Skip to playerSkip to main content
Humanoid robots throwing punches in stores? AI outages everywhere? Get the wildest updates on the future of tech, from robot showdowns in Russia to Tesla’s new Cybercab and data-collecting apps training our future robot roommates. Dive into the chaos (and fun!) of tomorrow, today. Don’t forget to subscribe and tell us in the comments what blew your mind the most! #robots #AI #technology #future #news

👉 This channel was created in collaboration with https://www.youtube.com/@realidadeimpressionante


The Inspired Business Technology by RomanSenykMusic

0:01 - Introduction: Robot Incident in Russia
0:50 - Humanoid Robot Incident Details
4:02 - Other Recent Humanoid Robot Safety Incidents
8:07 - Risks of Humanoid Robot Control Failures
9:33 - Hyundai's Humanoid Robot Plans and Labor Reactions
12:18 - Transformable Chinese Humanoid Robot: T01
14:29 - Simultaneous AI Service Outages
16:41 - Tesla Cybercab Launch and Figure AI's Data Collection


Category

🤖
Tech
Transcript
00:01A man pushes a humanoid robot inside a store in Russia. Seconds later, the machine takes a
00:07fighting stance and starts attacking. Employees have to run to try to restrain it. And the
00:12strangest thing is that this isn't the first case. In recent months, we've seen here on our
00:18eye other humanoid robots that have also been filmed going after people or showing somewhat
00:23violent behavior. So what could be happening? Are we just seeing isolated failures? Or are
00:32the robots that are increasingly sharing our spaces revealing a safety problem that's much
00:36bigger than we imagined? Let's try to understand what's going on out there and a few more things
00:42about new technologies. A video recorded by a security camera in an electronic store in Russia
00:54is drawing a lot of attention after showing a humanoid robot apparently reacting physically
00:59after being pushed by a human customer. The incident happened in Saratov, a city located
01:04in the southeast of Russia. The robot, called Syoma or Sema in different naming traditions,
01:10was used in the store as a kind of technology consultant to interact with customers. In the
01:15released footage, the man approaches the machine and initially tries to greet it. Shortly after,
01:20he puts his hand on the humanoid and pushes it. The robot steps back slightly, but then its behavior
01:25changes. The humanoid takes a stance similar to those used in martial arts. Then it starts performing
01:32a sequence of kicks toward the customer. Store employees quickly try to restrain the machine. Even while
01:38being held back, the robot continues moving its legs and some of the blows apparently hit the
01:42employees who were trying to immobilize it. The sequence lasts for about 20 seconds. After that,
01:48the behavior ends. Suddenly, the humanoid robot raises its arms, taking a position similar to a T.
01:55The customer apparently did not suffer any serious injuries, but the video quickly spread on social media
02:02because it gave the impression that the robot had acted deliberately after being provoked. And it's
02:08precisely at this point that there is a very important difference between what appears in the
02:12footage and what can actually be confirmed, you see. There is no evidence that the humanoid robot
02:17identified the push as an act of aggression and on its own decided to attack the man. It's also not
02:24clear which system was controlling its movements at that moment. According to futerista.com, the robot seems to
02:29be based on Unitreat Nitrogim. And I can also confirm that based on my experience, the robot looks like
02:34that Chinese robot that can be programmed to perform complex movements, including martial art sequences.
02:39But the footage doesn't allow us to determine whether the behavior observed in the store was performed
02:44autonomously. One possibility is that someone was controlling the robot remotely from behind the cameras.
02:51Another is that the push or some interaction triggered a sequence of pre-programmed movements.
02:57And there is also the possibility that the whole situation was planned as part of a promotional
03:02video for the store itself. And look, this hypothesis gained a lot of traction because the account
03:08associated with the establishment has already published other humorous videos using this same robot.
03:13A detail in the footage also caught a lot of attention. Right before the humanoid robot takes a
03:17fighting stance, an employee seems to touch the back area of the machine. It's not possible to determine
03:23from the video if the contact had any relation to the behavior that followed, you know. But this
03:28episode happens at a time when humanoid robot manufacturers are showcasing increasingly fast
03:32and complex physical movements. Robots like Unitreat G1 have already been shown performing punches,
03:38kicks, kung fu moves, and even fights between machines, you know. It even lost its head by the way.
03:44The video recorded in Saratov doesn't show that the robot decided to defend itself from a person.
03:49It also doesn't prove that there was an artificial intelligence failure in the robot what the images
03:54show is pretty clear, right? A humanoid robot being pushed and seconds later performing a series of
03:59fighting moves while the employees try to restrain it. But that wasn't an isolated case. Okay, there are
04:05several others. Let's take a look at some of them now. In recent years here at Realidade Impressionante,
04:12we've shown different robots that apparently advanced or performed dangerous movements near human beings.
04:18And then when we put these cases together in sequence, a problem appears that goes far beyond the idea
04:24of a supposed machine rebellion like the Terminator. One of the most impressive episodes, which generated
04:29a lot of engagement here on the channel, happened in February 2025 during a festival in China.
04:36The audience was watching the humanoids from behind the barricade when the H1 robot from Unitree
04:40started approaching people. Suddenly, the machine moved toward the crowd. In the footage,
04:45security guards quickly stepped in front of the humanoid and managed to drag it away. At the time,
04:51the video went viral with descriptions saying that the robot had tried to attack humans.
04:56We showed this case here on Realidade Impressionante and raised exactly this question that those images
05:01couldn't answer. Did the robot really try to hit someone? Or did it simply lose its balance and make
05:07sudden movements trying to regain its stability? The local authorities treated the episode as an unexpected
05:14anomaly during security. There was no evidence that the humanoid robot had made a conscious decision to
05:19attack those people. A few months later, however, another video involving a humanoid robot, also from
05:27Unitree, drew even more attention. This time, the setting was not a festival. It was a testing environment.
05:33The H1 appears suspended by a structure while two men work near the machine. At first, everything seems normal,
05:38until the robot starts to move its arms and legs violently. The movements quickly intensify. The
05:44structure is dragged, the monitor falls, and the men have to step back while they try to regain control
05:50of the machine. And we recently showed these images again here on the channel because when seen without
05:55the exact context, they seem to show a robot in a real state of rage. But again, there's a fundamental
06:02difference.
06:02You see? That wasn't an artificial intelligence deciding to attack the engineers. What the images
06:08show is a control failure during testing, and there's an important detail in that. Even without any
06:14intention, a machine weighing dozens of kilos with powerful motors and joints capable of making fast
06:19movements can pose a real danger when something doesn't go as planned. Actually, this problem goes
06:24back even before the current generation of humanoid robots, you know? Back in 2023, we showed here on the
06:30channel an accident involving an industrial robot inside the Tesla Gigafactory in Texas. The human
06:36engineer was working near the equipment used in production when he ended up being crushed and
06:40injured by one of the robots. The case quickly made headlines claiming that a Tesla robot had attacked
06:46the engineer. But once again, let me make it clear, there is no evidence of any intention from these robots.
06:53It wasn't any rebellious artificial intelligence. It was an accident involving an industrial machine.
06:58And maybe it's exactly this distinction that is becoming more and more important.
07:03Because traditional industrial robots usually work in controlled areas, separated from people by safety
07:09protocols, you see? Meanwhile, humanoid robots are being developed precisely to do the opposite.
07:14They need to enter our environment, walk among us, work by our side, move around in stores, enter hospitals,
07:21operate in factories, and in the future come into our homes. At the same time, their physical abilities are
07:27advancing rapidly. Today, there are already humanoids capable of running, jumping, withstanding impacts,
07:33and performing martial arts movements. The Toyota G1 itself, the model on which the Russian robot was
07:38apparently based, has already been demonstrated performing punches, kicks, and complex kung fu sequences.
07:43And this creates a completely new situation. A software error on a computer usually causes a message on the screen.
07:50But a software error in a physical machine can cause some kind of movement. And the stronger,
07:55faster, and heavier this machine is, the greater the consequences of that error can be.
08:00There are also other cases we've already shown here, for example, like that little boy who was kicked
08:05during a street performance. But there's still another side to the problem. We also showed here on
08:11Realidad Impressionante's study revealing that a single malicious command could compromise humanoid robots,
08:16and, under certain experimental conditions, make the attack spread to other vulnerable machines that are
08:22nearby. And this adds a different layer to the problem. We don't need to imagine a robot developing
08:28anger or self-awareness for there to be a risk in all of this, you see. A machine can perform
08:34a dangerous
08:34action because it received the wrong command, or because a sensor misinterpreted the environment, or even
08:40because it lost its balance. Because there was a programming failure, because someone with very bad intentions
08:45took control of it remotely, or even because the system, yes, was compromised. And now we come back
08:50again to that store over there in Russia. This time we're not seeing a robot in an isolated industrial
08:55area, not suspended by a structure during a test, and not separated from the public by barricades during
09:01a tech demonstration. We're seeing a physically powerful machine inside a store surrounded by ordinary
09:08people. And maybe that's actually the most important part of this whole story. The incidents don't show
09:13that robots are developing a will to attack human beings, but they do show something much more
09:18concrete. The more these machines leave the labs and enter our world, the less room there is for
09:23mistakes, right? Because humanoid robots don't need to want to hurt someone. They just need to make the
09:29wrong move at the wrong time. And in the midst of all this, Hyundai claims that their future humanoid
09:36robots will not be used to replace human workers. The statement was made by the CEO of Hyundai Motor,
09:42Jose Munoz. This happened during the Investor Day 2026 held in Seoul Pot. However, the timing of the
09:48statement drew a lot of attention. Just days before, Hyundai and its South Korean Union had reached a
09:54preliminary agreement. After weeks of strikes and also negotiations that included concerns directly
10:01related to artificial intelligence and automation, humanoid robots coming in, and the job security of the
10:07guys. During his presentation to investors, Munoz stated that the robots will be serve people.
10:13According to the executive, they should mainly take on difficult and repetitive jobs, while new job
10:18categories may emerge in areas such as robot maintenance, for example. And Hyundai is preparing
10:23a significant expansion precisely in that area. The company plans to start producing humanoid robots in
10:29a new factory in the United States in 2028. The expected capacity is 30,000 units per year,
10:35and a large part of this initial production has already been allocated. According to Munoz,
10:40the first 25,000 commercial units are already committed, starting with Hyundai's own operations,
10:45and also including industrial and logistics partnerships. And one of the main robots in
10:49this strategy will be Atlas, developed by Boston Dynamics, which is also South Korean, you know?
10:54The first large-scale deployment is also scheduled to take place in 2028 at Hyundai Motor Group's
10:59Metaplant America, located in Georgia. Initially, the humanoid robots are expected to work on
11:04sequencing parts. The task involves organizing and preparing components in the necessary order for
11:09different stages of production. Starting in 2030, Hyundai plans to expand the use of Atlas for more
11:15complex activities, including even assembling components for their vehicles. But while Hyundai
11:21prepares thousands of humanoid robots for its factories, South Korean human workers are still
11:26discussing what the effects of this automation will be. Hyundai's union had already declared earlier this
11:31year that no robot based on new technologies should enter Korean factories without an agreement between
11:37workers and management. Employee representatives argued that the use of humanoid robots could allow
11:44the company to reduce hiring and affect job security. Labor negotiations dragged on for several months and
11:51included partial stoppages, and in August, Hyundai's first full strike in South Korea in about a decade took place.
11:58Therefore, for now there are two very clear positions. Hyundai does intend to manufacture
12:04tens of thousands of robots and gradually introduce them into its industrial operations.
12:10But it claims that its machines will be used to work alongside people, not to replace them.
12:17A Chinese company has developed a humanoid robot that can modify its own physical structure depending on the
12:23task it needs to perform. Called T-01, the robot was developed by Hanfeng Hanfontech, a company based
12:30in Beijing that presented this robot during the 2026 World Robot Conference. The machine is described
12:36by the manufacturer as an intrinsically transformable humanoid robot, and its concept is different from most
12:41current humanoid robots. Instead of trying to permanently replicate the human body's structure with two arms and
12:48two legs dedicated to the same functions, the T-01 uses limbs that can take on different roles. On regular
12:54surfaces, the robot can use wheels to move around. This configuration allows it to perform tasks such as
13:00moving and collecting packages in logistics environments and warehouses, where walking on legs
13:05is not always the most efficient way to get around. However, when it encounters stairs and uneven terrain,
13:11the configuration can change. T-01's arms are transformable and can be converted into legs. With this,
13:17the robot goes from a wheeled platform to a machine capable of using bipedal locomotion to overcome obstacles
13:23that would be difficult for a system that relies exclusively on wheels. After crossing the obstacles,
13:28it can simply return to the configuration that is more suitable for regular surfaces. T-01 can also transform
13:34into a kind of personal vehicle. In this configuration, a person can sit on the robot while the structures
13:41used as arms serve as the controls. According to the information released about the project, in this
13:46mode, the T-01 can reach a speed of up to 10 kilometers per hour. For now, the company has
13:52not yet
13:52published detailed information about the T-01's range, load capacity, full dimensions, or commercial price.
13:59The robot is already listed by specialized distributors, but its price is only provided upon request.
14:05Yes, the project is still in the early stage of rollout and commercialization. The main feature
14:10already demonstrated is its transformable architecture. Instead of building a robot that
14:15needs to walk like a person in every situation, the company created a machine that can use wheels
14:19when they are more efficient, legs when the terrain requires it, and even transform its own structure
14:24into a vehicle when the task is to transport a person. And this one is interesting and quite enigmatic,
14:32you know? Three of the main artificial intelligence services on our planet experienced problems almost at
14:38the same time, leaving users temporarily without access to those tools that have already become
14:43part of everyday life for thousands of people around the world. On the morning of September 3rd,
14:492026, ChatGPT, Claude and Grok experienced outages during overlapping periods. There were also reports of
14:56issues with Google's Gemini, although the company did not officially confirm an outage in its services.
15:01In the case of OpenAI, the problems began around 11 a.m. Eastern Time. The company's status page reported
15:09high error rates on ChatGPT and also on Codex. For many users, this simply meant not being able to chat
15:16with their chatbot. But the outage affected a much larger number of features, you know? Logins, file uploads,
15:23voice notes, internet searches, deep research and image generation also had issues. OpenAI implemented
15:30mitigation measures and later reported that its services had recovered. Almost simultaneously,
15:36Enthropic was also facing its own problems. Claude, Claude Code and the API used by developers
15:42were all affected. The company identified the cause as an infrastructure problem and implemented a fix.
15:48Around 12.15 p.m. Eastern Time, the main outage had already been resolved. And Jego AK, developed by Elon
15:55Musk's XAI, also experienced difficulties. The problem started even earlier, around 9 and 9.30.
16:01Users encountered messages stating that the model was overloaded and suggesting to try again or choose
16:07another model. According to the information, the failure affected the GRTA versions for Android,
16:13iOS and Web. Later, XAI linked the problem to an outage at its data center in Memphis, in the United
16:19States.
16:20And the coincidence immediately raised a very disturbing question. How did three different
16:24companies end up facing problems at almost the same time? Similar problems? So far, at the time
16:30I'm wrapping up the script for this episode, there isn't a definitive answer. The companies haven't
16:35announced a common cause for these outages. If there's any new information, leave it in the comments here.
16:42And Tesla has finally started putting its cyber cab on the streets of Austin, Texas, about two years after
16:48first presenting the vehicle back then, remember? The electric model was developed specifically to
16:54operate as a robotaxi and has a feature that sets it apart from the conventional cars the company has
16:59used until now. It doesn't have a steering wheel and it also doesn't have pedals to accelerate or stop the
17:06car.
17:06The cyber cab has two doors, space for only two passengers and was designed from the start to
17:12operate without anyone driving. And the first units are now being added to the robotaxi services that
17:17Tesla already operated in Austin. Until now, the company had mainly been using adapted versions of
17:22the Model Y for this purpose. Are we believe that at the time, these vehicles began operating in the city
17:28in June 2025 and since they were originally developed as conventional cars, they had equipment with steering
17:34wheels and pedals. The cyber cab represents a very different proposal, you see. Tesla wants to provide
17:39a vehicle dedicated exclusively to autonomous passenger transport, thus eliminating the components needed
17:44only for a human driver. And its arrival on the streets is happening in a fairly low-key way, you
17:50know?
17:50According to what New Atlas reported, Tesla held a private event with a small number of guests,
17:55without inviting most of the press and without publishing the kind of announcement that is usually
18:00associated with a launch of this scale. And this approach contrasts with the original presentation
18:04of the cyber cab back in 2024, when the company widely promoted its futuristic vision for robot taxis.
18:11And Tesla intends to follow its own strategy to expand its network. In addition to operating the
18:16cyber cabs directly, the company also plans to eventually sell the vehicles to regular consumers.
18:21The proposal mentions that a person could buy a cyber cab for around $30,000 and put it to work
18:26on the
18:26streets in Tesla's network, of course, transporting passengers and generating income for its owner.
18:31And the company has even made a form available for people interested in purchasing one of these
18:36vehicles. And the American company figure AI has found a new way to collect the necessary data
18:43to teach its humanoid robots to perform tasks in our real world and in our homes.
18:49Instead of relying solely on the robots themselves to generate this information,
18:52the company started paying humans to record videos performing common activities in their homes.
18:57The platform responsible for this is called INDEX.
19:01And it started being developed quietly within an initiative known as Project Go Big and has
19:06now been made publicly available for iOS and Android.
19:09The process is relatively simple. Users can choose certain tasks through the app,
19:15record videos performing these activities and send the material to the company in exchange for payments.
19:20The recordings include household tasks like folding clothes, making beds and cleaning rooms,
19:25but also activities performed in restaurants, store stock rooms and logistics facilities.
19:30Another possibility is to hire workers through the platform itself to perform tasks,
19:35while their actions are recorded from a first-person perspective.
19:38The scale reached by the project is already significant, you know.
19:41During approximately four months of operation before the public launch,
19:45figures says that the INDEX received 16 million videos produced in 108 countries.
19:50The app reached 264,000 downloads and more than 44,000 active users per week.
19:57According to the company, approximately 30 minutes of new recordings are added every second.
20:02For every thousand hours collected, the company claims to record an average of 373 different tasks,
20:08more than 1,100 objects handled and 116 distinct environments, the company has already distributed
20:1515 million dollars to participants and plans to invest more than 1 billion dollars in data collection,
20:21computational capacity and all of this during just the next 12 months.
20:25Therefore, the company's strategy connects two parts of the same problem.
20:28On one hand, the company is drastically increasing the quantity and diversity of data,
20:33data showing how people are performing their physical tasks.
20:36On the other hand, it uses its artificial intelligence architecture to turn this
20:40learning into movements performed by humanoid robots in our real world.
20:43And FIGURE has already manufactured more than 1,000 of these FIGURE-03 humanoid robots,
20:48this new version, and has dozens of them in commercial testing,
20:51already working in operations at the BMW factory in South Pambur, for example.
20:55If you enjoyed this episode, you know what to do, right?
20:58Leave a like, check your subscription, and on your screen, there are two new suggestions,
21:01one here and one here. Click on one of them, and I'll see you in the next one, okay?
21:05I'll be waiting for you there. Thanks a lot, and thank you. Bye-bye.

Recommended