Latest / Automated with Brian Heater / Andie Zhang on Why Industrial Robots Still Beat Humanoids

Transcript

Andie Zhang on Why Industrial Robots Still Beat Humanoids · Automated with Brian Heater

• 0:00
  SPEAKER_00: Humanoid robots are designed to do many things.
• 0:04
  SPEAKER_00: They're supposed to be able to do A, B, C, and D.
• 0:07
  SPEAKER_00: The problem is that they don't do any of these things very well.
• 0:11
  SPEAKER_00: My career has been driven by doing something new.
• 0:14
  SPEAKER_00: I was from technology, I wanted to try business, and then collaborative robots was appealing because it was new.
• 0:19
  SPEAKER_00: I was the first person responsible for cobots at ABB.
• 0:22
  SPEAKER_02: The majority of roles that I see humanoids fulfilling requires people to be around, and safety is another huge mountain that needs to be climbed.
• 0:32
  SPEAKER_00: Absolutely.
• 0:32
  SPEAKER_00: Just because it looks like a human doesn't mean that it has this type of safety that is required for, say, human robot collaboration.
• 0:39
  SPEAKER_00: Collab robots that I have been working on is intended to be put into exactly these places where humans are working.
• 0:48
  SPEAKER_02: Do you think that like at any point in the foreseeable future that the big heavy industrial arm is going away?
• 0:55
  SPEAKER_00: I think for industrial manufacturing, cost, ROI, efficiency, reliability is always going to the main KPIs.
• 1:02
  SPEAKER_00: And there, a human-wide robot is not going to be competitive.
• 1:06
  SPEAKER_02: Uptime is a really big issue with robots.
• 1:08
  SPEAKER_02: You do need people there to interface with the robots.
• 1:11
  SPEAKER_02: And is there a role that companies can play to better position the people in those factories to interact with the robots and to be on the other side of automation?
• 1:42
  SPEAKER_02: Coming up, we have the second of two episodes recorded on-site at the inaugural Davos Tech Summit in Switzerland over the summer.
• 1:49
  SPEAKER_02: Andy Chang is the global portfolio and technology manager for collaborative robots at ABB.
• 1:56
  SPEAKER_02: She's got a lot of words in her title, which means she's very important, which is headquartered one extremely beautiful train ride away in Zurich.
• 2:05
  SPEAKER_02: Thanks to her, thanks to ABB, thanks to the event organizers for helping set up that interview and getting some room.
• 2:13
  SPEAKER_02: Thanks to you as always for tuning in.
• 2:16
  SPEAKER_02: If you're enjoying the show, you can help us out by liking and subscribing and checking out our newsletter over at automated.fm.
• 2:24
  SPEAKER_02: In the meantime, please enjoy this conversation with Annie Jang of ABB.
• 2:30
  SPEAKER_02: You studied engineering.
• 2:32
  SPEAKER_00: Yes.
• 2:32
  SPEAKER_02: Originally.
• 2:33
  SPEAKER_00: I did.
• 2:34
  SPEAKER_02: And then you went back years, like a couple decades later, it seems like, and uh for management, is that right?
• 2:42
  SPEAKER_00: Yes, you're right.
• 2:43
  SPEAKER_00: Actually, I was born in China.
• 2:45
  SPEAKER_00: And when I was 10 years old, we moved to Sweden because my father was doing his PhD in engineering at uh KTH, the Royal Institute of Technology in Stockholm.
• 2:54
  SPEAKER_00: And because both my parents are engineers, so I grew up thinking, okay, that's the only way to go.
• 2:58
  SPEAKER_00: So I also graduated was uh from the KTH.
• 3:02
  SPEAKER_02: I was curious where you you you grew up.
• 3:04
  SPEAKER_02: You grew up in the hometown of ABB.
• 3:07
  SPEAKER_02: And anytime there you look at one of these kind of like big multinational corporations that is in like everything, they always you go to their like Wikipedia page and it's always like in 1880s, you know, it started as like these comments, like so so long history, but usually, you know, somebody like you know, it'll be like Nintendo, you know, they started as like a playing card company, or I think like Nokia started as like a the shoe rubber company.
• 3:37
  SPEAKER_02: But ABB was pretty much like electronics.
• 3:41
  SPEAKER_00: Well, actually, am I wrong?
• 3:43
  SPEAKER_00: From the beginning, okay, the robotics part of ABB was ASEA uh in Sweden.
• 3:48
  SPEAKER_00: Yeah.
• 3:48
  SPEAKER_00: So that's where it started.
• 3:49
  SPEAKER_00: Actually, it started even further back with uh building uh power in Sweden.
• 3:54
  SPEAKER_00: There's a hydro power plant, still there in Vestuals.
• 3:58
  SPEAKER_00: But then the BB part actually comes from Brown Bovary, which is a Swiss company in 1988.
• 4:03
  SPEAKER_00: The ASEA, the A and the BB merged into ABB.
• 4:06
  SPEAKER_00: So that's where we get ABB today.
• 4:08
  SPEAKER_02: Yeah, I I'm part of the reason why I ask is the A and the BB both have this like, you know, more than a hundred-year-long history.
• 4:15
  SPEAKER_02: And uh I'm and I'm curious like how big of a presence it was in the town that you grew up in.
• 4:21
  SPEAKER_00: Oh wow, Investor was ABB was everywhere.
• 4:24
  SPEAKER_00: You know, ABB also went through phases of expansion where we acquired a lot of companies from nuclear power to building trains, etc.
• 4:31
  SPEAKER_00: And then also divesting some of those businesses and focusing on the core.
• 4:35
  SPEAKER_02: So in terms of engineering, like specifically, what were you studying?
• 4:40
  SPEAKER_02: What was your focus?
• 4:41
  SPEAKER_00: Yes, so actually uh this was so long time ago that it was for sure before the iPhone, even before color phones.
• 4:47
  SPEAKER_00: So it was uh black and white not being phones.
• 4:49
  SPEAKER_02: We had color TVs at the time, though, right?
• 4:50
  SPEAKER_00: There were color TV.
• 4:52
  SPEAKER_00: But um, so um I studied something that was quite new.
• 4:55
  SPEAKER_00: It was called multimedia technology at the time, and it was really looking at you know how do new types of media form, you know, that was before also uh streaming video.
• 5:04
  SPEAKER_00: So, you know, all these new things were coming on.
• 5:07
  SPEAKER_02: Yeah, we were sending, we were actually like sending Netflix in the mail at that time.
• 5:11
  SPEAKER_00: Exactly, exactly.
• 5:12
  SPEAKER_00: So also before streaming audio, streaming video.
• 5:14
  SPEAKER_00: So this was early days.
• 5:15
  SPEAKER_00: So I was really interested in how um human and computers interact.
• 5:20
  SPEAKER_00: You know, I got my first computer when I was a teenager, and then the internet came and I was like teaching myself to program web pages, and I was very fascinated by yeah, personal computer, internet, and the possibilities it's brought.
• 5:34
  SPEAKER_00: So I was always interested in the technology part, but also in the human part.
• 5:38
  SPEAKER_00: So my major was in human computer interaction then, uh, which now fits nicely with the human robot interaction piece that I'm also in.
• 5:45
  SPEAKER_02: Yeah, I was gonna ask how how how natural of the transition it is, because um, you know, looking down your CV, you you interned at ABB and then you transitioned to TetraPack, which is an automation company in packaging.
• 6:00
  SPEAKER_02: So it was a pretty natural transition from the this you know computer into robotics, or was that a bit of a heavy lift on your end?
• 6:09
  SPEAKER_00: Not at all.
• 6:10
  SPEAKER_00: Actually, um I was just very curious.
• 6:13
  SPEAKER_00: So, first of all, I was curious about the business part.
• 6:16
  SPEAKER_00: You know, technology is one thing, but customers, consumers, how do they use the technology?
• 6:21
  SPEAKER_00: How does uh technology add value?
• 6:23
  SPEAKER_00: But also more parts of the value chain.
• 6:25
  SPEAKER_00: You know, technology, you think a lot about RD.
• 6:28
  SPEAKER_00: But what about um building, like producing the machines?
• 6:32
  SPEAKER_00: What about the suppliers, managing the supply chain, signing contracts with the suppliers?
• 6:36
  SPEAKER_00: What is that like?
• 6:36
  SPEAKER_00: And then of course, it was really fascinating because TetraPack, we work with packaging machinery, but the packaging is ultimately used for packaging like uh beverages, you know, from milk, dairy drinks to juice and other types of drinks, which face consumers.
• 6:51
  SPEAKER_00: And TetraPack worked very closely with the companies that were producing the consumer products.
• 6:56
  SPEAKER_00: So I also remember doing consumer studies, following into the homes of regular consumers, open their fridge to see what's inside, follow them down the supermarket aisle to see how they pick products from the shelf.
• 7:08
  SPEAKER_00: And all of these pieces I think contribute to my work later with robotics, managing the product portfolio, because you have to focus on your customer.
• 7:17
  SPEAKER_00: Now, my customer is not consumers, but still, I need to understand the psyche, the behavior of my customers who are using the robot.
• 7:25
  SPEAKER_00: And what do they need from the robot?
• 7:26
  SPEAKER_00: How do they want to use the robot?
• 7:28
  SPEAKER_00: And how does this fit into the bigger picture?
• 7:30
  SPEAKER_02: So so to a certain extent, it it sounds like when we're talking about human computer interaction, um, you know, there's that sociological element, like that's the um the consumer-facing side, but but also like I mean, you're you you're an engineer, you studied engineering, so you're also bringing like a technical background to the table as well.
• 7:51
  SPEAKER_00: Yes, absolutely.
• 7:52
  SPEAKER_00: And that's I think really bridging, I think that's really my role right now is managing the portfolio of collaborative robots, is the bridge between my customers, my end users, the market, the businesses who have you know business needs to fill, and our deep tech RD research in robotics, in the technologies, right?
• 8:13
  SPEAKER_00: So there's a lot of cool things you can do with a robot, but you have to have customer need, a problem to solve, and then the product features, the technology, they have to match.
• 8:22
  SPEAKER_00: You have to solve the actual customer problem.
• 8:25
  SPEAKER_00: Otherwise, you just have cool technology, which is yeah, cool, but doesn't serve an economical value for the customer.
• 8:31
  SPEAKER_00: So really matching these two worlds, and that is what I'm most interested in.
• 8:35
  SPEAKER_02: What what what sorts of problems did a Tetra pack face, like from an automation standpoint?
• 8:40
  SPEAKER_00: So there's flexibility versus a productivity.
• 8:44
  SPEAKER_00: So the fastest filling machine speed at that time, that was 10 years ago, could do 11 packs per second.
• 8:52
  SPEAKER_00: So it's just so fast the speed of this machine.
• 8:55
  SPEAKER_00: But then uh you suffer the flexibility.
• 8:57
  SPEAKER_00: So you there are some high runners, just like in any industry, you have the high volume products, and there it justifies this type of automation that's super fast and uh high quality.
• 9:07
  SPEAKER_00: But then more and more the consumers are changing their taste, and that impacts the companies that know Nestle, you know, who are producing these products.
• 9:15
  SPEAKER_00: So the consumers are not, you know, Coca-Cola, now the packaging has like people's names on it, for example.
• 9:21
  SPEAKER_00: So that's challenges for the manufacturer, right?
• 9:24
  SPEAKER_00: So you can't keep having this super high volume fixed automation.
• 9:28
  SPEAKER_00: You have to add flexibility to it.
• 9:30
  SPEAKER_00: And that is the consumer demand, which feeds back to the equipment manufacturer, technology manufacturer.
• 9:36
  SPEAKER_00: How do we accommodate this type of new consumer challenges, which maybe change all the time?
• 9:41
  SPEAKER_00: So the manufacturer has to be able to pack, you know, have different packaging material, different labels, et cetera.
• 9:47
  SPEAKER_00: Another main challenge is the environmental aspect.
• 9:50
  SPEAKER_00: You know, plastic bottles, we don't make plastic bottles, but there is plastic inside the cut the paper packaging as well.
• 9:57
  SPEAKER_00: So how to make the packaging as environmentally friendly as possible.
• 10:01
  SPEAKER_00: So for example, caps.
• 10:03
  SPEAKER_00: Uh Tetropack has successfully replaced all the caps in the straws with uh renewable material, like from sugar cane, from paper, etc.
• 10:11
  SPEAKER_00: So this was also a big challenge, you know, how to uh bring environmentally sustainable products to the market.
• 10:18
  SPEAKER_02: Yeah, I want to talk a little bit about uh the, you know, as you were saying, the changes in taste and and and the turnover, like adjustments of the workflow of the machines.
• 10:28
  SPEAKER_02: This is something that's become really interesting to me as I've been talking to a lot more automation companies.
• 10:34
  SPEAKER_02: Like their seasonality, that's a big part of it.
• 10:36
  SPEAKER_02: You know, we had uh Rick Falk from Locust on the show uh not long ago, and they have mobile systems, right?
• 10:43
  SPEAKER_02: So they people will they'll do a RAS model and then people will like send the robot back to them, they'll refurbish it, things like that.
• 10:50
  SPEAKER_02: But when if we're talking to about basically like an assembly line, in order to replace or reconfigure the machine, you you kind of have to stop the assembly line to a certain extent, right?
• 11:05
  SPEAKER_02: And that's the last thing that a manufacturer wants to do.
• 11:08
  SPEAKER_00: Yeah.
• 11:09
  SPEAKER_00: So the for when it comes to assembly, which typically is in, let's say, consumer electronics, labor shortage is another big problem.
• 11:15
  SPEAKER_00: So shorter life product life cycles from I don't know, five, ten years to now two, three years.
• 11:20
  SPEAKER_00: So yes, that means reconfiguration of the line, but also making it easier for the workers because there might be a high turnover rate of the workers if you're producing in China or one of these countries where labor shortage is a big problem.
• 11:33
  SPEAKER_00: So the producer, the factory, they break down the work tasks into smaller and smaller pieces.
• 11:40
  SPEAKER_00: And absolutely, we need to do offline programming, offline commissioning of the product production line.
• 11:45
  SPEAKER_00: And this means doing more in simulation using a virtual twin.
• 11:49
  SPEAKER_00: So at ABB, we have actually a world-leading simulation software, robot simulation software called Robot Studio, where it's already possible to build a full digital twin of not only the robot, but the full line, all the physical components of the line, so that you can program the robot, you can program the robot motion, you can see how it tests the cycle time to optimize the cycle time, for example, and do as much offline as possible so that you can reduce the downtime online.
• 12:16
  SPEAKER_00: Another new feature that we're adding now as we're embracing physical AI and adding more AI capabilities to our robot systems is uh together with collaboration with NVIDIA, we're leveraging now the NVIDIA Omniverse to add synthetic images.
• 12:33
  SPEAKER_00: So traditionally, we can simulate all the hardware parts.
• 12:37
  SPEAKER_00: What was not possible to simulate is the vision input or sensor input of any kind, actually.
• 12:43
  SPEAKER_00: So imagine a camera, what the camera is seeing.
• 12:45
  SPEAKER_00: And if your application now assembly more and more is based on vision base.
• 12:49
  SPEAKER_00: So you use a vision system, you take a picture, and based on that you decide where to pick, how to pick this part, or you can even do a quality inspection.
• 12:57
  SPEAKER_00: Now, traditionally you install and you have to take lots, hundreds of pictures of the real objects.
• 13:02
  SPEAKER_00: Now we can use synthetic image to pre-train this system offline, which reduces then the commissioning time on site.
• 13:11
  SPEAKER_00: And we have already tested this with Foxconn, for example, who has uh reduced their commissioning time by up to 40%.
• 13:18
  SPEAKER_02: So are you using uh the Nvidia Isaac Sim at all?
• 13:22
  SPEAKER_02: Or you I mean, we I as you said you built your own simulation, so I assume that primarily you would want to stick to that.
• 13:29
  SPEAKER_00: Yes, yes, exactly.
• 13:30
  SPEAKER_00: So we do have um, let's say, our own virtual controller inside our robot studio system because this behaves identical to uh our real robot.
• 13:41
  SPEAKER_00: So it is optimized for the motion control of ABB's robots, right?
• 13:45
  SPEAKER_00: Now, what we leverage from NVIDIA at this moment is um the synthetic image generation, and that's how we're using, we're leveraging the NVIDIA ecosystem right now.
• 13:55
  SPEAKER_00: Now, of course, we are looking at how we can add more capabilities to our robots to make them more physical AI capable, adding more autonomous, more versatile.
• 14:05
  SPEAKER_02: Yeah, because like the you know, the the the big or one of the big problems with physical AI is is the physics part of it, you know, and it and it sounds like the thing that your simulation does particularly well is it knows your machines, um, but it's really useful to know the physics of the objects that you're picking up in the physics of of the world uh around you.
• 14:28
  SPEAKER_02: I mean, how much of that is actually baked into your system?
• 14:31
  SPEAKER_00: So that is already baked into robot studio, actually.
• 14:34
  SPEAKER_00: We already have this physics simulator in there.
• 14:36
  SPEAKER_00: So even we can simulate gravity, we can simulate how the cable harness of the robot moves, etc.
• 14:42
  SPEAKER_00: So this is already well incorporated into our system.
• 14:46
  SPEAKER_00: So what is new is really the sensing, the sensing part.
• 14:49
  SPEAKER_00: Yeah.
• 14:49
  SPEAKER_02: A lot of the physical AI startups that we're seeing right now are like essentially building AI for startups, right?
• 14:57
  SPEAKER_02: You know, they're they're they're they're building for those those embodiments.
• 15:01
  SPEAKER_02: Um you have an existing embodiment, you know, you have these arms that you've been building for a really long time.
• 15:07
  SPEAKER_02: So you have that base, you know, you have a huge deployment, like a huge customer base already out there that you need to serve.
• 15:15
  SPEAKER_02: But essentially, as the systems themselves become more capable and the hardware changes, um, then it like the the simulation also needs to follow and and and mature with the hardware.
• 15:30
  SPEAKER_00: Simulation is one aspect.
• 15:32
  SPEAKER_00: I mean, when it comes to physical AI, so we are definitely going in this direction.
• 15:35
  SPEAKER_00: That is clearly the strategy.
• 15:37
  SPEAKER_00: And uh what we're doing is making our industrial robot arms more versatile and more autonomous, and that is adding more capabilities to it.
• 15:47
  SPEAKER_00: So traditionally, it's fixed programming, it's hard programming.
• 15:50
  SPEAKER_00: Yeah.
• 15:50
  SPEAKER_00: And there is still a need for that today.
• 15:52
  SPEAKER_00: So the fact that there, you know, there are new capabilities brought by AI doesn't replace the current automation.
• 15:57
  SPEAKER_02: Well, let's let's talk about that real quick because um the well, one, when when when you talk about making them more capable, are you talking about systems that are actually deployed?
• 16:06
  SPEAKER_02: Are you talking about new systems that are coming to the market?
• 16:09
  SPEAKER_02: New systems.
• 16:10
  SPEAKER_00: So the systems that are deployed are usually installed in a already commission installed in a very good way in that they're generating value for the customer, they're solving the problem in a very reliable, high accuracy way.
• 16:21
  SPEAKER_00: Otherwise, they wouldn't be deployed, right?
• 16:23
  SPEAKER_00: That's the difference, I think, between us and startups.
• 16:25
  SPEAKER_00: So we're not doing POCs.
• 16:27
  SPEAKER_00: We're doing real deployment in customer size, fulfilling very strict productivity accuracy requirements.
• 16:34
  SPEAKER_02: Yeah, because the you know, the thing that that robots uh that like especially big, heavy, I mean, obviously, you know, ABB makes cobots as well, but uh these big, large uh robot arms do really well is is repetitive tasks.
• 16:49
  SPEAKER_02: Um so you know, so so they're already deployed doing that, but now there's an expectation that robots are gonna be able to do even more things.
• 16:58
  SPEAKER_00: Exactly.
• 16:59
  SPEAKER_02: Yeah.
• 16:59
  SPEAKER_00: So the customers now, it's often the same customer.
• 17:02
  SPEAKER_00: They already have a lot of industrial robots deployed.
• 17:04
  SPEAKER_00: If you look at a car OEM, body shop, the painting line, highly automated already.
• 17:09
  SPEAKER_00: But if you look at final trim and assembly and final testing, that's a that's where the human labor is.
• 17:14
  SPEAKER_00: And all these companies are looking at how to automate those tasks.
• 17:17
  SPEAKER_00: But if you you know ever see how a car is produced, it's you know on a moving line, and you have just lots of people going in, you know, you have to remove the doors and you have to do stubble assembly on the doors, and everything is moving constantly at this pace, right?
• 17:30
  SPEAKER_00: You put in the dashboard, you have so much human doing all these things, and this has traditionally still is today, very hard for robots to solve.
• 17:38
  SPEAKER_00: And in order for robots to solve them, they need to be able to perceive their environment, uh, which in this case is more unstructured.
• 17:47
  SPEAKER_00: So they need to be able to perceive, they need to be able to understand, and then they need to be able to act accordingly in an autonomous way.
• 17:54
  SPEAKER_00: So, okay, based on the vision sensor, this part is moving, it's here.
• 17:59
  SPEAKER_00: Okay, I'm going to, you know, mount this part here and while it is moving, for example, and apply it a force.
• 18:06
  SPEAKER_00: That's also difficult.
• 18:07
  SPEAKER_00: You know, humans can feel with their hands.
• 18:09
  SPEAKER_00: Robots, grippers, uh mostly actually don't have a tactile sensing.
• 18:14
  SPEAKER_00: So how much how hard do I press until it clicks into place?
• 18:18
  SPEAKER_00: So these are the new skills that robots need to be equipped with in order to automate more tasks in the field.
• 18:24
  SPEAKER_02: Is that uh is something like that?
• 18:26
  SPEAKER_02: I mean, without without sensing, without feedback like that, is is that a matter of of hard coding?
• 18:32
  SPEAKER_02: Like how do you actually do that?
• 18:34
  SPEAKER_00: That's the thing.
• 18:35
  SPEAKER_00: With hard coding, typically it's position-based.
• 18:37
  SPEAKER_00: So then it has to be the part is in one place, I just move into position, and that's it.
• 18:42
  SPEAKER_00: So you can't.
• 18:42
  SPEAKER_02: So how do you how do you do that without like the right sensors?
• 18:45
  SPEAKER_00: Without you can you program the position that it goes in, or we also have force control, but it's quite complicated.
• 18:51
  SPEAKER_00: And this is why these tasks today are not automated, because they're so difficult to automate.
• 18:56
  SPEAKER_00: And that is what actually physical AI unlocks are these tasks which today cannot economically be uh automated or technically feasibly be automated with current uh traditional automation.
• 19:11
  SPEAKER_00: And there are also other tasks such as um, you know, picking, like in logistics centers.
• 19:16
  SPEAKER_00: Now you have, for example, fast fashion, you know, people are shopping more online.
• 19:20
  SPEAKER_00: And imagine all these clothes are produced typically in South in Asia, China, other Asian countries.
• 19:26
  SPEAKER_00: They're putting these poly bags, you know, flexible plastic bags, uh, and there are so many different kinds, variants of these clothes.
• 19:33
  SPEAKER_00: You're putting to cardboard boxes, shipped to distribution centers.
• 19:36
  SPEAKER_02: Like a shine.
• 19:37
  SPEAKER_00: For example, but I mean, all the major fast-fashioned companies, right, have uh e-commerce, also small ones.
• 19:43
  SPEAKER_00: They go into the logistics center and somebody has to open the boxes, somebody has to pick them out and singulation.
• 19:50
  SPEAKER_00: I mean, you have to separate every single one of them.
• 19:52
  SPEAKER_00: And this is currently done by hundreds of thousands of human workers because every one of them is different.
• 19:57
  SPEAKER_00: You cannot program, pre-program how they look like because they're flexible, they're soft, and nobody knows how they look like.
• 20:03
  SPEAKER_00: And now, for example, we have developed a system using AI vision.
• 20:07
  SPEAKER_00: So the AI algorithm is pre-trained on this similar type of objects, but then even when it sees a new type of, let's say, polybag filled with clothes, it will automatically recognize what it is, where the boundaries are.
• 20:19
  SPEAKER_00: It will decide by itself autonomously how to pick this part and then place it on the conveyor belt.
• 20:24
  SPEAKER_00: And it will also recognize if there's, you know, the flap from the cardboard, interlace layer, etc.
• 20:29
  SPEAKER_00: So this system that we call the fashion inductor, this is really AI-based automation, which solves problems that could not be solved before without AI.
• 20:39
  SPEAKER_02: Yeah, I mean, the the the the heart the harsh reality of of manufacturing when it comes to the human labor is, you know, in a lot of cases, the reason why it's moved to where it is is because like that that's where the human labor is the the cheapest, right?
• 20:54
  SPEAKER_02: So there's the least overhead.
• 20:57
  SPEAKER_02: And for, you know, for like as long as I've been in this industry, people have been talking about decentralizing manufacturing to a certain extent.
• 21:04
  SPEAKER_02: There's been that conversation.
• 21:06
  SPEAKER_02: It hasn't really happened.
• 21:07
  SPEAKER_02: You know, we have seen some manufacturing come to the US, but um the and then the pandemic happens, and then the supply chain breaks, you know.
• 21:19
  SPEAKER_02: So all of a sudden, like there's even more demand for this idea.
• 21:23
  SPEAKER_02: And it seems to me that you know, if automation ever gets to a point where it is that precise and it can do those jobs, then at very least, like there's more of a possibility of having manufacturing closer to the customer.
• 21:40
  SPEAKER_00: Absolutely.
• 21:41
  SPEAKER_00: And actually, this is what many industries want uh shorter lead time to the end customer.
• 21:46
  SPEAKER_00: We have robot factories in Sweden, in China, in the US, in order to have shorter lead times to our customers in all of those regions.
• 21:53
  SPEAKER_00: So absolutely, and when it comes to, let's say, labor, I worked 10 years in China, and even if it's It's a low labor cost country.
• 22:01
  SPEAKER_00: Labor shortage is top of mind for every company I visit, every manufacturer.
• 22:06
  SPEAKER_00: Because the new generation, Gen Z, they don't want to work in factories, right?
• 22:10
  SPEAKER_00: I mean, factory uh managers I talked to in China, they say their biggest headache is after the Chinese New Year holiday, which is like Christmas, half the workers don't come back.
• 22:19
  SPEAKER_00: So they have to, you know, train new workers, on board, train new workers.
• 22:23
  SPEAKER_00: It's a huge headache for them.
• 22:24
  SPEAKER_00: That's why they want to deploy automation.
• 22:26
  SPEAKER_00: That is why China is deploying so much automation.
• 22:29
  SPEAKER_02: Yeah, there is there's something kind of kind of wonderful in you know the idea of like uh taking like taking a little vacation, just being like, yeah, all right, I'm done.
• 22:38
  SPEAKER_02: You know, I'm all right.
• 22:39
  SPEAKER_02: Yeah.
• 22:40
  SPEAKER_00: They want to become a gig worker.
• 22:41
  SPEAKER_00: Yeah.
• 22:43
  SPEAKER_00: Or you know, travel Uber.
• 22:44
  SPEAKER_00: There are many things that they can.
• 22:46
  SPEAKER_02: Yeah, yeah.
• 22:46
  SPEAKER_02: And I mean, you know, it this is interesting.
• 22:49
  SPEAKER_02: This is a this is a difficult conversation.
• 22:51
  SPEAKER_02: No, nobody, I don't know that anybody's giving certainly nobody's giving me a perfect answer, but um, you know, uh we talk a lot about upskilling, right?
• 23:01
  SPEAKER_02: So the idea is that like some jobs, in invariably some roles at least will be replaced by automation, right?
• 23:10
  SPEAKER_02: Like that's gonna happen.
• 23:11
  SPEAKER_02: At least for the foreseeable future, uh will humans be required for uh, you know, in order to interface directly with the robots themselves, and is there a way to upskill the people who are already in those uh those blue-collar jobs to do that work?
• 23:31
  SPEAKER_00: Yeah, I think definitely.
• 23:32
  SPEAKER_00: I mean, I have visited over a hundred customer plants in various industries from you know, industrial manufacturing, lab automation, food and beverage, automotive, electronics, and uh all the time.
• 23:44
  SPEAKER_00: The reason why the factory management wants to deploy robots or automation is because they're growing their business, they can't find enough qualified labor.
• 23:52
  SPEAKER_00: I can give welding as an example.
• 23:54
  SPEAKER_00: Um, you know, welding, when I started with COBOS 10 years ago, we didn't think that cobalt would be doing welding, but now it's one of the main applications.
• 24:02
  SPEAKER_00: And why?
• 24:03
  SPEAKER_00: Because there are not enough qualified welders.
• 24:06
  SPEAKER_00: I remember a statistic in 2025 there would be over 400,000 missing welders uh in the US alone, which is like crazy numbers.
• 24:15
  SPEAKER_00: So what happens in the factories is that uh there's a lot of small welding jobs, manual welding.
• 24:21
  SPEAKER_00: So now what you can do with, for example, cobalt is that the welder is using an easy-to-use interface.
• 24:27
  SPEAKER_00: It can be a visual or graphical user interface on a tablet or a device directly mounted on the cobalt flange with buttons, where they can just move the robot to position A, position B, they can say circular weld, straight weld, and they teach this weld.
• 24:43
  SPEAKER_00: And then they move on to teach another robot, and the robot is performing these welds.
• 24:47
  SPEAKER_00: So in this way, a company can double their productivity with welding without adding new welders.
• 24:54
  SPEAKER_00: And you have to think of, you know, there are just fewer and fewer people who are doing manual welding uh today and in the future.
• 25:00
  SPEAKER_00: So this is an example of a type of labor that needs to be automated.
• 25:05
  SPEAKER_00: And the new uh people that join the workforce, they might want to do other types of tasks, such as teaching the robots.
• 25:12
  SPEAKER_00: So the welder no longer has to perform the weld themselves, they are now upgraded to teaching the robot how to do the task.
• 25:18
  SPEAKER_00: Yeah, yeah.
• 25:18
  SPEAKER_02: And uh yeah, obviously, like every every every day there's there's new people entering the workforce, but you know, I'm I'm always thinking about the people who are like there who have, you know, who have been trained on on the warehouse floor themselves.
• 25:31
  SPEAKER_02: And that's what I'm curious about, you know, is when I talk about upskilling or reskilling, I guess um some people call it uh uptime is a really big issue with robots, right?
• 25:43
  SPEAKER_02: Um, you know, you do need people there to interface with the robots.
• 25:47
  SPEAKER_02: And and you know, I'm I'm this is this is this isn't really an A B B question, or maybe it is, but you know, is there a role that companies can play to better position the people in those factories to interact with the robots and to, you know, to be on sort of like the other side of automation?
• 26:10
  SPEAKER_00: Mm-hmm.
• 26:10
  SPEAKER_00: Yeah, absolutely.
• 26:11
  SPEAKER_00: I mean, uh not only does ABB support vocational schools and education as a general effort, because we want to train more, you know, robot programmers, robot operators for the future to be prepared for more robots entering the workforce.
• 26:24
  SPEAKER_00: When you say supported, what do you mean by that?
• 26:26
  SPEAKER_00: As in, we uh we have special, let's say, education robots and special collaboration with vocational schools in different countries and special packages where we also develop the training material and we can provide uh certification to vocational schools for people learning to be robot programmers or robot operators.
• 26:46
  SPEAKER_00: But also that inside the factory, we of course want the people uh inside the factory to be self-sufficient.
• 26:52
  SPEAKER_00: Because imagine for a lot of, let's say, a global company, they typically have an automation team centrally located, I don't know whether in Germany or somewhere, but the actual factories are could be in Eastern Europe, could be in Mexico, could be in many remote, remote places without professional engineers who know automation.
• 27:09
  SPEAKER_00: So then the people, the operators and maintenance staff in those plants have to be able to manage the robot themselves.
• 27:16
  SPEAKER_00: So for us, what we do is that of course we provide training, but we also make the user interface easy, intuitive, so that those people can, you know, uh manage when something goes wrong, but also maybe do limited reprogramming of the robot.
• 27:32
  SPEAKER_02: Yeah.
• 27:32
  SPEAKER_02: Yeah.
• 27:33
  SPEAKER_02: I the uh the the uh obviously like the uh there there has been a lot of excitement around the the humanoid form factor, but I think that that is well, I know that I've been watching it and reporting on it, a lot of it, a lot of it has has come back down to earth to a certain extent.
• 27:50
  SPEAKER_02: I think people are being um more pragmatic about it.
• 27:54
  SPEAKER_02: Um obviously, you know, ABB along with every other company is is interested in different sorts of embodiments and and how you know the role that that physical AI is gonna play.
• 28:04
  SPEAKER_02: But I do you think that like at any point in the foreseeable future that that that the big heavy industrial arm is going away?
• 28:14
  SPEAKER_02: I mean it's it's good at what it does.
• 28:16
  SPEAKER_00: Yeah, no, I don't see the big heavy industrial arms uh are going away.
• 28:21
  SPEAKER_00: And I also don't see factories filled with humanoid robots.
• 28:24
  SPEAKER_00: The thing with humanoid robots, I mean, it's very attractive to imagine that because they look like a human, they can do the things that my human workers on the shop floor are doing today.
• 28:33
  SPEAKER_00: But as um, you know, some uh people have mentioned these days at the Devils Tech Summit, is that we're limiting our imagination on what robots can do.
• 28:43
  SPEAKER_02: If you're if your factory is producing backflips, then you're fine.
• 28:47
  SPEAKER_00: Yeah, exactly.
• 28:48
  SPEAKER_00: But also the robots can do more.
• 28:50
  SPEAKER_00: There is no reason to limit, for example, having eyes facing front, why not having eyes facing back?
• 28:55
  SPEAKER_00: Only one thumb instead of two thumbs, fingers have been backwards.
• 28:59
  SPEAKER_00: So for me who have been building robots for 10 years, the form follows the function.
• 29:03
  SPEAKER_00: So first you need to understand the problem, and then you need to understand, okay, what is the best robot or form factor?
• 29:09
  SPEAKER_00: What are the features that's required to solve this problem?
• 29:12
  SPEAKER_00: And you know, humanoid robots are designed to do many things, right?
• 29:17
  SPEAKER_00: That's the beauty of it.
• 29:18
  SPEAKER_00: They're supposed to be able to do A, B, C, and D.
• 29:21
  SPEAKER_00: The problem is that they don't do any of these things very well, at least not better than industrial robot.
• 29:26
  SPEAKER_00: If you look at some of the industrial demonstrations of what the humanoid robots do, for example, loading a metal part onto a fixture, it is, you know, kind of walking like uh shakily.
• 29:38
  SPEAKER_00: It is, you know, not very precise in how it places a part, and it's much slower than industrial robots.
• 29:44
  SPEAKER_00: So once the humanoid robot can actually do this task, it will be pitted against other robots that can do the same task, but faster and cheaper.
• 29:53
  SPEAKER_00: And for those companies who are manufacturing cars or others, cost, ROI is so critical that they're not going to be willing to pay for another extra joint that doesn't add value to this application.
• 30:06
  SPEAKER_00: This is why I think for industrial manufacturing, cost, ROI, efficiency, um, reliability is always going to the main KPIs.
• 30:15
  SPEAKER_00: And there a humanoid robot is not going to be competitive versus uh industrial robot.
• 30:20
  SPEAKER_02: Yeah, I think I think another big piece to consider too is in in at least most cases that I've seen, um, I don't I don't know this this is necessarily universal across the board, but the big robot arms that we're talking about right now are usually in a cage.
• 30:36
  SPEAKER_02: You know, they usually have oftentimes like a physical safety system.
• 30:40
  SPEAKER_02: And you know, one, obviously that's that's for human safety, but two, it's because uh largely speaking, the jobs that they're doing, like they don't need humans within close proximity to do the jobs.
• 30:50
  SPEAKER_02: But um at least for the moment, uh, you know, the majority of roles that I see humanoids fulfilling requires uh people to be around.
• 31:00
  SPEAKER_02: And and you know, safety is another huge mountain that needs to be climbed.
• 31:06
  SPEAKER_00: Absolutely.
• 31:07
  SPEAKER_00: And that's something that people are not talking about when it comes to humanoid robots, right?
• 31:11
  SPEAKER_00: Today this humanoids have no safety, certainly not, you know, dual channel redundant safety that is required in industrial manufacturing.
• 31:19
  SPEAKER_00: And the fact that, for example, the the low um the two legs, you know, just standing up up right, this requires a a lot of battery power and two poses also safety problems.
• 31:30
  SPEAKER_02: Dynamically stabilized.
• 31:31
  SPEAKER_02: If the power goes out, they fall down.
• 31:33
  SPEAKER_00: Exactly.
• 31:33
  SPEAKER_00: So safety is something that needs to be solved.
• 31:35
  SPEAKER_00: How it interacts with the human operator also needs to be solved.
• 31:39
  SPEAKER_00: Uh now, collaborative robots that I have been working on is intended actually to be put into exactly these places where humans are working.
• 31:47
  SPEAKER_00: You know, we started with a dual UMI robot, which was designed exactly like a human form factor to fit onto the electronics assembly line next to human co-workers and interact with them in a collaborative way.
• 32:01
  SPEAKER_00: Since then, we have moved on to more single-arm collaborative robots for cost efficiency productivity, which is what we, you know, after talking to our customers, realize that is what the customers are looking for.
• 32:13
  SPEAKER_00: They need to solve the problem, but it needs to be simple and cost-efficient and very reliable.
• 32:18
  SPEAKER_00: And of course, safe.
• 32:20
  SPEAKER_00: So moving out of the steel cage doesn't necessarily need to be a humanoid robot.
• 32:26
  SPEAKER_00: And just because it looks like a human also doesn't mean that it has this type of safety that is required for, say, human robot collaboration.
• 32:35
  SPEAKER_02: Looking again, looking at your resume, like within ABB, there's a really interesting uh role in there, which is internal incubator manager, right?
• 32:45
  SPEAKER_02: So you're working within ABB robotics, obviously, like the the entire time that I'm coming from a startup background, right?
• 32:53
  SPEAKER_02: I was at TechRunch for nine years.
• 32:55
  SPEAKER_02: Um you you you talked to me about incubators.
• 32:58
  SPEAKER_02: I'm thinking about startups.
• 32:59
  SPEAKER_02: I mean, what what is what does an incubator mean within the context of a company like ABB?
• 33:05
  SPEAKER_02: Is it related to what we're talking about in terms of like diversifying form factors, in terms of getting into physical AI?
• 33:13
  SPEAKER_02: Or, you know, are we talking about startups or spin-offs, things like that?
• 33:17
  SPEAKER_00: Um, not startups or a spin-off, but it's an internal innovation effort, if you will.
• 33:23
  SPEAKER_00: So we have, you know, organization structures, processes.
• 33:26
  SPEAKER_00: But if we want to enter a new market, do something new, work with something different, sometimes we need a more lean structure to that's over to overcome the existing organizational structure and to have a different governance.
• 33:38
  SPEAKER_00: So that was the idea.
• 33:40
  SPEAKER_00: And I actually pitched the idea for this incubator to our top management.
• 33:43
  SPEAKER_00: And because I realized that you know, collaborative robots is not just um, you know, a form factor or safety, the technical parts, but it's so much the human robot collaboration part, the user experience, how to make this robot really useful for the users.
• 33:58
  SPEAKER_00: And also we're targeting a new type of customers, small and medium-sized companies, typically companies who are using robots for the first time.
• 34:06
  SPEAKER_00: So we can't follow the same playbook that we have been doing for the large industrial arms.
• 34:11
  SPEAKER_00: So that's why I pitched to start this internal focus group, a SWAT team, if you will, and also to bring in diverse talent.
• 34:19
  SPEAKER_00: Diversity and inclusion for me is very important.
• 34:23
  SPEAKER_02: Human diversity.
• 34:24
  SPEAKER_00: Human diversity and inclusion, absolutely.
• 34:26
  SPEAKER_00: So to bring in a cross-functional team, so we have people from also extern, I mean, a lot of external talent uh combined with internal, but cross-functional from communications and um user experience to application development, to software mechanical development to application expertise.
• 34:47
  SPEAKER_00: So really a mixed team to uh work together to figure out the solution of this problem, to really go to the customers, test the robots at new customer size.
• 34:55
  SPEAKER_00: We literally had a robot under our arm walking, knocking on doors in the Munich area that we were in, uh, small businesses or people who never tried robots before to get them to try our robots.
• 35:08
  SPEAKER_02: Like a door-to-door salesperson, whatever wants to.
• 35:10
  SPEAKER_00: Get them to come and say, hey, robots is not scary.
• 35:13
  SPEAKER_00: Come and try it out.
• 35:14
  SPEAKER_00: We want to do user testing with people like you because we want to enable people like you, the small businesses, to be able to use the robots.
• 35:22
  SPEAKER_02: Yeah, well, one of the things I was I was gonna ask, uh, you know, is it with within you know both of these companies, so you know, you were at your last company for 10 years, you're now at at uh ABB for I guess about 10 years in 2026.
• 35:36
  SPEAKER_02: But you've had a lot of different roles within each of the company, and it sounds like maybe to a certain extent, part of that is just sort of being proactive and like suggesting a new role or a new like something that doesn't even already exist within the company.
• 35:52
  SPEAKER_00: Yes, of course.
• 35:53
  SPEAKER_00: I mean, for me, my career has been driven by doing something new, like we talked about earlier.
• 35:58
  SPEAKER_00: You know, I was from technology, I wanted to try business, and then you know, collaborative robots was appealing because it was new.
• 36:04
  SPEAKER_00: I was the first person responsible for COVID at ABB.
• 36:07
  SPEAKER_00: It didn't, the category didn't exist before.
• 36:10
  SPEAKER_00: And then, you know, starting this incubator, focusing on the user, user journey, end-to-end user journey.
• 36:15
  SPEAKER_00: This was also new for ABB.
• 36:17
  SPEAKER_00: Uh and yeah, now physical AI is uh it's a new thing that is exciting, brings down so many new opportunities and possibilities.
• 36:26
  SPEAKER_00: So that is what I mean.
• 36:28
  SPEAKER_02: So the category so the category didn't exist but within ABB.
• 36:32
  SPEAKER_02: Um I mean, what what is that conversation like?
• 36:36
  SPEAKER_02: You know, again, like you're this big like multinational corporation.
• 36:40
  SPEAKER_02: Hey, you know, why don't you try something completely new?
• 36:43
  SPEAKER_02: Um I I as as somebody with a hardware background, you know, we talk about this a lot in terms of innovation.
• 36:51
  SPEAKER_02: This is why we see a lot of innovations from startups, is because um they're they're smaller and they can they can take those risks.
• 36:58
  SPEAKER_02: It's a lot harder to take a risk within a large company.
• 37:01
  SPEAKER_02: So it's very um very bold of you to go to a company and say, hey, like this is something entirely new that we should give a shot.
• 37:10
  SPEAKER_00: I would say the person who brought me into ABB from TetraPak is the one who was very bold because he thought he knew that collected robots is something new and he needed a new perspective and not something, somebody within ABB who is using the old playbook.
• 37:25
  SPEAKER_00: So then when I went in, I had free hands.
• 37:28
  SPEAKER_00: I could decide how to define a strategy, how to build a product roadmap.
• 37:32
  SPEAKER_00: Now, when it comes to pitching the internal startup, I think that is, like you say, for a big company, you know, this kind of internal incubator or internal startup is a way to manage new experimentation while managing the risk.
• 37:45
  SPEAKER_00: So, but of course, we also work with external startups and we also leverage the new technology and innovation from external startups.
• 37:54
  SPEAKER_02: It's an interesting transition, and and I'm kind of curious like what your sense is in what he saw in your existing role that he was like, oh, this is the person that I want to bring on to run Cobots for AUV.
• 38:07
  SPEAKER_00: I he wanted somebody who was customer-facing, market-oriented.
• 38:12
  SPEAKER_00: He said uh yeah, some of the previous uh product managers were more focused on you know spending more time with RD, but not out on the field visiting customers.
• 38:21
  SPEAKER_00: And for me, I mean, Cold Boss is new.
• 38:24
  SPEAKER_00: So that's the only thing I did for the couple of years, first couple of years was I was just constantly on the road traveling to different countries, meeting the local team, of course, but also visiting customers in, I mean, I was in toy factories in Thailand, electronics factories in Vietnam, and uh, you know, car companies in the US, and you know, personal uh, sorry, white good companies in in Italy, I mean, just everywhere.
• 38:50
  SPEAKER_00: Uh, I've even been to factories in Slovenia or, you know, so just fast, of course, all around China looking at the different industries and factories there.
• 39:00
  SPEAKER_00: I mean, it's just so fascinating.
• 39:02
  SPEAKER_00: And what amazed me is that in even high labor cost countries such as the US and Germany, you still see a lot of manual labor doing very simple things such as machine tending, machine loading CNC machine loading, manual, just pick and place loading.
• 39:16
  SPEAKER_00: I mean, there's so much potential for robot automation.
• 39:20
  SPEAKER_00: A lot of the barrier for them was one they they thought robots were complex.
• 39:24
  SPEAKER_00: You know, the traditional large industrial robots behind fences, they thought it was expensive, complex.
• 39:29
  SPEAKER_00: They never used a robot before.
• 39:31
  SPEAKER_00: And that is what I wanted to achieve with the cobots to make them more uh user-friendly and more approachable for these types of users so that they can feel, wow, I can, you know, this is the first robot I and that's often the case.
• 39:44
  SPEAKER_00: The cobalt is the first robot these small, medium-sized companies deploy.
• 39:48
  SPEAKER_00: Then when they feel more comfortable, they can um upgrade or not upgrade, but expand their automation uh park to incorporate more industrial robots because they actually work in the same way.
• 39:59
  SPEAKER_02: When when was it clear, you know, with within the context of ABV robotics that that like you it was time to really start focusing?
• 40:10
  SPEAKER_02: It sounds like really the kind of the line share of your efforts on physical AI versus the hard coding, which we've been doing for you know decades now.
• 40:20
  SPEAKER_00: Well, I mean, the cart hard coding continues, right?
• 40:22
  SPEAKER_00: So you we have a lot of um real applications.
• 40:25
  SPEAKER_02: But but you know, we were talking earlier, and and it sounded like you're like, all right, this is like kind of going forward to a certain extent, like physical AI, like that's what we're focused on.
• 40:34
  SPEAKER_00: Yes.
• 40:34
  SPEAKER_00: So that's the new application.
• 40:36
  SPEAKER_00: So the things that we can solve with today's um programming reliable, we we solve them that.
• 40:42
  SPEAKER_00: So we're not trying to apply physical AI to all the problems.
• 40:46
  SPEAKER_00: But there are problems that cannot be solved with today's systems, which require, for example, working unstructuring environments, require, you know, adaptation.
• 40:54
  SPEAKER_00: The robot needs to adapt its behavior based on the sensor input.
• 40:58
  SPEAKER_00: And this is where now we can apply physical AI to solve this problem.
• 41:02
  SPEAKER_00: I mean, another problem like random bin picking.
• 41:05
  SPEAKER_00: Yeah, right?
• 41:06
  SPEAKER_00: When I started 10 years ago, this was the holy grail of robotics.
• 41:10
  SPEAKER_00: It was very difficult to solve.
• 41:11
  SPEAKER_00: And even if it, when it was solved, it was not at the speed that was required to meet the cycle time.
• 41:16
  SPEAKER_00: I mean, I personally have been with multiple random bin picking projects which failed because the cycle time was just not fast enough because the vision system needs time to take pictures and to analyze the data.
• 41:27
  SPEAKER_00: Now this has improved with time.
• 41:28
  SPEAKER_00: And now with AI-based vision, you can more advanced with random bin picking.
• 41:34
  SPEAKER_00: So the same thing with other problems that um some of the problems have traditionally been difficult for automation to solve.
• 41:42
  SPEAKER_00: Now with additional AI capabilities on top when it comes to sensing, to interpretation of the sensing material, to also enhance dexterity to more autonomy of the AI brain, if you will, or the AI model behind the robot can handle some of these flexible tasks.
• 42:00
  SPEAKER_00: Um, handling flexible components also is another very hard problem.
• 42:04
  SPEAKER_00: Yeah.
• 42:04
  SPEAKER_00: Yeah.
• 42:05
  SPEAKER_02: Yeah.
• 42:05
  SPEAKER_02: The the again, like when traditionally, you know, when I when I think of ABV, um, really when I think of like the big four, again, I think of of arms, you know, in in manufacturing.
• 42:17
  SPEAKER_02: But um if you if we're talking about more flexibility and and the things that something like the i opens up, like I think that invariably is going to have to open you up to the possibility of like a much broader range of hardware form factors in order to sort of fulfill some of those tasks.
• 42:38
  SPEAKER_00: Well, we look at form factor that makes sense.
• 42:41
  SPEAKER_00: So absolutely, we're not limited, I mean, we're not limited to any kind of form factor.
• 42:45
  SPEAKER_00: In fact, robots today already have maybe there's just not so dramatic variations such as having a humanoid robot.
• 42:52
  SPEAKER_00: For example, for palletizing, you can have four-axis robots because that is the fastest type of industrial palletizer.
• 42:58
  SPEAKER_00: For welding, you can have additional access when it comes to positioners, and you have often, you know, linear rotational tracks, which enable the robot to reach more places, et cetera.
• 43:08
  SPEAKER_00: So often and also mobile mobility, right?
• 43:11
  SPEAKER_00: Robots that are now often on a mobile platform can be more mobile.
• 43:14
  SPEAKER_00: So all of these actually count as additional form factors for the robot.
• 43:18
  SPEAKER_00: Now, if you're talking about multiple arms, we it's possible for AVB to put two arms, three arms, four arms uh uh in such a configuration.
• 43:27
  SPEAKER_02: So now if you I think it's so funny because I think is really funny is like, you know, we we we've always been having this conversation, you know, what what makes a humanoid, and it's just like it just seems like it's just getting more and more abstracted.
• 43:41
  SPEAKER_02: You know what I mean?
• 43:42
  SPEAKER_02: To the point where it's like, all right, I mean, you can you can put two you can put bidexterous manipulation on an AMR, you know, and that is kind of a humanoid, and that's kind of uh in a sense what you're talking about.
• 43:57
  SPEAKER_00: Oh yeah.
• 43:57
  SPEAKER_00: I mean, now I think more and more there are new companies.
• 44:00
  SPEAKER_00: Companies like um Genesis AI release a NEO, which has two arms and I think a head, but it doesn't look like a human.
• 44:06
  SPEAKER_02: I think there are it looks like a bunch of iPads.
• 44:10
  SPEAKER_00: Uh but I mean there, I just everybody I see like new companies doing something.
• 44:14
  SPEAKER_00: Okay, I'm manual manipulator, but that don't look like a human.
• 44:18
  SPEAKER_00: And I think this is exactly the point that the form should follow the function that you know, the robot should be designed to do the function it that's solve the problem it needs to solve.
• 44:27
  SPEAKER_00: It does not need to look like a human.
• 44:29
  SPEAKER_02: Yeah, but but obviously, you know, um uh bidexterous manipulation is a huge, huge unlock.
• 44:36
  SPEAKER_02: I mean, dexterous manipulation is a really big lock.
• 44:38
  SPEAKER_02: Everybody's chasing a hand right now.
• 44:41
  SPEAKER_02: You don't need a hand for a lot of certainly a lot of manufacturing jobs, but a lot of them you, you know, it having dexterity would be huge.
• 44:50
  SPEAKER_00: More dexterity, absolutely.
• 44:52
  SPEAKER_00: This is one of the skills that we're adding to our physical AI skill stack, if you will.
• 44:57
  SPEAKER_00: So absolutely, we're looking at how to make our robots more dexterous.
• 45:00
  SPEAKER_02: And would that would that require uh new or different hardware?
• 45:04
  SPEAKER_00: Uh we're not, I mean, I'm I'm already building the next generation of robots as we start.
• 45:10
  SPEAKER_00: So I mean, so of course we we look at you know, how can we solve the problem and what makes sense um yeah to solve those future problems.
• 45:17
  SPEAKER_02: Yeah.
• 45:18
  SPEAKER_02: Um but but the the other side of it is, you know, uh while two arms and mobility are huge unlocks, these are also like really big, these are really hard problems to solve, right?
• 45:32
  SPEAKER_02: I mean, no, you gave me a look like it isn't, but like dexterity, I mean, like it seems like nav navigation itself is like pretty solved right now.
• 45:41
  SPEAKER_02: Uh, you know, there are still some some issues there, but actually like dexterous gripping on a moving platform is seems like a real hard thing to do.
• 45:50
  SPEAKER_00: Dexterous gripping is a hard problem.
• 45:52
  SPEAKER_00: Yeah.
• 45:52
  SPEAKER_00: Five-finger gripper, this is uh very hard.
• 45:55
  SPEAKER_00: I mean, I think that's the new holy grail of uh, let's say, robotics or physical AI right now, because you can't just program all the 20 joins in this five-finger hand.
• 46:04
  SPEAKER_00: But um, I mean, for us, we uh we don't build grippers.
• 46:07
  SPEAKER_00: Actually, we have a third-party partner ecosystem who work together with us and they build like plug-and-play N-of-arm tooling that can be used.
• 46:15
  SPEAKER_00: And also very often, uh our system integrators design dedicated N-of-arm tooling depending on what needs to, what object needs to be handled.
• 46:23
  SPEAKER_00: But obviously, dexterity is really important in order for a robot system to handle different objects without tool changing, because tool changers, you know, that takes time, et cetera, and it's less flexible.
• 46:34
  SPEAKER_00: So, this is definitely a topic that we're looking into.
• 46:37
  SPEAKER_00: Uh, I think another important part is the human-robot interaction.
• 46:41
  SPEAKER_00: Yeah.
• 46:41
  SPEAKER_00: How are you know humans supposed to interact with robots?
• 46:44
  SPEAKER_02: Yeah, you've been a co-op person for like a decade now.
• 46:47
  SPEAKER_00: So and uh, you know, for for an operator on the shelf floor, it's hard to understand how the robot works, and it's hard for them to um to, let's say, program or teach or instruct the robot.
• 46:57
  SPEAKER_00: So this is something that I think there needs to be more work done.
• 47:01
  SPEAKER_00: You know, now there's a lot of AI startups working on, you know, dual arm manipulation, et cetera.
• 47:07
  SPEAKER_00: But how are you going to teach the workers to teach the robots?
• 47:11
  SPEAKER_02: That's what Rethink was trying to do a long time ago.
• 47:13
  SPEAKER_02: Yeah.
• 47:13
  SPEAKER_02: A long time ago.
• 47:15
  SPEAKER_02: Maybe a little bit early, but that but that is another, I think that's another answer to the question that I asked before, as far as like keeping human beings relevant when there are robots on the floor.
• 47:26
  SPEAKER_00: Yeah, absolutely.
• 47:27
  SPEAKER_00: Um, I mean, now in all I think AI, any kind of AI application, there's human in the loop, right?
• 47:32
  SPEAKER_00: And in fact, a lot of these human-white robots are deployed using teleoperations, where at least there's always a person ready to teleoperate when as needed.
• 47:40
  SPEAKER_00: So right now, I think human is definitely in the loop.
• 47:43
  SPEAKER_00: But when it comes to human robot collaboration, it's also about, you know, you talk about by manual, mobile, et cetera.
• 47:49
  SPEAKER_00: You have more degrees of freedom.
• 47:51
  SPEAKER_00: And that makes more complex programming, teaching, instruction, whatever you want to call it.
• 47:57
  SPEAKER_00: So, and that increases the complexity in the system.
• 48:00
  SPEAKER_00: So we need to make it usable for the target end user.
• 48:03
  SPEAKER_00: And also we have to define the target end user.
• 48:06
  SPEAKER_00: Like many years ago, you know, in the beginning of COBOS, some people were saying that um the collaborative robots will be so easy to use that.
• 48:13
  SPEAKER_00: You just move the arm and then but the end user can teach them themselves, right?
• 48:17
  SPEAKER_00: You don't need the system integrators.
• 48:19
  SPEAKER_00: Yeah, 10 years later, that didn't happen because there's so many, the devil is in the details.
• 48:23
  SPEAKER_02: Yeah, that's why Rethink didn't really work out.
• 48:26
  SPEAKER_00: Exactly.
• 48:26
  SPEAKER_00: So system integrators are needed for COBOS today.
• 48:29
  SPEAKER_00: They might be you know integrating much less than a conventional industrial robot, but they're still needed to make sure the application actually works for the end user.
• 48:38
  SPEAKER_00: And the end user often they don't want to do this.
• 48:40
  SPEAKER_00: They're producing their potato chips or whatever.
• 48:43
  SPEAKER_00: They don't want to be, you know, an automation person as well.
• 48:46
  SPEAKER_00: And this is also valid for the bimanual mobile robots.
• 48:51
  SPEAKER_00: We need to deliver them also with a user interface that is usable for the target user, whether the target user is a new type of physical AI system integrator or the end user.
• 49:02
  SPEAKER_02: Yeah, they they it seems like the system integrators are kind of like the unsung Unsung heroes.
• 49:12
  SPEAKER_02: Um, as you're as as the focus is really evolving, like specifically within um ABB robotics, I mean, to what extent is that uh a product of the company?
• 49:26
  SPEAKER_02: Obviously, like there's a transition happening at the company, of the company effectively changing hands from ABB to soft bank.
• 49:33
  SPEAKER_00: Yeah, so within the soft bank group investments ecosystem, their goal is artificial superintelligence and um you know a leader in physical AI.
• 49:42
  SPEAKER_02: It's a modest goal.
• 49:44
  SPEAKER_00: It's a modest goal, yeah.
• 49:45
  SPEAKER_00: But um I think this is very much in line with what ABB robotics was already trying to do in you know making our robots more autonomous, more versatile, adding more AI skills to it.
• 49:57
  SPEAKER_00: And um within, you know, I what I think makes sense is that within the SoftBank group, they already have investments in the AI companies, whether it's open AI or skilled AI, and also in um chipmakers such as ARM and the infrastructure related to AI.
• 50:12
  SPEAKER_00: So under SoftBank, I think we can accelerate our pace towards fiscal AI and really unlock those new opportunities that it brings.
• 50:20
  SPEAKER_02: Yeah, it it's interesting.
• 50:22
  SPEAKER_02: You know, um obviously like SoftBank has has been involved in robotics in in different ways throughout the years.
• 50:30
  SPEAKER_02: And yeah, it it sounds like they're very much building this this portfolio.
• 50:34
  SPEAKER_02: Um as of this recording, the deal hasn't been finalized yet.
• 50:39
  SPEAKER_02: But do you get the sense that what they're working towards is really uh like sort of cross company within the portfolio collaboration that you'll be working with the other teams to build?
• 50:52
  SPEAKER_00: Oh, um I imagine, yes.
• 50:55
  SPEAKER_00: Uh after we're part of the the Sound Bank group, absolutely.
• 50:59
  SPEAKER_00: Um this is part of their strategy towards artificial superintelligence is to have a strategic portfolio of companies that add value towards this uh this goal.
• 51:09
  SPEAKER_00: Yes.
• 51:09
  SPEAKER_02: Yeah.
• 51:10
  SPEAKER_02: Yeah, it's it's interesting too.
• 51:11
  SPEAKER_02: I you know, it's uh I feel like I feel like Japan is really kind of starting to make a kind of comeback on robotics.
• 51:18
  SPEAKER_02: I I I I I I'm not entirely sure why, but I've maybe it's just getting more coverage, but I've been seeing a lot more Japanese companies in robotics than I had over the last several years.
• 51:28
  SPEAKER_02: They were incredibly early to it, obviously.
• 51:31
  SPEAKER_02: Um a lot of that is uh aging population, assistive robotics.
• 51:37
  SPEAKER_02: Um but now, you know, people are talking about the US, China, to a certain extent Europe.
• 51:44
  SPEAKER_02: Obviously, Japan's a really small country, but it's it got outsized influence.
• 51:49
  SPEAKER_02: Um but it seems like more and more like Japan is really becoming like a bigger and bigger player within robotics.
• 51:55
  SPEAKER_00: I think actually if you look at the big four, I mean, you know, Japan has always been a big player in robotics.
• 52:02
  SPEAKER_00: Uh I mean they're not um when it comes to the manufacturing base, like the deployment size, of course, you know, China, Germany, US have much more manufacturing base and therefore more robots deployed.
• 52:15
  SPEAKER_00: But Japan as a supplier.
• 52:17
  SPEAKER_00: Sure.
• 52:17
  SPEAKER_02: I I suppose I'm skewed because I like I deal primarily with startups.
• 52:21
  SPEAKER_02: Yeah.
• 52:22
  SPEAKER_00: But also components, you know, Japanese uh technology has been supplying the core components of a lot of the robots.
• 52:29
  SPEAKER_00: Yeah.
• 52:29
  SPEAKER_02: So as as we're uh at time right now and wrapping up the conversation, there was one more thing that really jumped out at me in your resume.
• 52:39
  SPEAKER_02: Um 2022 to 2023.
• 52:43
  SPEAKER_02: Do you do you remember what your role was?
• 52:46
  SPEAKER_00: I was uh I was a mother.
• 52:48
  SPEAKER_02: It says full-time parent.
• 52:49
  SPEAKER_02: Yeah, which I really uh I really appreciate.
• 52:52
  SPEAKER_02: Like one, actually like having that on your LinkedIn, in your CV.
• 52:56
  SPEAKER_02: Obviously, like you've had this great career, you've been in the industry for a really long time.
• 53:01
  SPEAKER_02: Was it a difficult decision to make to sort of step away for a year?
• 53:07
  SPEAKER_00: Um, I think when I was younger, yes, it did feel like it was very difficult to take time off from the career.
• 53:14
  SPEAKER_00: But then I realized, you know, it has to happen sooner or later.
• 53:17
  SPEAKER_00: I was in a good place, I would say.
• 53:19
  SPEAKER_00: You know, I had my incubator, it was running really well.
• 53:22
  SPEAKER_00: Yeah.
• 53:22
  SPEAKER_00: So what I did to help with the transition was I was preparing my succession plan already, you know, way ahead of time.
• 53:29
  SPEAKER_00: So I had a plan.
• 53:30
  SPEAKER_00: I was able to hand it over to somebody that I trusted.
• 53:33
  SPEAKER_02: You seem like someone with a lot of plans.
• 53:35
  SPEAKER_02: Like you're you seem like you got things planned up until like 2030 at this point, at least.
• 53:40
  SPEAKER_00: Uh no, I wouldn't say so.
• 53:42
  SPEAKER_00: But uh let's say uh six months to one year.
• 53:45
  SPEAKER_02: Okay, it's better than you know, I'm like two days out max.
• 53:49
  SPEAKER_02: Okay.
• 53:49
  SPEAKER_00: Well, in this case, because this is a big project, right?
• 53:52
  SPEAKER_02: So I was so succession plan is the project being uh building a family or being okay.
• 53:59
  SPEAKER_00: So so I think it's important to make sure there's a succession plan in place, you know, and also uh one topic we haven't touched upon is how I have always encouraged diversity inclusion in the technology sector.
• 54:12
  SPEAKER_00: I do a lot to encourage more women uh in STEM fields, girls in STEM fields, but also I'm an advocate for hiring more women in robotics.
• 54:22
  SPEAKER_00: You know, I have with my incubator, I have more women in it.
• 54:25
  SPEAKER_00: And I mean, technical, very skilled technical women still today in my team.
• 54:29
  SPEAKER_02: And and that's I mean, that that is something that like that I see like I've seen like visually, like anecdotally, like changing.
• 54:37
  SPEAKER_02: Just at events that I go to, it seems like the the ratio is is changing year to year.
• 54:44
  SPEAKER_00: Yes, that's great, but not changing fast enough.
• 54:46
  SPEAKER_00: I mean, there's still far less women.
• 54:49
  SPEAKER_00: So, you know, I I always I often get comments from male managing management colleagues who are like, how do you hire all these great women?
• 54:56
  SPEAKER_00: You know, and I'm like, okay, so there are certain tips and tricks I can give to all your listeners as well.
• 55:02
  SPEAKER_00: Okay, so number one, you need to look at how your job ad is written, right?
• 55:07
  SPEAKER_00: If you write five, 10 years experience in robotics, you're never gonna find a woman because 10 years ago there were even fewer women in the field than today.
• 55:15
  SPEAKER_00: And even for men, there's few that have 10 years experience in robotics.
• 55:19
  SPEAKER_00: That's a shit.
• 55:20
  SPEAKER_00: You need to remove those kind of hard criteria.
• 55:22
  SPEAKER_02: Yeah, so so so there's experience, but but you but do you do you feel like there's the the pool as far as uh when we're specifically talking about um uh women that as far as like uh what their focus was in higher education, that there are more and more of them that have been studying robotics?
• 55:42
  SPEAKER_00: Well, I think 10 years ago, even when I was in university, more than 10 years ago, mechatronics, mechanical engineering was not appealing to me.
• 55:49
  SPEAKER_00: Software engineering and something within more human human aspect of it, like uh human robot interaction or human computer interaction, that was more appealing to me.
• 55:58
  SPEAKER_00: So that's why I think there are fewer female mechatronics mechanical engineers, probably.
• 56:03
  SPEAKER_00: That's why if you're going to find good ones, you need to maybe relax your criteria.
• 56:07
  SPEAKER_00: You know, everybody knows we learn on the job.
• 56:10
  SPEAKER_00: So I focus on the soft aspects, you know, the learning, the team, uh teamwork capability, especially now, you know, with AI, people are able to learn things faster.
• 56:21
  SPEAKER_00: But what really stands out for a person, how they deliver in your team, is their ability to learn new knowledge, their ability to collaborate with others and their open-mindedness and not sticking to one point of view.
• 56:35
  SPEAKER_00: And also having a diverse team adds to that, right?
• 56:38
  SPEAKER_00: If you design a product or software with only one type of demographic, you're going to miss requirements from other parts, especially today when you know when it comes to robotics.
• 56:48
  SPEAKER_00: China is buying half, right?
• 56:50
  SPEAKER_00: More than half of the world's robot.
• 56:52
  SPEAKER_00: You know, if it's designed by European engineers in Europe, how are you going to succeed selling this product to China?
• 56:57
  SPEAKER_00: Yeah, right?
• 56:58
  SPEAKER_00: You have to, of course, understand, you know, the Chinese perspective.
• 57:01
  SPEAKER_00: You need to have people from different cultures, backgrounds on your team.
• 57:05
  SPEAKER_00: You need to have, of course, women and men on your team.
• 57:07
  SPEAKER_00: So you have to open up for this diversity in order, you know, designing your robot for left-handed, right-handed people, your interface.
• 57:14
  SPEAKER_00: What if somebody's colorblind?
• 57:15
  SPEAKER_00: 20% of men are color weak.
• 57:17
  SPEAKER_00: So all of these things need to be considered.
• 57:20
  SPEAKER_02: The the greatest thing, so I'm a cisgendered white male.
• 57:23
  SPEAKER_02: I don't know if you noticed that about me, but uh that that is in fact the case.
• 57:27
  SPEAKER_02: The single thing that really like brought that into sharp release for me is there is a video online of one of those electronic like soap dispensers, and a white hand goes and it and it dispenses a soap, and a black person puts their hand underneath and like it doesn't recognize it.
• 57:43
  SPEAKER_02: The sensor doesn't notice it.
• 57:44
  SPEAKER_02: And I think that's that like really boils down what you're talking about.
• 57:48
  SPEAKER_00: Very good example.
• 57:49
  SPEAKER_00: But yes, things such as that.
• 57:51
  SPEAKER_00: So that's why you know diversity and inclusion is very important.
• 57:54
  SPEAKER_00: Also, when you're designing a robot, we're selling robots to factories with a lot of um um operators, workers.
• 58:02
  SPEAKER_00: You know, a lot of those factories, the workers are females, right?
• 58:05
  SPEAKER_00: So of course you need to consider the users, the people who are selling next robots, and you need to get females on your team.
• 58:10
  SPEAKER_00: So I was saying number one is look at your job app.
• 58:14
  SPEAKER_00: Don't focus on the hard requirements that you probably can know that the females are not.
• 58:18
  SPEAKER_00: And the best way is to run it by some of your female colleagues and colleagues from other backgrounds.
• 58:23
  SPEAKER_00: Let them read it.
• 58:24
  SPEAKER_00: Does it sound appealing to them?
• 58:26
  SPEAKER_00: Would they apply for this job?
• 58:27
  SPEAKER_00: If not, change it.
• 58:28
  SPEAKER_02: Yeah, I I want to, because I I wanted to make a point on that too, because you you know you said relax requirements, and I don't think it's even relax requirements so much as sort of like change what the focus of the skill set is.
• 58:40
  SPEAKER_02: Like you do obviously have to get really competent people who are really good, but as you said, who can like learn on the job for example.
• 58:49
  SPEAKER_02: So in a way, it's almost like not relaxing so much as sort of like uh adjusting the criteria.
• 58:56
  SPEAKER_00: So instead of writing master's degree in engineering mechatronics, I write people who have who are interested in robots have been playing automation or technology.
• 59:05
  SPEAKER_00: You have demonstrated um, you know, experience using, designing, or in some way playing with any kind of technology, software, hardware, mechatronics for a number of years, right?
• 59:17
  SPEAKER_00: Because all of these skills are then transferable.
• 59:19
  SPEAKER_00: If you're interested in technology, if you're a tinkler, you know, you then you can transfer your technology from software to um adjacent.
• 59:27
  SPEAKER_00: So and it's a cross-discipline.
• 59:28
  SPEAKER_00: You're building a robot, it's not just hardware, it's software as well, right?
• 59:31
  SPEAKER_00: It's a very cross-built-disciplined uh field.
• 59:35
  SPEAKER_02: Yeah, I I uh uh my my psychiatrist once told me, uh she said, uh you're very interested in people, and I appreciated that.
• 59:44
  SPEAKER_02: And I realized I'm like, oh, that's like the one thing that's gotten me by on my job all these years, is is is having that curiosity, you know, and being able to ask those questions.
• 59:55
  SPEAKER_02: And I I think like that's also kind of what you're getting at, too, is like a really um important part of of intellect is is that curiosity and and and and that desire and drive to learn new things.
• 1:00:11
  SPEAKER_00: Absolutely.
• 1:00:12
  SPEAKER_00: For you, people skills, this is transferable between different jobs.
• 1:00:16
  SPEAKER_00: You know, I often at some events uh with you know young women who are asking me, like, oh, I'm interested in robotics, but I don't have this background, how can I get in?
• 1:00:24
  SPEAKER_00: I would say focus on your transferable skills, right?
• 1:00:27
  SPEAKER_00: There are many skills that are transferable, no matter what field you're in.
• 1:00:30
  SPEAKER_00: Focus on promoting them when you're you know applying for a job because chances are they're they're useful in in that company, in this company as well, whatever the product or field is in.
• 1:00:40
  SPEAKER_00: Yeah.
• 1:00:41
  SPEAKER_02: Andy, thank you so much for taking the time.
• 1:00:44
  SPEAKER_00: It was a pleasure talking to you, Brian.
• 1:00:48
  SPEAKER_02: Thanks again to Andy, to ABB, and the organizers of the Dodos Tech Summit.
• 1:00:53
  SPEAKER_02: If you've been enjoying the show, you can help us out by liking and subscribing and checking out our newsletter.
• 1:00:59
  SPEAKER_02: You'll find that over at automated.fm.
• 1:01:02
  SPEAKER_02: Thank you for tuning in, and we will see you soon with another episode of Automated.