[Music] this robot will take over the world that is what Elon Musk is hoping for with his Tesla bot also known as Optimus a humanlike Android that Elon believes will someday outnumber human beings on the earth and on Mars too now a robot takeover might sound like a bad thing to some but it might also be exactly what we need to save the world more on that later first let's get into what these robots can actually do and how they are doing it this is something that seems to be largely misunderstood which is fair it
was only 5 years ago that a robot like Optimus would only exist in science fiction and now it's real life that's a big adjustment to make here is the largest deployment of the Tesla bot so far the company's cyber cab reveal event also called wi robot on the surface this was supposed to be centered around robot cars but it ended up being the robot people that really stole the show and made headlines not all good ones though there was a lot of misunderstanding going on about how we train robots and that's what this event was
really about tea operation what does that mean in simple terms it means that the robots are remote controlled by people which is definitely something that we were seeing some of the time during the wi robot event but of course nothing is actually that simple there are three main ways to control a robot robot toe operation is one and there definitely is a market for that right now in fields like bomb disposal feeding tigers or cleaning up toxic waste jobs where people are likely to get hurt then there's hard coating that's basically every robot we've seen
up until a couple of years ago the computer inside the bot is running software that was written by a person with a set list of parameters that the robot can follow this has been highly successful in manufacturing and shipping Industries where the robot performs one task over and over this tends to work great up until something unexpected happens that isn't covered in the robots hardcode that's why Amazon warehouse robots have cages between them and the human workers even still workplace injuries tend to be 50% higher in robot equipped warehouses the wait is over our Black
Friday sale is live and so is the complete new collection we have everything from Bold statement designs like the move fast break things crew neck to Retro inspired favorites like the I'd rather be on Mars T and right now you can take 20% off your entire order with Code Black Friday but don't wait these designs are exclusive and stock is limited once they are gone they are gone and the sale ends on Monday click the link in the description to shop the full collection now and don't forget to use Code Black Friday at checkout for
20% off third is robots that are controlled by neural networks or artificial intelligence this is the direction that the industry is rapidly moving in as we speak this is the solution that will actually make robots safe and useful in pretty much any scenario because they'll be able to reason and problem solve and learn just like a human but we are not there yet the thing about AI is that it's not like traditional software it can't just be written into existence by some dude sitting at a desk AI needs to be trained you you have to
show not tell the neural net has to learn from examples so that it can begin to recognize patterns and make connections on its own this is why teleoperation is a huge part of the Opus robot program at Tesla right now the guys with the VR headsets and the motion capture suits are demonstrating tasks for the neural net to learn from think of it like on the job training and the Cyber cab event at Warner Brothers Studio was the biggest robot Training Day ever held some of the robots that we saw were running fully autonomously on
neural Nets others were running on neural Nets that were actively supervised by human operators and some robots were basically just functioning as avatars for teleoperators anytime you see a robot that is walking or dancing this is fully AI driven these are trained actions and we can see the training process of teaching Optimus how to walk over the years in videos released by Tesla we can also see the evolution of of the robot's dance moves then there are the robots that are interacting with the event guests in a controlled scenario like handing out the snacks and
drinks and things Tesla has demonstrated that these actions are trained and can be performed autonomously but when we see the robot going outside of that training like pouring the beer from the tab for example then suddenly their motion gets a lot more awkward and less coordinated which is a good sign that a human operator has intervened to take control this is exactly what we've seen being done with autopilot and full self-driving in Tesla vehicles for years in the early days the car would only be able to perform some limited Maneuvers in highway driving scenarios like
making Lane changes and going around curves anything more complicated would require the human driver to take over then as autopilot trained and improved the car was able to handle more complex operations but the person would still need to be ready to take over when the car got confused now we are quick quickly approaching a level of capability where autopilot can drive all on its own without the need for any supervision or input from the human being at all that took many years to accomplish and training the Tesla bot to interact with the real world is
going to be a similar process it might not take as long because we've got a lot better at neural network training just over the past 2 years but it will still take time and this is why the Tesla Bots that we see out there in the crowd mingling and interacting with the general public were all being remote controlled by people with headsets Optimus had never been exposed to a scenario like that before the bot has been interacting with Tesla employees sure but those are trained professionals who know how the robot works and what it's doing
this giant crowd of influencers and random people doesn't know any of that they haven't been trained on how to safely approach a robot many of them have already had drinks at this point and now they're just having a spontaneous encounter with an Optimus this could have gone disastrously wrong but according to the Tesla bot engineer Milan kovak there was only one incident of a bot falling down in the entire 4-Hour event and that was caused by a purse getting caught on a robot's hand this was what we mean by people not knowing how to interact
with a robot a Tesla employee wouldn't approach Optimus with a bag swinging around on their body but some random person definitely would so Optimus had no training data on how to operate in a scenario like this or it didn't but now it does all of that data generated by the teleoperated Bots interacting with all of those different people is going to be fed into the neural net training process and through that Optimus is going to begin understanding what it's like to operate in a world full of human beings it's going to know for next time
that getting your hand caught up in a purse trap is really dangerous going back to the idea of a programmed robot versus an AI robot it's not that that hard to get a robot to operate perfectly in a controlled environment you can do that with hard coding but getting a robot to function in the unpredictable chaos of the real world is preposterously difficult no amount of computer code will make that happen and this is where Tesla is at right now with full self-driving it wouldn't be that hard to make an autonomous vehicle if you knew
for sure that everyone would always follow the rules of the road at all times and if all roads were built to the exact same standards and all infrastructure was maintained perfectly if that was the case then Elon probably would have been right 5 years ago with his self-driving predictions but that's not the case so it's really hard if Elon is right about the Tesla bot just going out in the real world and buying your groceries or walking your dog then there's going to be a very similar learning curve and what we're seeing right now is
just the beginning okay let's talk about hands one strange reveal at the Wii robot event was what appeared to be the upgraded gen 3 hand of the Tesla bot what we've heard from Elon Musk is that this new hand will increase the robot's dexterity from 11° of freedom to 22 that basically means there are 22 different potential directions of movement between the wrist the thumb and the fingers what we are looking at is the closest that Tesla has come to a direct copy of human anatomy and it's not that Tesla couldn't have made a more
complex more human-like hand to begin with it's more more like they were trying to avoid doing so if at all possible we know that Elon is a dedicated subscriber to his best part is no part philosophy and the way this works is to remove as many parts from the process as you can until you reach a point where you have to start putting things back in to make a product work the way you want so if we know that the human hand is our Baseline we can consider the first generation Tesla bot hand to be
what happens when you take out as many of the parts as possible you basically get two flaps and a thumb moving on to the Gen 2 Optimus hand that's what it looks like when you start adding Parts back in four fingers but still missing some joints now looking at the Gen 3 hand it kind of looks like Tesla has ended up putting almost all of the parts back in three Knuckles on each finger an extra degree of freedom to bend the pinky in towards the middle and the ability for each finger to spread out and
in giving this hand the unique ability to make the Vulcan salute this is all being controlled by tendons kind of like the puppet strings that lead into the hand through the wrist and are being pulled by actuators in the forearm section of the robot this approach is more typical with animatronics like what you would see in a high budget movie or an expensive theme park so does that mean Elon musk's philosophy was wrong well no it just demonstrates that in this particular application more is actually more one of the few scenarios where more capability will
actually make the robot more efficient so the biggest difference between Optimus and other Factory robots is that Optimus is a generalized bot while everything else in the factory is built for a specific purpose this is the argument that you hear the most often that it would be better to build specialized robots for specific tasks in theory this could be more efficient and there is some truth to that a generalized robot can only be truly efficient when it can genuinely do anything not just most things but all things so the reality is that in order for
the robot to be equipped for those edge cases like hand sewing or playing a piano it's going to have to be overbuilt for the majority of cases like carrying a box or cleaning up garbage most of the time when we see the Tesla bot being trained it's on Factory specific tasks like carrying a tray of batteries around but we've also seen Optimus being trained on more generalized skills like folding shirts and putting items on and off a shelf and that kind of brings us back to the whole point of making these robots to do the
jobs that people don't want to do the boring repetitive and dangerous tasks that are essential to a functioning Society not only are there fewer people willing to do these jobs there are quickly becoming fewer people that are able to do these jobs there is a very large number of people currently retiring from the workforce but they're not retiring from society they're definitely not retiring as consumers and over time they will require more services and more labor just to stay alive which is something that people are tending to do for a much longer time these days
so there's a point coming soon not immediately but not that far away either it's a point where the jobs that people don't want to do will simply not get done and the social disruption that this would cause is going to be huge introducing robots into the workforce will also cause social disruption but the sooner we begin to do so the lower the impact is going to be in the long run and that is where we should be looking right now
