AI Powered Humanoid Robots

Rinat: 00:00

It is happening. It is happening faster than you thought. The science fiction that we've thought about or we've imagined even as recent as five years ago, is very close to come to life, and that's what we're gonna talk about today in our podcast with Amit And Rinat, we talk about tech, everything related to tech and today's tech is AI powered humanoid robots. We've talked about humanoid robots before, but the landscape have has changed massively within this short period of time. So we thought, let's tell our listeners about all the advancements that happened with the integration of AI.

Rinat: 00:50

So very exciting time indeed. And we are excitedly and a little bit worryingly waiting for what's gonna come in future on this topic. So let's dive right in. Let's talk about AI powered humanoid robots. Thank you Amit, for coming up with this topic. Um, What's your thought on the tech landscape right now with this.

Amit: 01:15

Thanks Rinat for that amazing introduction. I think you've captured it really well. Things have progressed a lot. The last episode we recorded on humanoids was in 2023. So in last two years, a lot of things have happened and we just wanted to cover it from the perspective of AI.

Amit: 01:30

When we talked about. In 2023, we spoke about the Boston Dynamic robots, the Tesla robots, and few other companies. But now there are a lot more companies who are in the game who are developing humanoids. But why humanoids and why AI is what we are planning to discuss. So I just for the benefit of our users.

Amit: 01:45

I just want to quickly tell first, what is a humanoid and then we can talk about AI humanoids. So humanoid is basically a robot that looks like a human. That has two legs, two arms. It has a face that has a camera. It has the same range of motion, maybe more or less than a human, and it has a torso. So it looks like a human but of course it's not physically exactly the same. It doesn't have skin, et cetera. And most of the robots that we had seen previously, they were remote controlled. So you had to operate them in a remote controlled environment.

Amit: 02:18

So you take them to a physical environment with using a remote control and you ask them to go somewhere and do something, perform certain tasks. That's where it was. And then you had certain intelligence. It's not like there was no intelligence, there was certain intelligence. So it could do only specific task in specific environments. When we talk about AI humanoids, it means you are adding intelligence to it. And what does intelligence mean? It means that you can put it in any environment. It'll learn what is in that environment and it'll learn how to interact or be in that environment.

Amit: 02:46

Say you are in a warehouse, you are told, okay, sort objects. So it'll first look where the objects are. It'll first align itself, and then it'll start interacting with the objects and maybe sort them, pick them up, or move them from one location, one rack to another rack; so it can do all those things.

Amit: 03:02

Now, what we have noticed is that a lot of players are emerging in this software space, but that's not where, that's not where the revolution or the next trillion dollar company will happen. It'll mostly be in the hardware side because you have intelligence on the software, but we have now democratized software.

Amit: 03:20

Creating a software is very easy at the moment, so you can create whatever you want at very low cost. So that will not be where we add value to the human life, we will add value in the physical world. And where is that? So it is places where it's unsafe for humans to go. It is places where we need a cook or help, but we can't afford because we have low income.

Amit: 03:43

Then there are various risky jobs like repairing a high powered electric cable. Or repairing locomotives, et cetera. You can replace all of that with humanoids. Of course, it'll have an impact on jobs, but because they are now in the real world, in an environment that humans can interact with, it is far more useful. I. Compared to a software. Because a software might be useful to specific use cases, but a humanoid can be useful in a far more broader use case, which is much more tangible. So whatever intelligence we are now developing, imagine you club that intelligence with a robot and that is what you get, and that's what's called a AI humanoid.

Rinat: 04:24

Wow. That's very interesting. You've touched on quite a few things there. First of all, it's really good to distinguish between different kind of humanoid robots and what it was before. And what it is now and what is what it may going to be. So that's really good to know.

Rinat: 04:39

We talk about these kind of AI and robot and all of these kind of automation related topics a lot. And as soon as you were talking about AI and the sort of learning ability of a humanoid robot, I was just thinking, wow, this is so relevant with AGI that we talked about before, artificial general intelligence. And I, one of the examples I gave at that time was that a robot or an AI or an a digitally intelligent thing can play chess very well. But if a piece of. Chess, a piece of pawn is picked up and thrown outside of the board. It wouldn't know what to do next.

Rinat: 05:16

And with what you just said now. That is the kind of that is where we are heading. And so far I've read a lot of articles and a discussion about artificial general intelligence being software based. But then [00:05:30] that's another insight that you just talked about, is that where the biggest impact will happen is in the physical world because that's where there is a good room for improvement there.

Rinat: 05:41

And yeah, it's very exciting and if you're not absolutely focused on all the [00:05:45] things that are happening, you're gonna miss out. So definitely keep listening and keep tuning in. Now, as we talk about the abilities of AI powered humanoid robots. Obviously we mentioned humanoid robots are [00:06:00] robots who look like humans. But now I.

Rinat: 06:01

What we can achieve mechanically with only 206 bones. Robots have to have a lot more intricate engineering to achieve the same kind of motion and movement, which the robots are doing. [00:06:15] And in, in some cases they're doing even better.

Rinat: 06:17

One of the things that we all 8 billion of us have and walking around with it is the most sophisticated technology there ever was, is our brain. And that's the most [00:06:30] difficult part. Even now, you might talk to talk to chat GPT or other AI systems and think, oh wow, my brain isn't close enough to all the things that it knows.

Rinat: 06:40

So they, but all of those knowledge are actually virtually stored in a cloud. [00:06:45] But to put all of those processing power in a brain sized space is the biggest difficulty when it comes to AI powered humanoid robots. How is this challenge looked at and what [00:07:00] are the companies doing to solve this problem?

Rinat: 07:02

Yeah, there are ai, they are very efficient AIs out there, but they're all digitally located in a cloud, in a, utilizing massive amount of processing powers, and that's how they achieve that. But in [00:07:15] order for a robot, human-like robot to have AI power, it has to fit all of that technology in within its body, it doesn't have to be head because it's not a human, but within its body as well as all the [00:07:30] mechanical connections, as well as a massive battery that it needs to power itself. Where would it store the AI powered technology?

Amit: 07:39

That's a good question and I don't have an answer for that. I'll be very honest with you, but I think what's happening is [00:07:45] we are looking at staged development, just like we saw with large language models. Now, I think you've touched a very important point of the brain and let's first talk about the human body itself.

Amit: 07:54

If you look at the human body, we are very fragile. But we we have a lot of strength. We have a lot of [00:08:00] endurance, and we can heal. Robots will have to be repaired. They cannot auto heal, at least not yet with the current technology that we have. So they cannot regenerate skin, they can't repair any of the broken parts, et cetera.

Amit: 08:11

So we have humans have the capabil capability that we can heal when it [00:08:15] comes to brain power and when we look at intelligence. So let's break down the intelligence. You receive certain inputs from your different senses, like you, from your eyes, your nose, your ears, your mouth and your touch. So you receive certain senses.

Amit: 08:25

You process that sense. You process that information in your brain and then you come up with an output. [00:08:30] But you do that very fast and you do it constantly. And the same thing has to happen with a robot. If you look at the self-driving cars, they are having cameras and lidars or radar sensors and they're constantly monitoring the environment and the moment they identify certain objects, then they will [00:08:45] quickly take the necessary action.

Amit: 08:46

Here, they're not using language as an output, they're using action as an output, so they're doing something instead of saying something. So that's the first difference. Large language model versus. This so they don't have to understand the entire coding library. They don't have to understand [00:09:00] the entire encyclopedia Britannica. They don't have to understand the entire internet. They just need to understand, sorry, your.

Amit: 09:05

They don't even have to understand how to speak.

Amit: 09:07

Exactly at least not at this point in time when they, when humanoids have to interact with humans at home, they would need to under, they would need to learn how to speak. And if I [00:09:15] ask it to do something, it should be able to respond to me saying that, yes, it has understood what I've asked. And then it performs that action.

Amit: 09:21

And then it says, okay, is there anything else? Or do you want me to go sit down and charge? So you've coupled multiple aspects, so that's why I want to break it down. So firstly, it's action. Instead of language. Language communicates a lot of things, but it's actions that they need to replicate. So if you look at Boston [00:09:30] Dynamics, it was mostly focused on action. It was not focused on language. It was trying to figure out, okay, this is the environment where I have a lot of obstacles, how do I interact with those obstacles? And if I've learned that, then it's easy for me. Wherever I go into a new [00:09:45] environment, I can try to adapt to that environment using the training that I've got. So that's the first thing.

Amit: 09:51

The second thing is when it comes to inference, the brain is trying to infer something based on the sensory in input that we've got. To do the [00:10:00] inference, it requires very less power. But for the compute that we are currently talking about in large language models, it requires a lot of power. It requires a huge data center, so the cost to generate a token. It is very high, and the energy used to generate the [00:10:15] token is very high. But with with the human brain, it is very low. And that is what we are trying to replicate. How can we generate the next thing the next word, at a very low cost? Because our human brain. It doesn't heat up when we are talking or when we are thinking, when we are doing things right. It [00:10:30] has got that context.

Amit: 10:32

The other thing is the energy usage. So humanoids will have to be charged. So as you said, they'll be electrically powered. So there will be the charging constraints. Where do we carry the battery pack? If I want to have an environment in [00:10:45] my house, do I want it to carry the battery pack everywhere in the house? How do I charge it so that aspect is there? And there was the brain. And then what was it?

Amit: 10:51

And the last thing is context. So humans have a good grasp of information that they have on which they have specialized. Say for example, you're a mechanical engineer, [00:11:00] but you are not a computer engineer, I. So you would have a good grasp of mechanical concepts, but not coding concepts.

Amit: 11:06

So you will struggle with it even though the information is out there. Plus you have information about mechanical, only in certain areas . Not [00:11:15] all aspects of mechanical. Say you're specialized in robotics. But not in automobiles or you're specialize in automobiles and robotics, but not in constructions or manufacturing.

Amit: 11:23

So you have limited ideas there. But with a robot or with a AI model, what you can do is you can train them across all fields of [00:11:30] mechanical. So it could be construction, robotics, manufacturing hydraulics, anything, control systems, CNC machines, lathe machines, et cetera. So you program them. And they have this context and they have this connection about everything.

Amit: 11:44

The problem with [00:11:45] us is we cannot grasp all this information and then connect the dots. That's why geniuses come, because they sometimes can connect the dots across multiple domains of knowledge. Okay, but it's very difficult. Geniuses can, but regular human beings, [00:12:00] it's very difficult to connect the dots over a large expanse of information.

Amit: 12:04

But robots, because if they're AI powered, they will be able to connect a lot of things. So they might be able to connect a flight. With formula One car and then maybe come up with a new [00:12:15] way to operate these different things. And same with manufacturing and other things. So there are a lot of these aspects that that AI humanoid robot can do.

Amit: 12:24

But what how is it storing and how is it inferring? I think at the moment it is just focused [00:12:30] on action rather than language. So it doesn't need to know everything around.

Rinat: 12:35

You've mentioned Boston Dynamics. And last time when we spoke about humanoid robots, we covered a lot of what Boston Dynamics are doing. We're a big fan. We still are. [00:12:45] But nowadays what I, what I'm looking at is there's so many startup companies and various initiatives who made a lot of progress in terms of integrating AI and humanoid robots. Boston Dynamics is just one of many now, [00:13:00] and all of these different things that they're doing are very impressive. And some of them are environment dependent, so they for example robots are working in warehouses where they have set points where they have to go to carry out a particular task. And some of them [00:13:15] are not even depending on environment and as you said, can be trained and, understand context of where it is and what to do based on its input information that it collects from lidar and camera, et cetera. So what's the latest [00:13:30] advancement, is there any startup that's now more advanced than Boston Dynamics or

Amit: 13:35

there are a lot of companies now in this space, apart from Boston Dynamics. There is a company called One X, Figure, Humanoid, Unitree, and there many other companies and they are all building humanoid robots [00:13:45] and the way they're trying to solve it is that, okay we know AI is coming. We know that we need robots. How can we build the next robot that can do multiple things?

Amit: 13:54

So take for example you have a big police force in your country. But the problem is [00:14:00] you can't recruit enough people and your population is increasing and nobody wants to join the police force. Now you want to replace that police force, but you don't have people to replace them. Robots are a good example to do that.

Amit: 14:11

The next thing is you need a lot of drivers. So that drivers is, we are eliminating [00:14:15] that and we are getting driverless cars. The next thing is you need house help. Because you have a busy family, you are working, your wife is working, you have children, you don't have the time to cook, you don't have the time to clean the house. So that's another area.

Amit: 14:27

The other aspect, as you mentioned, is the [00:14:30] warehouses. So there are companies like Figure, that are trying to figure out, okay. Robots are working in this environment. So how do we optimize it for that? Now the other thing is you are not having one robot. You're having hundreds and thousands of robots, and these thousands [00:14:45] of robots have to be aware of what everyone else is doing inside the warehouse. They need to be able to communicate with each other and coordinate.

Amit: 14:53

Say you want to lift a heavy object, you can't lift it by yourself. A robot can't lift by itself, [00:15:00] a human art robot, and you don't have a crane. So you need two robots to coordinate together to lift a box in a warehouse, take it out from a rack and put it into another rack. Simple example, and in which two robots have to coordinate and [00:15:15] communicate. So how would that interaction work. So that problem is also now currently getting solved. Because most of the examples that we saw when Boston Dynamics was there is one person, one robot, trying to do one single task. Now it's multiple robots [00:15:30] trying to work together to solve one task. So that's the next step.

Amit: 15:34

The other is self-defense. So suppose you want to have a security guard. Security guard is a very boring job. You have to wake up the whole night and you have to make sure that you are alert, and if someone attacks you, you can defend [00:15:45] yourself. And also subdue the attacker down right.

Amit: 15:49

Now, what if your robot could fight? So we have now seen examples where a robot could actually fight, can do some kung fu moves. It can keep its balance while doing the fight, while recovering [00:16:00] from a kick or a punch. And then try to put down the attacker. So that's another example. So there are various aspects of our lives where robots can play a very important role.

Rinat: 16:12

Interesting problem and even more interesting [00:16:15] solutions. So what you've said, I. At excites what you said made me excited, but also made me quite worried about the last thing you just said. One of the things that was a safety net for me as I was finding out about new advancements in AI and robotics is that [00:16:30] there will always be this underlying code within the robot that do not harm any humans. And that was like a final safety net. But now what happens if these robots are in wrong hands and then they [00:16:45] can use that to harm regular human beings .

Amit: 16:49

But that's already happening, but that's already happening in many things, right? There are rogue agents everywhere in our life. The rogue agents can take advantage of any tool and they can use it to kill or harm any [00:17:00] human being. So for example, you have a car, I. The car you use to drive to take your kids to school or to buy groceries.

Amit: 17:07

Very simple, useful tool. A simple thing that everyone uses, but the same car can be used by thief and they can run over [00:17:15] some people I. Same tool, different purpose. So it's already happening with us. Of course, there will have to be safeguards. So I'm thinking about password control or voice activated or offline devices that just the AI on their on their on their physical body and it can't [00:17:30] interact with outside the world or internet. And only it uploads the data when it's needed, say during downtime hours when it's quiet, et cetera. So I'm pretty sure there will be safeguards in place and we need not worry too much about that. And I'm sure people are thinking about [00:17:45] this, but you're right, that's a valid concern.

Amit: 17:47

What if it's comes into rogue hands who can use it to harm other people? But that's already the case with any technology that we can think about. People are already using it to harm people. So it's all about education, it's all about control. So how do you [00:18:00] safeguard, say, selling a knife to a person who can use it to kill someone?

Amit: 18:04

Like we know that in London there are a lot of cases of knife stabbings because we don't have a gun culture here in the UK. For people who are not aware, we don't have a, we can't buy firearm that easily as you can do it in the US, but we can, [00:18:15] what we can do is we can buy a knife very easily and these knives can be very dangerous.

Amit: 18:18

And there has been a lot of reports of knife stabbings in the uk, especially in London o over the last couple of years. It's under control, but it's still quite a lot right? Compared to say [00:18:30] shootings like you have in the US. Now there is a device, it can be used to do useful things, but then people are buying it to harm others.

Amit: 18:39

And there, of course, police are now saying that, okay, we need to check everyone's license and we need to check if they are [00:18:45] adult, if they have any criminal records, et cetera. So there are safeguards in place, like with any other things. So the same things I'm guessing will happen with humanoids.

Rinat: 18:53

Okay. No that's a very good point. Yeah, it's already happening. It's just because it's humanoid and it's a lot more advanced [00:19:00] than the whatever current technology we have. And I'm probably just thinking like science fiction that everyone, or half the people are walking around with a humanoid robot following them, but that's not gonna be an overnight

Amit: 19:12

not gonna be overnight. It's, yeah.

Rinat: 19:13

very slowly and [00:19:15] within that time the legislation and enforcement will catch up and maybe they will also have various robots and then they will protect the malicious robots. And we'll have awesome, we'll see, awesome robot fights. But aside from that, one of the things. That.

Rinat: 19:29

One of the other things, one of [00:19:30] the macro things that came to my mind and now I'm just I'm focusing on the societal impact of this technology. The scene that comes to mind is in Ironman one and two, when Ironman created that suit, government was very [00:19:45] interested.

Rinat: 19:45

They wanted that technology and when Ironman didn't wanna give it they forcefully took it, but okay that story aside, this technology as it advances, this is not gonna be like chat GPT or other digital AI tools that [00:20:00] everyone can sign up and start using.

Rinat: 20:02

If you're a small. Business who does manufacturing, you probably wouldn't have enough budget to buy a lot of robots, but if you're a big business like Amazon with their warehouses, they'll probably [00:20:15] overnight replace and buy all the robots.

Rinat: 20:18

So the wealth gap in a society will increase. This is opposite of what a software based AI is doing because software based AI, everyone signed up and barrier to [00:20:30] entry to any project was eased because of availability of knowledge Chat GPT and similar technologies provided.

Rinat: 20:37

But with physical robots, they're very costly and initially they're gonna be ex out outside of reach of even medium sized [00:20:45] businesses. Only government and massive conglomerates will be able to afford it, and as soon as they afford it, they will become. They, big businesses will monopolize the whole market because their efficiency will be a hundred times more than any com [00:21:00] competitors. So that seems like a negative impact of this technology. And I don't know how to what could be a solution for this?

Amit: 21:10

I think that you have raised a couple of very interesting things. First is the weaponization aspect. [00:21:15] So if you talk about just robots becoming weapons as you mentioned the Ironman example, I think that's a very relevant thing and I think it's good to have robots as soldiers because then you are saving human lives. Essentially robots fighting robots. The best thing is human lives are

Rinat: 21:29

[00:21:30] Yes. That's only if robots are fighting robots.

Amit: 21:33

I'm guessing yes. I'm guessing that, you wouldn't want to deploy robots, fighting humans. That would be an unfair advantage, but I'm guessing some states might do, some state actors might do. Yes, [00:21:45] there is always the scenario where there are states who take advantage of the situation just because they have that upper hand, because they have more robots.

Amit: 21:52

The other thing is, the wealth cap. The wealth gap. Let's first break down why there is a wealth gap between software and hardware. So if I want [00:22:00] to make a smartphone. Okay, so let's say I built the first smartphone.

Amit: 22:04

It's very expensive. Now I need to make a second smartphone. It'll be still expensive. It'll not be very cheap. If you look at software, I make a file or a content. Then I just [00:22:15] do control C, control V on a Windows machine, and I immediately create a copy of it. Digital content, the copy part is very simple.

Amit: 22:23

That's why it's easy to scale, and that's why a lot of big software companies have been able to scale because the cost of [00:22:30] replicating is zero. But the cost of replicating a physical product is there and it's tangible, it's significant. And then you talked about that initially it'll be only owned by big businesses i. Till the cost comes down. And how does the cost come down? It comes down by usage. When there is [00:22:45] more demand, so let's say for example, Amazon had huge data centers which they were using for serving their customers. Then they realized they have expanded too much. They have spare capacity.

Amit: 22:55

So what do we do with that spare capacity? They release it as cloud. That's where you had [00:23:00] AWS, that's where we had the first cloud, and that's where the people started using it. It created more demand. And it created companies like Microsoft Azure, Google Cloud Google Cloud platform, et cetera. And with that, the cost of hosting on cloud came [00:23:15] low.

Amit: 23:15

So initially what you saw is industries using it for their own advantage, like a big data center, using it for their own good. Then they had extra capacity, so they leased out that extra capacity, and then the cost started falling down. I'm guessing the same thing will [00:23:30] happen with robots. So the companies, they will do, they will have a lot of robots initially, then they will have extra robots, which are spare because they have optimized their business operations.

Amit: 23:39

So with that spare capacity, they'll start leasing out. As they start leasing out, people will start renting it initially, [00:23:45] because in cloud you don't own anything, you rent. Basically, you pay a monthly subscription. So similarly you will rent a robot and once you rent a robot, you realize that it's good, so you don't have to own the robot.

Amit: 23:54

You can just rent a robot and it's affordable. And then because more and more people [00:24:00] rent the robot, the demand increases and the cost comes down. It's a very simple example and it has happened in our, in front of our eyes. I'm just connecting the dots.

Rinat: 24:07

It does make sense, but I don't know if I can fully agree with that. Even using your example, yes. In the process of Amazon providing the [00:24:15] cloud service, creating for its own, go own usage and then outsource, selling it, in that process, Amazon made a lot of money. And it killed a lot of comp competitions it had at that time. So those businesses had to, had no other choice but to die out. [00:24:30] And then a new kind of industry or a new landscape appeared based on all of that. So now, whichever company has access to enough wealth to buy the first batch of robots, they will have surplus robots at one [00:24:45] point, and then they will rent it. And in this whole process, they will probably become the biggest

Amit: 24:50

Same thing happened with Apple. Same thing happened with many other companies. So they were the first who came up with iPhone, the first actual smartphone, and they are now the biggest [00:25:00] tech company in the world. I. One of the biggest, I'm not saying the biggest, so it's the first mover advantage.

Amit: 25:04

And most companies will have, and you're right, if you look at the companies that are going to buy these robots in bulk, but maybe not just buy, but who are the companies that are manufacturing it? Because you'll have to be able to [00:25:15] manufacture this at scale. And we are not talking like. One one robot or two robots.

Amit: 25:18

We are ma talking about a car for every consumer. So a robot for every consumer. You're talking about billions of robots. For that you need to solve the problem of scale. You cannot just build 1 million robots. You have to solve [00:25:30] the manufacturing process of how to build these robots. How to make sure these robots are safe, deploy them at scale, et cetera.

Amit: 25:36

So that. Part, whoever solves that first, they will definitely have that advantage. And we have seen this with any industry. If you look at the Ford Motor Company that came out with the [00:25:45] first car, the first proper car that was scaled to production it must have killed so many jobs, which were dependent on horse riding, right?

Amit: 25:52

Because the earlier locomotive was a horse cart, so that industry died and then a new industry emerged. So the same thing will [00:26:00] happen. So there will be some. Industries that will definitely be impacted, which is with case with any other technology. Like when Apple came, Nokia started falling down, Blackberry started falling down, and we don't hear about these companies, which were a big name previously.

Amit: 26:12

Now there is no mention of [00:26:15] these companies, right? So it's always the case, and this will happen with these places as well. And that's why there are so many companies now who are actively involved because now you have intelligence and now you have a physical capability of building these robots. Whoever makes these robots, they'll be [00:26:30] the next trillion dollar company. It'll surpass apple, it'll surpass any other company that we currently know of. Because you, a robot will be able to do any menial task. So say for example, we I went to a garage recently to get my car fixed. And in the process I saw [00:26:45] that. What are the conditions in a garage and what the humans are working on, and they have to do a lot of repetitive task. Now, imagine a robot has to do that same thing, for example, a cook. So you have a house cook, the cook has to cook the same thing again and again. Or you have a maid servant that comes and cleans your [00:27:00] house.

Amit: 27:00

They have to do the same thing again and again, maybe not just in your house, but multiple houses. You replace that with all the fleets. Now it's not like we are not going to need these people. We are going to need these people, but these people can be easily replaced by robots, and that is where the, that is where the large scale job losses will come, it'll not get [00:27:15] replaced by AI. It'll get replaced by robots. So these low wage jobs, you have construction workers, maid servants, cooks, factory workers, et cetera. These jobs will get replaced once these robots are there and they will disrupt the market like crazy because currently you [00:27:30] have high cognitive task jobs and low cognitive task jobs. The low cognitive task jobs are far more than high cognitive task jobs, and that is where the disruption will happen.

Amit: 27:41

AI will of course a So I was recently thinking about this after listening to a [00:27:45] podcast knowledge work and knowledge worker. So a knowledge work worker does the knowledge work. The knowledge work could be, creating a software code. The knowledge work could be coming up with a new marketing campaign.

Amit: 27:52

The knowledge work could be coming up with a new design. The knowledge work could be building a website, et cetera, right? Knowledge work and who's doing that knowledge work? A knowledge worker. So now the knowledge work will be done by AI instead of the knowledge worker. [00:28:00] So whatever work the knowledge worker was doing, we, AI has replaced that.

Amit: 28:03

So knowledge worker now, has to do knowledge work that can't be done by AI, it has to perform at a higher level. So you have currently increased the barrier to perform higher cognitive task. [00:28:15] So if you can't perform a higher cognitive task, your job will be replaced by someone that can do a lower cognitive task, and that is AI.

Amit: 28:23

And the same thing will happen with robots. So robots will do replace the the the, these low wage jobs. It'll also replace surgeries. So routine surgeries like knee replacements [00:28:30] or keyhole surgeries. It's all will be replaced by robots. So then you need robots, which are specialized.

Amit: 28:36

So higher cognitive task, and the same thing applies to humans and people are not realizing it, but this is coming. This is the next wave that people [00:28:45] are not thinking about. Knowledge work, knowledge worker, higher cognitive task, lower cognitive task. If you break down things in this way, then you understand what jobs are going to be replaced, what jobs are not going to be replaced.

Rinat: 28:57

Yeah, no, that's actually a very good [00:29:00] insight on what's happening and what might happen. That is a very thought pro. Provoking view on, on, on this topic and something that I'm going to think a lot about after this. And hopefully our audience is also intrigued by this conversation 'cause I [00:29:15] certainly am, there's a lot of a lot of sort of topics to think about as. As it happens in front of our eyes, we are very close. We're getting closer every day. We're dangerously close, one might say to having a serious breakthrough. Think about Deep Seek . That the main point of why it became so popular is because it could achieve the same or similar or sometimes even [00:29:30] better results than Chat GPT with a lot less power and processing requirements. So as we have more and more breakthroughs like this, we could have, a self-contained AI humanoid robots, which doesn't [00:29:45] need to outsource its processing online like Alexa does nowadays. It won't need that. It will have a self-contained decision making ability and context understanding ability, everything [00:30:00] within that package. And if all of these robots are behaving and, supporting their owners. Then, as you walk out into the park, you'll see various robots, behaving on behalf of their owners and that [00:30:15] would be a true science fiction. And I'm, I dunno how quickly I want it, but I definitely want to experience it in my lifetime.

Amit: 30:23

You'll definitely experience it in your lifetime. I think. Two other things that I want to mention is that these robots, they can't be built without [00:30:30] materials. And without chips. And if you look at Elon Musk, I think the strategy behind Ukraine is to get the rare earth minerals from Ukraine. We are not talking about political aspects, just bear with us. We are not endorsing any opinion here. [00:30:45] We're just trying to figure out and connecting the dots as we speak. So what I'm guessing is Elon Musk has already realized that if we want to build robots, we need chips. And if we need those chips, we need rare earth materials. And if we need rarer materials, which are the countries which we can negotiate [00:31:00] with. One is China. They can't negotiate with China. The other is maybe Ukraine. And they're trying to see if Ukraine can give access to those materials. Because in order for us to build these robots, we need those chips that can be programmed with the intelligence.

Amit: 31:13

And to build those chips require a lot [00:31:15] of raw materials. And that raw materials has to be physical. It can't be a digital raw material. So that raw material, first it's needed and then you have to scale it. So you need the manufacturing process in place so that you can scale it.

Amit: 31:26

It'll be like robots will be building robots, but it won't be [00:31:30] humanoids, building humanoids. It'll be robots like like we see in the manufacturing process, a single arm robot like you have seen in on the factory floors. The humanoids are the built products.

Amit: 31:39

And then the second part is the energy. So these robots will require energy, and these require robots will require charging. [00:31:45] Just like our cars require fueling, electric cars, our phones need charging, these humanoids will need charging, so we'll need to have a charging infrastructure if we want these humanoids to work 24 7 around ourselves.

Amit: 31:58

Inside the house, there should be a charging station outside the house. There should be charging station inside the warehouse. There should be [00:32:00] charging station. So say you have a fleet of a hundred robots. Out of that, maybe 80% will be at full capacity and working 20% will be charging. So you'll have a downtime.

Amit: 32:09

So if you want that a hundred percent efficiency, you need maybe 120 robots to compensate for the [00:32:15] 20% charging at any given point in time. So you have to think about all these things. So you cannot just have a hundred robots and thinking you'll get a hundred percent efficiency. No, it'll not be the case because some of them will be charging themselves.

Amit: 32:25

So you'll need fast charges. So you need to have that charging infrastructure and these robots have to go and sit somewhere to [00:32:30] do the charging. So all these problems have to be solved when we think about scale. And these are the challenges, which now companies are thinking also. And there will be a lot of companies in this space, like how do we rent robots?

Amit: 32:43

So there might be a company [00:32:45] that's renting out not Airbnb style houses, but renting out robots. So there could be the next big platform. Maybe we should build that Rinat.

Rinat: 32:55

That is ahead of time idea. And why don't we just have the platform ready when the time is [00:33:00] more appropriate. That would be the one that's first in the market. Definitely a good idea and we should definitely talk about it. Yeah, audience, if you, one of you are interested, reach out and maybe we can do partnerships.

Rinat: 33:11

You've answered me before about the wealth gap, but I wasn't fully [00:33:15] convinced at that time, but now you inadvertently said some of the things, which actually convinced me that actually the advancement of robotics in this way will also create a lot of entrepreneurship [00:33:30] and a lot of opportunities for small businesses to flourish and grow as well. How? It's from my own life as I started my career as a mechanical engineer. The first company I worked with, they had just one customer, which was Rolls Royce. So this [00:33:45] was a small business of about 150 people. It's still a small business and all their task is to supply to Rolls Royce, a big company. All of these this small business, but obviously the owner is probably a millionaire and [00:34:00] creating 200 jobs. So all of these are created because of this large company, Rolls Royce. And then my second job. One of their biggest customer was Jaguar Land, land Rover, and I've also even designed a lot of things that was [00:34:15] supplied there as well. Again a massive company, but there are a lot of dependent, smaller company which were only possible to flourish and grow and exist because they were supplying to that big company.

Rinat: 34:28

As you mentioned, with robotics, all of [00:34:30] these other things that are req required charging stations. So an electrician who does house related. Electrician's job, they can easily transfer their skills to install cha charging station and repair robots and all of that. So there is a lot [00:34:45] of transferable jobs and opportunities that are gonna become available.

Rinat: 34:48

And then. Yeah, a, a company with the next trillion dollar company will be the robot manufacturer, the final seller of the robot. But then they will buy at scale batteries from one place, [00:35:00] and then motors from another

Amit: 35:01

Raw materials. Electric cables, actuators, various things, even chips. They may not be manufacturing the chips themselves. They'll need someone like say, Intel or TSMC to manufacture the chips and provide them. So you're right. I think I [00:35:15] also didn't realize while talking about it, but I think you hit the point. Because Yes. But the same thing, we can't say about AI because AI is happening in a digital world.

Rinat: 35:23

Yes. Actually, from that perspective, AI is making a bigger gap, but which one's making a bigger gap or which one's shortening the [00:35:30] gap is obviously too big to calculate and remains a controversy and up for intriguing discussions, which we are all up for. If any of our listeners reach out, definitely do with a potential topic or anything that you want to [00:35:45] discuss.

Rinat: 35:45

I think the main differentiating factor is people who take actions and people who wouldn't. AI in digital world, you could learn and yes, obviously in third world countries the barrier of entry [00:36:00] is a little bit higher.

Rinat: 36:01

But aside from that, if you have access to internet, it's up to you to I. Utilize the freely available AI models that are there. And in future when you'll be able to [00:36:15] potentially rent a robot it would be up to you to decide whether you wanna invest on that or instead something else.

Rinat: 36:22

Yeah I think it's, say for example, if you're an electrician, it's up to you to think about what could be the charging ports [00:36:30] that a robot must need. And you might patent that before anyone else and become a trillionaire yourself. So there are opportunities for you right now that would decrease the wealth gap. So I think taking and not [00:36:45] taking action is probably the biggest differentiator. It is a very interesting topic Amit and a lot of the things I didn't know and thank you for opening my perspective in so many different ways.

Rinat: 36:56

Hopefully our audience enjoyed this topic very [00:37:00] much. Yeah. Thank you everyone, and we wanna see you next in our next episode. Thank you.

Amit: 37:04

Thanks everyone.