Introduction
Welcome back to another episode of TJ's Technology Tuesday. What is this show all about? At its core, it's about assessing technology realistically — especially for executives and business leaders. And there is one topic that is regularly underestimated. What do I mean by that? I mean robotics. Let's take a look at how far robotics has already come today. How far along is it really? We'll explore this using a specific video as an example. Playing table tennis against a robot, for instance, is not a good idea — it will beat us convincingly. Spoiler: AI generates all those videos showing people playing badminton, tennis, or other fast-paced sports that require dynamic reactions. Robotics is not yet that far advanced. So what is the reality? Let me show you what the original video actually looked like. The original showed a real player in red, who was essentially replaced by a robot in the edited version. And then comes the impressive move. What do we notice? We notice that two things are interesting. First, we overestimate how far robotics has come — robots are not yet able to react that quickly and dynamically.
Robots at BMW in Action
What is also underestimated, on the other hand, is how good today's video technology already is at placing a robot over a real human. Let's also look at the current speed of actual robots. This is what BMW, for example, is deploying in Spartanburg. In their production line in Spartanburg, they are using these robots. And this is the speed these units operate at — but even when something is not positioned perfectly, they can correct themselves, jumping to the other side. What the robot does here — adjusting because a part is not sitting perfectly — is something AI systems could not do until recently. You may have also seen the Chinese New Year celebrations not long ago. What did that look like? There was a mix of real people and robots, and it looked impressive. You might say: surely they can improvise too? Yes and no. The speed and agility are great, but there are also some behind-the-scenes videos showing that these robots were trained intensively and precisely.
Why Robots Are Not Yet on Par with Humans
What becomes clear is that there is no unplanned interaction with humans whatsoever. That means if something unexpected happened — say someone suddenly fell over — the robots would have serious problems, because the entire routine was essentially pre-programmed. Just as the human dancers had learned the choreography by heart, the movements were 100% identical so that humans and robots could dance in sync. This was trained, with carefully maintained distances, and nothing that was not planned and programmed in advance. So the takeaway is: robotics has not come that far yet either. Let me turn to CES, which just took place in January 2026. This is the pace at which robots currently operate.
We see Atlas as the ultimate tool, we know that it's going to be going into environments and doing tasks alongside people for a long time. And so with Atlas we just see work changing and we've designed Atlas to work alongside people and.
And what these colleagues said — it works alongside humans. That is also why it looks like a human: it has the same constraints we do in most areas. It has a certain size, which means it can only reach a certain height, just like us humans. And it cannot get under things that are too low, because it is too tall. Of course, as we just saw, it can rotate 360 degrees, which we can only do to a limited extent — unless you're Meryl Streep in that wonderful film. What we see here is that we have moved significantly further forward than we were a year ago. We are not yet where some seem to think we are — as comedian Paul Panzer would say — but we are in a space where robotics is becoming much more capable. That said, if you look at the robots sorting laundry at trade shows — at IFA in Berlin last year, or at CES recently — I would always say: step aside. I can sort that five times faster than those machines.
So we are not yet where we want to be, and I say deliberately: not yet. Because the biggest mistake is to say, "robotics is still useless, I'm not doing anything with it." But equally mistaken is always trying to picture robots in the environments we are used to. Ford invented the assembly line at some point and defined what that looks like: a person standing there, turning screws. And so robots are being built to the same size — so that they fit into existing roads, existing tools, existing environments, including at home: dishwashers, kitchen cabinets — all designed around us as limited human beings. And so we are artificially limiting ourselves by building robots that mimic humans. Interestingly, even a company that builds humanoid robots does not use them in its own production. If I look at what Tesla has done, for example, Tesla said: wait a moment, I'm building something completely new here. I don't have existing situations where I need to replace humans in their environment. I can start from zero.
China's Lead in Robotics
And that is a good approach. Incidentally, KUKA — originally a German company, now Chinese — and Boston Dynamics, now also a Chinese-owned company. That means China is clearly leading here, because as a rule they think not just in electoral cycles but in generations. Their political decision-makers plan in timeframes where they know they will no longer be in office — or even alive — but they simply think longer-term. I may not always agree with what China does in terms of human rights and democracy, but it is a remarkably efficient system — and one that is increasingly powered by renewable energy. The era of the big diesel-belching machine is over; China is ahead there too. Another area where we consistently misjudge things is self-driving cars.
Waymo Outperforms Humans on Safety
Waymo, a subsidiary of Google, has been driving around for years already. They have roughly 41 million kilometres on the clock, and after those 41 million kilometres they analysed how many property and personal injury incidents occurred during that time. With AI — and I'll cover the label here with "AI generated" — there were nine property incidents and two personal injury incidents.
They compared this with how many property and personal injury incidents real human drivers caused over a comparable period. That was 78 property incidents and 26 personal injury incidents — significantly more. Of course, a human driver would never make the mistake of the white Tesla that drove into a white lorry. On the other hand, an AI would never get the idea to rear-end the car in front because it was quickly checking WhatsApp messages on its phone. And by the way, most accidents involving Waymo vehicles are caused by a human driver crashing into the back of the Waymo car. There are five levels of autonomous driving, one through five. Level one means essentially driving yourself; level five means fully autonomous operation. Waymo drives at level four, not five — this is not widely communicated. They also still operate with a human in the loop. What does that mean? If a vehicle does not know how to proceed, it stops. That is the moment when someone occasionally rear-ends it. Then two things can happen: either the car manages in a very short time to decide that the obstacle is gone and drive on.
Or it signals that a human needs to step in. And then there is the person with the job I would have loved to have as a teenager: they remotely drive the vehicle through San Francisco. So this is a technology where a human is still in the loop. And what is important here — as the colleagues at Gartner have always highlighted so nicely — they look at how expectations around a technology evolve over time. When you look at those two axes, there is always the same pattern. I'll illustrate it with the Metaverse: the Metaverse arrived, and everyone said that everything we now do on LinkedIn live, via video, or online — and everything we do in person — will be replaced. We'll all just be wearing these glasses — I don't have mine to hand right now, they're charging in the back. Yes, those VR headsets. The Apple Vision Pro came along, and Meta Quest 2, and Meta Quest Pro. Then people said it was a flop, that nobody was buying the Apple Vision Pro, that AI is far more exciting.
VR Headsets Revolutionising Sales and Training
And while the tabloid press declared the Apple Vision Pro a flop, Lufthansa was quietly introducing it for in-flight training. Or the next time you go to buy a Porsche, there is a good chance that when you say "do you have this one — the one standing there — not just in black, but also in yellow?" the salesperson will put on the headset, switch, and — there it is, the car in a different colour. You sit inside and say: "these are black leather seats — do you have them in light beige buff?" Suddenly you're sitting in a vehicle with light beige leather upholstery. Then you say: "actually, maybe it is a bit tight — perhaps I should buy a Cayman instead." As you can see, I don't drive a Porsche, so I'll say I want an SUV. What do I do? The Macan, for example — as you can see, yes, an SUV. Then I can switch to full virtual reality, and suddenly I'm sitting on a seat but virtually inside an SUV. And then I say: "let me look at another model," and I can switch to it without physically getting into another car.
Or Lufthansa, which says: I can put a pleasant passenger next to me, or a difficult one, and train for it. The idea here is very clear: with AI we always expect something like the Terminator — though we are also not always sure whether it will be benevolent or malevolent. I'll cover that topic in more depth in one of the upcoming episodes, because it is very important to assess, of course, that there are risks and dangers with AI as well. And when I look at what we experience when we ask Siri things — cynics say Alexa is like Siri, just with a high school diploma — and even Alexa sometimes leaves me despairing at home. But here is another nice example: an image I generated about a year and a half ago. At the time everyone said, "that's clearly AI generated." But that was two or three months ago. I simply said: create a photo with Steve Jobs and it should say "if you want to make people happy, sell ice cream." What came out? The point is: do not write off AI by saying it cannot do things. It can do something — but AI also increasingly means it can do more and more. And that is why one of the biggest mistakes you should not make is putting unrealistic expectations into AI. Underestimating technology is a bad idea. Overestimating technology is equally a bad idea. Two weeks ago I did an episode on this exact topic — Cloudboard or Moldboard, for instance — where most people said "brilliant, superb, saves the world." But now it is becoming clear that this may also carry some security risks, which I also pointed out at the time. In the meantime the whole thing has been acquired by OpenAI, so it is heading back to the US. You sometimes have to be careful about always being the very first pioneer — sometimes being an early follower, operating on the basis of truly realistic expectations, lets you unlock real productivity gains. That is the idea behind the realistic use of artificial intelligence — always combined with common sense.
Conclusion
And when does AI make sense? It makes sense when it increases productivity, improves the customer experience, or enables new business models. If you would like a sparring partner on this topic as a Personal IT Coach for executives, feel free to reach out — your Thorsten Jekel.
Key Takeaways
- Robotics is regularly both overestimated and underestimated: fast, dynamic movements are not yet second nature for robots.
- BMW is already deploying humanoid robots in its production line in Spartanburg, and these robots can self-correct as well.
- At the Chinese New Year celebrations, robots and humans danced together — but only thanks to an exactly pre-programmed choreography with no unplanned interaction.
- Waymo (a Google subsidiary) caused significantly fewer property and personal injury incidents after 41 million kilometres than human drivers — 9 property and 2 personal injury incidents compared to 78 and 26 respectively for humans.
- Waymo operates at level 4 of 5 of autonomous driving and stops when uncertain, until a human takes over remotely.
- VR headsets like the Apple Vision Pro are dismissed as a flop by the tabloid press, but are already being used by Lufthansa for cabin crew training and by Porsche dealerships for virtual vehicle configuration.
- KUKA and Boston Dynamics are now owned by Chinese companies; China leads in robotics because it thinks in generations rather than electoral cycles.
- Humanoid robots are built to human size so that they fit into existing human-scale environments — tools, roads, household appliances.
- AI makes sense when it increases productivity, improves the customer experience, or enables new business models.
- The biggest mistake is having unrealistic expectations about AI and robotics — neither overestimate nor underestimate, but assess realistically.
Frequently Asked Questions
How far has robotics really come today?
Robotics is more advanced than many people think, but not as far along as viral videos sometimes suggest. Fast, dynamic reactions — such as in sports — are not yet possible for today's robots; many of those videos are AI-generated or show precisely pre-programmed movement sequences.
Where are robots already being used in businesses today?
BMW is already deploying humanoid robots in its production line in Spartanburg. These robots can self-correct when a part is not perfectly positioned — a capability that AI systems have only recently developed.
Why can't robots yet interact spontaneously with humans?
Current robots are designed to execute exactly pre-trained movements. Unplanned interactions with humans — such as when someone suddenly falls — still pose major problems, because they cannot flexibly respond to unforeseen situations.
How safe are self-driving cars compared to human drivers?
Waymo, Google's self-driving car unit, caused only 9 property incidents and 2 personal injury incidents after 41 million kilometres. Human drivers caused 78 property incidents and 26 personal injury incidents over a comparable period — significantly more.
What does level 4 of autonomous driving mean?
Autonomous driving has five levels: level 1 means the human drives themselves; level 5 means fully autonomous driving with no human input. Waymo operates at level 4, stops when uncertain, and is then remotely driven by a human if needed.
Why are humanoid robots built to human size?
Humanoid robots are constructed to human size so that they fit into existing environments built for people — including roads, tools, dishwashers and kitchen cabinets. This adaptation to human infrastructure simultaneously places artificial constraints on the robots.
Why is China so far ahead in robotics?
China leads in robotics partly because political decision-makers think long-term in generational timescales rather than short electoral cycles. Even formerly German companies such as KUKA and Boston Dynamics are now under Chinese ownership.
Are VR headsets like the Apple Vision Pro really being used, or are they a flop?
Despite negative coverage in the tabloid press, VR headsets like the Apple Vision Pro are already in practical use: Lufthansa uses them for cabin crew training, and Porsche dealerships use them to show customers vehicles in different colours and configurations without having to change the physical car.
When does using AI in businesses actually make sense?
AI makes sense in a business context when it increases productivity, improves the customer experience, or enables new business models. The key is to have realistic expectations and to neither overestimate nor underestimate AI.
What is the most common mistake when dealing with new technologies like robotics and AI?
The biggest mistake is holding unrealistic expectations: either writing off a technology entirely because it is not yet perfect, or massively overestimating it because individual demos look impressive. Realistic assessment — combined with common sense — leads to genuine productivity gains.
Tools & Resources Mentioned
- Artificial Intelligence (AI) — overview of AI topics on digital4productivity.de
- BMW Spartanburg — deployment of humanoid robots in automotive production
- Waymo (Google subsidiary) — self-driving car at level 4, already 41 million kilometres driven
- Boston Dynamics (Atlas) — humanoid robot, now under Chinese ownership
- KUKA — originally a German robotics company, today Chinese-owned
- Apple Vision Pro — VR headset, used among others by Lufthansa for in-flight training
- Meta Quest — VR headset mentioned in the context of Metaverse expectations
- Tesla — builds humanoid robots but in its own production uses robot-friendly environments built from scratch
