of a minute. Give it back to the human. And this is what amplify our intelligence. So, uh, talking about jobs, a lot of jobs have disappeared and we have a challenge to replace them by human job. Imagine the future when you have a human and an army of robots, either digital robot, which is agent, uh, I bot whatever you wanna call them, or physical robot that listen to you.

That's amazing. But we [00:14:00] don't want like bad ideas to come and use the these robots against us because also it'll be awful to have an army of tax collector instead of having robots working for a humanity and we don't pay taxes. You see, you can have the two visions. So I'm very hopeful about the future.

Now you mentioned a lot of other things, which of them? It's a question like a JI like new tech companies. Like super intelligence. I think there, there is a moral issue here. Do we need to create super intelligence? I don't think so. Why we need to do that. It's really a step into, let's call it dangerous territory.

We don't know what will happen if we create really super intelligence and that's intelligence that belong our imagination. But something else, let's be humble. We don't know how to do it anyway. I hear these people going into arena and forums and discuss, yes, EGA two more. They have been sewing that for a year.

Now I, I'll have exclusivity for you. They will not do it [00:15:00] and they will package something that is super, let's call it good generative, uh, ai. That means, uh, a very good bad ai and they will call it a, just to save their face. They cannot do it, not because they are not intelligent enough, because let's be humble, we don't know how that even works.

We have no idea yet. Not because you have strong machine, very powerful algorithms that are very speedy and cast illusion of intelligence. You mentioned, uh, transformers. I tell people the generative AI is like linear and jab. It's not smart. We didn't figure anything yet. It is just big and rapid and giving us this illusion.

So, no, we'll not figure it out. Last thing I tell people, if you are that obsessed by creating or recreating human intelligence, do it the right way, which is their biological way, and bring more babies. This is recreating intelligence. We have it already. And [00:16:00] make them well educated. Make them smart. Give them the ai, and you will achieve super intelligence by a new generation of humans, not by a new generation of machine.

Mehmet: This is sort provoking for me, Sam, honestly, I agree with you on something, uh mm-hmm. Which is if you, I come, I, I studied, uh, you know, computer engineering as well. And now if someone asks me what does a GI, you know, I said. Okay. They will find that multiple definitions online, right? Mm-hmm. I say, okay, but how you define that a machine, you measure it as smarter than a human being.

You know, like, yeah, they make these tests like, I don't know, right? So they, they said like, it can pass the PhD exam and they can pass. I don't know what I say. So what, like, because honestly, if, you know, these exams are kinds of programs if instructions, so you, if I have the power. To [00:17:00] memorize and analyze and write or type them fast.

Yeah, I will be myself. Super intelligence, right? I mean, if they want to call intelligence as doing something fast and accurately, I think the issue is. My opinion, my 2 cents. Of course, I'm, I'm, I'm not the expert at all. I don't claim this, and I told many of my guests and you know, a lot of people I speak, I, I, I say, I think the biggest challenge for us is defining the word intelligence, because when we were like kids in the school, I would ask you also, like maybe Sam, you have some, you can shed some light.

If a kid, for example, used to memorize the. You know, multiplication table. Oh, he's smart, right. Or his intelligence. And you find the other guy who was like, oh, he's, he's like, sorry to say like, he's, he's dumb, right? He's, he's, he, he can't do anything. You go back in life, you meet these two people or two persons after I.

20 years you find, you find that the one who was like doing the multiplication table, he's [00:18:00] like working as a, uh, humble accountant. No offense of course, like normal career. You see the other guy is a big businessman, an entrepreneur. So for me, you know, the word intelligence is kind of debatable. I'm not sure if you agree with me on this or no.

Before I go to the next question. 100% 

Sam: and, and somehow you proved my point. I told you first of all, we are creating machine jobs. So the guy who is memorizing things, I'm sorry. That's useless. It seems impressive because we human, we like knowledge, you know? So when someone is reciting something for us, we are impressed.

The truth when you mention the other guy who become a businessman, he's using his ultra intelligence. The intuition, his intuition and business go hands in hand. You make decision not based on data because you need to do it quickly, but based on your intuition, you know it's right. You need, you want to prove it's right.

It's weird, and I'm not the first one to talk about that. I have this, uh, famous Nobel Prize, like, let's call it, uh, [00:19:00] uh, living, uh, legend that we still have. Hopefully he will still live for a while, for a longer time. His name is Roger Penrose and Roger Penrose said it in his book a long time ago, like, you know, it's right.

That happens a lot in mathematics. He didn't prove it, it's right. And then he will spend hundreds of years, a lot of mathematician to get the proof. We have stories about that, like firm theory and other stuff. I told him, if firm approve this stuff, never. He said it's right. I proved it one day and people are struggling for hundreds of years to prove it.

And finally they did. So. What happens in this process of human intelligence? So complicated, and you have a point we cannot describe it. So how you are going to program it if you cannot even describe it. And I like test and benchmarks because it's trying to give a number for intelligence. And I, people ask me sometimes what's an IQ test?

And I used to tell them it's just a number that sets. A test given by a human [00:20:00] trying to give a number in this way because it means nothing. You can have a guy who, who succeeded all this tests, and then he's a, a machine. He's working in a, in a company as a humble programmer. There is nothing wrong in that, by the way, but you have another guy who, who will never succeed with high numbers.

Who use his intuition to do business. And he is very successful in life. And by the humble definition of intelligence, used by families, not used by, uh, let's call it, uh, scientist, especially in late life, as you said, oh, he's intelligent, he made a lot of money, not, he is intelligent. He can memorize a lot of numbers, right?

So at the end of the day, human intelligence is still a mystery. I like how this generative AI revolution, which I tell people all the time, it's just linear. Stop pretending. It's more than that. It's just like mattresses, multiplication. You have input output, and in the middle you are figuring some, [00:21:00] let's call it, uh, parameters, and that's how they call it.

Now when they tell you about large language models, obviously there is this number of billion of parameters, right? It's the parameters that will figure the output as an equation of an input. This is linear and J algebra. This is not intelligence. This is not, not, okay. This machine can can guess the solution without knowledge.

No. I heard some people who use also misleading terms. Right? I love these terms because they're funny, but it's not related. Like when they tell you machine learning or deep learning, you have the impression it's mimicking our learning. It's not. What they call machine learning practically, it's linear algebra, interpolation, it's guessing the parameters, algorithm.

We need to guess it. Okay. And then have an output. That makes sense. Now, what was surprising for all the human, honestly, is the scale. When we did that at scale, we had this like illusion of intelligence and some people said, yeah, it is how we. [00:22:00] Work. This is how the human brain works. I told them, okay, maybe.

And here I put a big maybe 'cause I wanna be scientists, so I'll not tell you no. But honestly they don't know either. It's just a guess. It's just a name, a label. So good luck with that. It's not how we need to think. I think humans. I'm not only also the first one to say that we are a mix of what we know as computing or mathematical computing and quantum computers.

We have even some new studies that prove that our cells have a phenomenal, we call it quantum resonance, which somehow I will not really you, it's prove we are quantum computers. We are not there yet, but it's a hand that it's maybe true. Quantum computers is crazy. We're not going to start to talk about it, but it's not generative AI at all.

It's something else, something crazy. When something can be true and false at the same time. When you can do something, uh uh, you think in the real world, or let's call it the big [00:23:00] physics, it's impossible. So again. This is, I think, and this is only my humble opinion, we are quantum computers. We didn't figure how our brain works yet.

That's so complicated, and I agree with you. We don't know really what's inte human intelligence really mean machine intelligence. Yes, we have a way to define it very well. 

Mehmet: Yeah. So again, to your point, Sam, and again, and I think, you know, I get it from you, we are not saying this to underestimate the power of this technology actually, and I can get it exactly from you, Sam.

So what Sam is trying to say, like we don't need to exaggerate on the. On the wards, right? So we don't need to to say, oh yeah, yeah. Deep. I don't know what, you know, like as you said, so it's just a technology that humans came up with. It's, it come from mathematics a hundred percent. So I tell people like generative ai, I say like, think about it, it's like automation on steroids, right?

[00:24:00] So. Because I'm, I'm a big fan of automation, you know, like, I like to automate. When I was like, uh, working as system admin, I liked automating. I was questioning like, why you have to do this in seven, eight steps can be done in like two or three. So it's always in my mind and you know, generative AI is basically, you know, actually it's automation if you think about it.

Yes. It's like going searching pattern. It does some pattern, you know, recognition. And this is where what we call machine learning. So we teach the machine actually. How we think and then the machine tries to imitate what we do and then we automate the whole process to come up with this. So a hundred percent now, but let's come Sam to the practical aspects of this.

You know, amazing technology, right? Again, we are not like saying it's bad or, or no, actually, it's, it's revolutionary. And some, I, you just mentioned this now, if I want to look at its particularity from business perspective, uh, like use cases [00:25:00] is what are you seeing in that space that really, you know, uh, excites you?

Sam: First of all, I agree with you. Automation is the best part of that technology. Interacting with AI Chatbot is good but not good enough. It's better if it's was an information and in terms of, it's coming. Let me go back to business. So ourself, we are enjoying that technology, so we are going to apply it everywhere.

Uh, I still think it's in the early days, like I don't believe we have a full system that can, you can use AI and forget it. You still need the human in the loop. This is first to, to supervise somehow, like, uh, people who are using it for content generation will understand what I'm talking about. Like we a catastrophe.

If the AI start hallucinate and instead of generating one article online, it can generate thousand and you have thousands of hallucination, that's a catastrophe. So I'm not [00:26:00] undermining the value of the technology. It's amazing. In fact, we built systems using this technology. Uh, in our company, texa. We built a system that will help people generate content, manage client, and also you have a virtual employee that can handle your clients.

Uh, talk with them on the social media, interact, take information. He'll be convert more client. But also, uh, go and do a campaign, uh, out, um, reach campaign using ai. Uh, and that all that is amazing in terms of sales, but also in terms of healthcare. I think using AI for healthcare is amazing. It can help doctors.

Uh, generate, uh, more and more data easily. Like if they can do, as you mentioned, automation to handle their, um, data when they meet, uh, patient and how they can find the issues using ai. That's also amazing. [00:27:00] It goes on and on. So AI is amazing and it's improving very fast. Sometimes I have difficulty making, um, uh, like following up with the news because every week you have something in you and sometimes you tell people this is the best AI on the planet.

And then in one week you are absolute, there is a new one that is beating it. It becomes like a race. So all of that is amazing and people can use it in a different way. Different way. Now, it can amplify our intelligence. So I will go back that it's not the same anymore. I might call it the singularity of hope because when we, I.

Define the ity is used when machine outsmart us. I tell them yes, and then what? When the machines outsmart us, we'll have a lot of innovation, a lot of things. I tell them, yes, we'll have that, have a lot of innovation. A lot of improvement every day in our society, but not by the machine outsmarting us by using the machine to amplify our intelligence.

So mixing [00:28:00] both. Imagine you mentioned the kid and the businessman, uh, kind of intelligence. One is data, the other one is intuition. Mix them both and you will see miracles. You will see. Someone who has suddenly all the knowledge, but also how to process the knowledge because this is what AI is. It's a machine then can process the knowledge, not only give it to you in thousands of pages.

No can give you really a summary. It can give you an analysis for the text that you can make sense of it. So when the machine learns, it gutters data. But when humans use it, it uses wisdom. It's something we define as well. In terms that sometimes it's, it's very complicated to explain what all human understand, like wisdom.

Hope this term is very difficult to define for a machine, but every human being can understand that, 

Mehmet: right. Sam, you, you, because you've built multiple ventures, I gotta talk about this [00:29:00] collaboration again. Uh, can we see AI as a co-founder? In a sense, uh, I have a very good business skills. Mm-hmm. You know, this is, this is my, my natural skills, let's call it this way.

And then I bring the AI to be my technical co-founder or vice versa. I'm, I'm a 

Sam: I'll guarantee too much credit. That's, that's, uh, but let me tell you something. If you bring a human. Is Master of AI manipulation. This is your technical co-founder, right. Okay. Um, if you see science fiction, I give at people all the time.

So the best example is ves. In, in Marvel universe you have these ves that always assisting Ironman, and Ironman is talking with it all the time. It's so smart. It has all the knowledge and her not asking me to do, asking this machine to do complicated things sometimes because we know it's just fiction.

Uh, there is a human mimicking the voice, right? But we're almost [00:30:00] there. And this is what we'll have. So you will have a better Jarvis, you will not have a partner with you, uh, in the, in the company, but you will have the perfect Jarvis. And that's dangerous. Why? That means you will have not have a technical co-founder, but the lower level of jobs will disappear.

You will have a machine that execute for you. And then you need to judge using your intuition, using your empathy if it's good or not. And this is what I call human AI augmentation loop. Like the machine give you something and then we using your intuition, you don't need data, you judge it. You say, that's good, that's not, that's, I like, that's, I don't like to use your empathy and love here.

And you give them second round and a third round. And this augmentation, we are building a software now in Osis. It can measure that can tell you how good you are using ai. This equation, how good you are using AI amplification. If you, of your intelligence, if you [00:31:00] measure it, you know, am I using it right or am I doing what I call brain obesity?

Which just means you are satisfied with the first answer of ai. You do wanna work. Give you an article. Ah, okay, let's publish it. It's good enough. Instead of reading it and improve it and how you improve it, you use the human intelligence. What's a human intelligence knowledge? You can judge as good or not.

That's, that's by the way, amazing how we can do that. And now some people tell me, but hey, HGI is when this mechanism of human interaction is to meet. Um, no, I don't think so. I dunno, you know how to do it. And I'm trying, by the way, to make a lot of automation in my company, like the other company that is a business, uh, oriented one.

We are trying our best, but we are always surprised by ation because AI is not aware. I. It's giving you something sometimes outta proportion. You need you as a human. Judge it. If you give you a business plan [00:32:00] that look amazing, what means, I don't know, 10 times the amount of people that you have, but you as a human, as expert in your domain.

You can't say that then machine. No. Now will it get it better? Yes, you will have a very excellent Jarvis that you can trust, but you will not have another member of the Avengers. You will not have a few. If you saw the Avengers movie Vision. Like really a human, uh, 

Mehmet: yeah, uh, 

Sam: robot. A humanized robot that is 100% human, which was why weird.

And they are funny in the science fiction to describe it. They told, okay, there is magical things called mind is known that give it that intelligence because computer-wise will not figure it out. So again, maybe this mind stone is a quantum computer. Maybe we'll figure it out one day. Not, not now. We are not there yet.

And I don't think it's five, 10 years things. I think it's much more because of how complicated quantum computers are. So no will not have a technical co-founder, but you will have an excellent Jarvis, you will have maybe, [00:33:00] uh, uh, an assistant that is perfect. I tell people now the age of enter assistant may be gone.

I don't need one. I have an AI who could be my assistant. It cannot be my cofounder, and if I trusted too much, I think I would be doomed. I saw that in my engineers who trust AI sometimes too much, and the hallucination of code. Give them in the loop that they cannot handle anymore. Too much wrong things.

It seems good, but this is, I cannot handle that here. You see the magic of the humans, how we do that. We don't have that knowledge. We don't have that data. Still we can do good and bad. I like it. I don't, well, weird 

Mehmet: specialized agents is if like the, maybe the next thing. Agent. Yeah, it is. 

Sam: It is. And it's already happening.

Like. Sometimes we talk about something and it's already happened. So specialized agent, or let's call it in a very different way, narrow AI yet, [00:34:00] yes. Yeah. Maybe we can call it that. It's, it's degrading for the machine, but maybe it's the truth. Yeah. You are narrowing the context of AI usage, so it give you less errors, helps estimate less.

You are controlling it, but narrowing it. And, and that's, yes. I think that's amazing. That's what's working so far. That's what we do right now. Like you give it really a specialty, a context, a limit. And in this case it's amazing. It's really what you want. You want a specialized journalist. A lot of journalists.

Yes. And some people think a GI would be, the journalists not know it all. Maybe that's a good definition for the humble ai, not the real definition, which is supposed to be human-like intelligence, right? So yes, a super, uh, Asian is, is achievable, will have it, but people will prefer always the narrow, uh.

Intelligence, the specialized agent, because it's safe. You have an automation, you have a neuro intelligence specialized agent give you the result in a deterministic [00:35:00] way, and everyone is happy, so you have the perfect employee. We start by saying, we human, we wanted people to become machine, right? We want the perfect employee that execute orders and never say no work all the time.

Uh, never do mistakes, right? Specializations is the closest thing we have today. 

Mehmet: Absolutely just one thing, which is again, it's my humble, you know, thinking about something you just, uh, came out, which is related to hallucination. I always tell people it looks like, and of course like in, it's like fictionally, when we train these AI models, we train them on the laziness of human beings, which is in our nature, we, we like to find shortcuts, right?

As, as humans. Yeah. We always like. We always like to, to go and find the shortest path for something to sometime. I, I just joke with Frank, I said like, you know, like some of these AI tools, it looks like they try to get things shortcut. The problem is we are also lazy because to your point, like the first, uh, let's [00:36:00] say piece of content that is generated, but let's say by Chad, GPT, I go and, and go and publish it, right?

Because I'm lazy. So sometime even like chat, GPT might be lazy. I notice something now. I think they are doing it in the background. Again, it's a theory, and I'm not like saying this to say like they are bad or good when first Chat GPT start to come out after like couple of months, especially where, you know.

You start to to interact with it and not within the same chat, because I know people will tell me, yeah, you have to open a new chat. Even when you open a new chat. Looks like the performance of the model degradates a lot. A lot. It becomes dumb. It becomes dumb, like it start to not answer you. It start to try to avoid and then all of a sudden we see like, hey, let's go from three to uh, 3.5 to four, and then they start to go from four point, I dunno, they start these weird namings and this, the others did it well.

And then, you know, I noticed recently that. Open [00:37:00] ai, they stopped changing the versions of the, you know, model. They just have release notes saying, Hey, like we just tweaked for example, four O and we added this. And even they did something catastrophic and they have to to, to revert it back. Um, so again, this is my point.

You know, Sam, which is maybe related to, to, to the. To the book about the laziness. What do you think about us being lazy and want to rely all too much on technology 

Sam: first of yes, uh, we are lazy, but that's also a mechanism we have. We call it minimum action. 'cause I studied physics. I know that mechanism is amazing.

It's using the minimum energy of time. Time to, uh, to do a, a task, you know? And this is like part of our way, we think, uh, and it's not how AI thinks, okay? Because this is how we think. Right? Some intelligent people implemented that. Yes, it's true. And I, you mentioned a lot of things. It's amazing. You always [00:38:00] put a lot of questions when a question, but I will start by the hallucination.

So hallucination, let go back to the mathematical model. You have input, have output, and between them you are guessing parameters. So sometimes you have enough parameters, sometimes you don't. When you have, when you don't have enough parameters, you have bad output and you see it. Or when you see the models giving you bad load, right?

But what happen if you have enough parameters or you have too much parameters? That's the hallucination. That's when the machine doesn't know anymore what to do. It has too much parameters, so it can synthesize data behind what it has, you know, because it's cast giving you data that it has in the train.

It's not immediate data. It's transformed. It's in the transformers. It's a new way of representing data, but still it has too much because it has too much parameters comparing to the problem. [00:39:00] This is explanation. What's hallucination is, it's in fact overrepresentation. Too much parameters for the same problem.

And I give sometimes basic example for high schooler. It's when you have an equation of two unknown, but one equation. In this case, you have too much solutions. So what are you doing to give people which one it becomes? Ian comes over thing. Now it mentioned something in the beginning of the ai, and that's, by the way, it's a real thing.

It was when you told me the AI become lazy. Stop answering me. But I will tell you, that was the beginning of the magician giving away his tricks because they are not telling you we program the machine to do that, but they did. Once you interact too much, you are consuming their machine. So they put the program to stop you doing this by stop giving you data.

It's not the machine that is tired or lazy, it's that algorithm. Hey man, stop. We are paying money to run that machine. [00:40:00] So now it become boring. Go away. Come back tomorrow, we'll give you more. And it's true. This is what they are doing. And it was a big scandal recently. You talked about it yourself, but rolling back an improvement.

They made an improvement to cheer people up. AI start giving us too much compliment and supporting bad ideas. People figure it out and they told them, Hey, what? What did you do? Now, even if I give the AI bad idea, it tells me, oh, it's a great idea. And they said yes, yes. We rolled it back. That's only three weeks ago, or four weeks ago.

And they did. So somehow they are admitting, they're manipulating us. They're manipulating the user by casting intelligence and playing with you. So what the cheerful one is just an experienced one drunk? No. People guessed out that, what did you do? You make it like give us good feelings, right? But it's too much.

It's become annoying, it become wrong, so they rule it back. Now we got back to things when you have too much [00:41:00] models, I'll tell you why. Because they didn't figure it out. They're trying to improve their algorithms and then something else happens, and. If you are a computer engineer, you know what I'm talking about, you improve something or something goes wrong.

So instead of figuring it out, which is so complex for them, they start naming them different models. So you have all four, it's great enough, but it's not good in writing. You have a G four and half. Uh, it's good in this four old generate images, but it cannot think so they start putting it on you. The human.

Okay. We give different models and you guests do whatever you want with them because we as engineers, it's too complex for us to do a JI. We cannot do it, or the super agent if you want. So instead of this, we put it on the human. You would use whatever model you want according to your intuition. We don't call it this way, of course.

We'll give it to you and you will use, and I know people will tell me yes for writing, I use the four and a half. They would name soon 4.1 I think if [00:42:00] you are right about the funny names. I don't know what's the, there is no logic, you know, and oh three is good in big data, or oh four is good in math, you know?

Um, and sometimes post three, I dunno if you saw that. It gives you a lot of tables. You don't ask it, but it gives you tables. You know, the other one. It, it will never give you table. Maybe it should be, if you ask it very closely, please put that in the table. Uh, then it give you bad table just to frustrate you.

So going back to the early ai, yes. After 15 messages, me messages, I had the number, it become awful, becomes stupid. Start giving you result. Not because really is tired. It's a machine, but because the people behind it, they don't wanna spend more money. You are, you are cost them a lot. Remember the early ai, it was free by the way.

Now we are paying 20 and there are also 200 uh, dollars a month. 

Mehmet: Yeah. 

Sam: And become arrogant. You know, sometimes I'm challenging these new models with, with things that I know is wrong, and [00:43:00] I know data on the internet is not available. So I asked them some tough questions and they're arrogant. They, no, you are wrong.

Oh my God. So, so they didn't just cast us an illusion and they're lying to us, the big, uh, tech, but they are also creating elegant machine. And don't tell me they cannot do it. I did a small test. I can't tell it, tell you about it. Try to tell, uh, especially open your eye. Uh, give me a joke about a man.

It'll give you, tell them, okay, give me a joke about home and tell you it's not appropriate. Really, I. It? No, someone program it this way. Censorship is programmed into the machine. It's not really intelligence man. They are playing with us and it's not the first time, right? People on social media experience that once they see interaction, they give you more.

So is there a human giving you more? No. It's an AI tool that figures out what we do. And, uh, for, to amplify interaction, it gives you more or less manipulating your emotions. Same with Chad gt, [00:44:00] unfortunately. 

Mehmet: Yeah. Um,

I don't know how we can, you know. Do with this manipulation. So like, are, are we living in, in kind of a big, uh, lab now and then, you know, these companies are, let's, 

Sam: let's address it. I will tell you we need to address these people. They are, they are evil and we need to, to tell them that, okay, like this big, uh, tech companies stop penetrating the humans.

Tell us the truth. Stop using gripe and say AI is doing good. No, it's not ai, it's you. Okay. I like one I will not name anymore. Uh, there is one of the big tech, uh, CEO who came and tell you, Hey, we are doing it ethically, responsibly. I. Really, we trust you, you who used to manipulate us in social media for your game.

Now we trust you again. Why? We need to trust them. So here when come the community role [00:45:00] of really putting rules and, and we want accountability. We want to see what's the code. Know, so we need to see what's the code because, and we did that in our company. We have developed our own ai uh, model out of a three model and out of three model.

When you develop a model, you see how manipulation, they put a lot of manipulation, they, they are not realistic. And when you make it no manipulation, it's not fun anymore. Somehow when people are not engaged, it's a boring machine. Right? That's also an issue, you know? So you think the other one is good because it's well done by the way.

Mehmet: Yeah. Now the question, and you are a serial entrepreneur also Sam, so in this era, how startups can, I would say, find their ways in, in all this. 

Sam: Oh my god, that's a tough domain, especially if you are relying only in technology. I will tell you the future is very different. We don't need to create only software [00:46:00] because this ship has already sailed.

Any software you create today, I will guarantee in three to six months using ai, we are going to engineeringly, uh, reversing it and create it with a fraction of cost. We saw it deep seek, uh, event. It was a worldwide, uh, working out with uh, right $6 million. They replicated what open AI did with the billion of power and they used the open AI tools.

In the beginning, the deep seek used to tell you up open ai, you know, like they didn't hide it, that they used AI tools or to recreate AI tools and that's amazing. So I advise people to have somehow a physical things. A human component in any startup. 'cause if you are purely digital, I guarantee to you someone will copy you within month.

And that's a crisis. And I talk about it in my book too. It's the singularity crisis. It's when innovation would [00:47:00] be very cheap and very quick, it become a commodity. This is new for us humanity. We are used to create. A technology company and benefiting from it for years before it's got challenged with the new tech.

Now, that would happen in March, soon, in weeks. We are not used to that very fast. It's a very different era. It's the singularity, it's, that's what we call the singularity is when things happen so fast every day you have an innovation and whatever you do, someone will challenge it. So I'm advising people that will not happen with biology.

You need lifetime in biology. So if your innovation has a alive things with it and using AI here, you have more, let's call it lead time over other competition. So this is my advice. Don't do pure digital things. Mix it always with physical innovation that will give you some lead, go back to the real world.

Digital will be copied very, very easily. It's a challenging time. [00:48:00] 

Mehmet: Wow. Another example if you allow me, Sam, is uh, you know, they made a lot of noise about Manus. Uh, and then they figured out like it's built, you know, the model is clouded from Tropic, right? So mm-hmm. And all what they did is they, Dr of course, I, it's not, I'm saying it's easy, but you know, they, they wrote, uh, a lot of, uh, automation in the back end, so it looks like, you know, it's doing this fantastic stuff.

So, anyway, 

Sam: um, it's amazing. No, it's amazing. It's, it's, um, it's amazing. Yeah. 

Mehmet: Uh, again, we are not, we are not saying this to underestimate or underrate these technologies, but I, I think people are getting it like we need to follow real innovations like the ones that you mentioned, right? Physical one. 

Sam: So manners for example, it's what I call cursor meet NN somehow.

So cursor is what? It's generating files and revising them. It's funny how programmers would create, they will, will make absolute, the [00:49:00] first job is the programmers. Why? Because they understand it very well. So they're trying to make it more and more automated. So somehow, uh, that a small guy will go, we know from code generation.

That the, let's call it architect or the smart guy, will still need, be needed to create, uh, uh, programs. But the basic programmers that used to create basic stuff, we don't need. We have AI who created, so Manus is the same thing, but not only creating files like cursor, creating any task. Going into the internet and it's amazing.

But at the same time, I show you how hallucination and automation has the limit. 'cause once something goes wrong, it goes massively wrong. You cannot like stop it. And I know, uh, our or not do promotion, but several, let's call it automation company now. Release the new component. If you are familiar with automation, you know the steps like in vague.com or Zapier, they add a new one called Human in the [00:50:00] Loop when you verify the output of ai, and it's a real thing they released just last week because they said, okay, people are tired of one error and instead of one article wrong, you have thousand.

Start to become, you know, uh, a problem. So Manus is amazing for automation. It's called it that task. It's like an, an expert. An expert will never do a small errors, only big errors, right? I. Because he already did all the small stuff, so the new problems, that's difficult for him. But anyway, it's an amazing software when you see it working for the first time, you think you help solve the word problem when they start giving you errors massively.

You said, okay, I'm not using it myself. I said, okay, no, it's too dangerous. I don't mean dangerous, the machine outsmart us. I mean, it makes so much errors if it goes wrong and you cannot throw it back. And I did that. I used it for a advanced version hemo. They gave it to me because also I am, [00:51:00] they considered me early technology adapters, so they gave me access and I created a big website with it.

One time it was a mistake. I couldn't roll it back. What happened? I restored my website from the backup. It went so wrong with ai. It destroyed the website quickly. One error, you know, and done. So that's also, uh. Dangerous over automation with this AI capacity and someone in automation yourself. You understand what I mean?

One small thing goes wrong and what are you going to do? That's what we did. We did the, uh, restore the version of yesterday. I. 

Mehmet: Absolutely. Uh, Sam, I really, you know, I, I enjoyed, you know, the, the, the conversation today before I let you leave. Uh, I know you do like, uh, you know, uh, speaking engagements. Are you coming to the Middle East any soon?

This is my first question. If you have some days you can share with the audience and where people can, can get in touch and learn more, and, you know, of course get, get the [00:52:00] books. 

Sam: So, uh, for the books and my, and let's call it, uh, uh, activity as an author and speaker, you go to sandman.com, S-A-M-M-E-N e.com.

And this is my personal website. I have also social media. You can follow me on my social media. Uh, I produce content all the time. In terms of speaking engagement. I will be in the founder. CXO conferences in Dubai. That will happen in December, every December. So I will be the keynote speaker for the founder conference.

I will be opening the conference and, uh, I'll be happy to engage with people in person there. Uh, in terms of social media, I'm accessible. We have a lot of, uh, active channels from my personal one on. The channel on YouTube and we have all kind of content from content about my fictions to content about the companies that I have created with my wonderful team.

Mehmet: Great. Uh, I'll make sure, I will put the links in the [00:53:00] show notes so people can reach out and, um, we'll be looking forward to, to. See you face to face here in Dubai. Sam, you are here. I really enjoyed. I promised audience we're gonna do another episode, like be, because you know when, when I have guests like yourself.

I always, you know, ask for a second. Oh my God, 

Sam: I'm, I'm, uh, I'm thankful for that. And we forgot to talk about Mars, my fiction, the Republic of Mars. But anyway, maybe that will be a another. Yeah, so this 

Mehmet: will be, this will be, this will be, I promise this will be part two and you know, the disc, I'm sure like by that time in couple, maybe of months as we can do another episode.

So we, we can talk about this. Uh, as I said, thank you very much, Sam, for being here with me today. And this is for the audience guys. If you just discover this podcast by luck. Just, I need a small favor. Subscribe and like, and share to your friends and colleagues, and if you are one of the. People who keeps coming again and again, thank you very much because of your support.

The podcast is ranking in the top 200 Apple Podcast charts across multiple countries at [00:54:00] the same time. So as today, as May 14th, it was in four countries, Malta, Jordan, uh, Slovenia, and Azer region. So thank you very much. I'm, I'm, I'm seeing new countries, so thank you very much. And as I say, always stay tuned for a new episode very soon.

Thank you. Bye-bye. 

Sam: Thank you.