UMA unveils its vision for the next generation of robots, Remi Cadene at Machina Summit

[00:00:02][music] >> This is our AI.

[00:00:08]We develop it at UMI.

[00:00:11]It controls a standard robot arm to automate the task of picking and sorting anchors of various colors.

[00:00:19]It's a single neural network.

[00:00:22]So, here nothing here is scripted.

[00:00:27]No pre-programmed trajectories, no hardcoded rules that tell it where to go or what to pick next.

[00:00:35]This robot sees the world through raw pixels and decides what to do next.

[00:00:44]Perception to action end to end.

[00:00:48]It's a bit like you and I when we reach out for a cup without even thinking about it.

[00:00:54]Yeah, this kind of behavior it's always amazing to watch.

[00:00:59]What we show with this AI is that it can operate for hours robustly with precision adapting to what it sees.

[00:01:11]And so what you see here is the beginning of something profound.

[00:01:17]AI is moving into the physical world embodying the next generation of robots.

[00:01:26]Robots that are mobile, they can move around, they can choose, they can manipulate uh different type of objects, and they are smart.

[00:01:36]They learn and adapt. And this robots will help us in our daily lives.

[00:01:42]They will be in manufacturing, in agriculture, and they will be even in services and in your homes.

[00:01:50]In labs, they will turn years of research in just a few months.

[00:01:54]We won't be limited by our capability to make but only by our imagination.

[00:02:02]We are entering a new era of limitless creativity.

[00:02:07]This might feel like a distant future, but it will happen much quicker than we expect because soon enough those robots they will be absolutely everywhere and they will be absolutely needed.

[00:02:20]And this is because we are facing three crises in our generation hitting us at the same time and converging on one bottleneck that these robots will solve.

[00:02:34]And so we need to build these robots now and we need to build them right.

[00:02:40]And that's why 9 months ago I co-founded UMA to build the Universal Mechanical Assistant.

[00:02:48]So let me tell you about these crises through a series of examples that might be a bit confusing at first sight.

[00:03:00]So first crisis demographic.

[00:03:04]In this image you can see a factory in southern Germany.

[00:03:09]Profitable but shutting down.

[00:03:13]Not because demand fell but because they could not hire the machinists, the welders, the technicians.

[00:03:20]They tried to raise wages.

[00:03:22]They tried to pay for relocation but nothing worked.

[00:03:26]And this is a well-documented phenomenon happening right now across Europe and accelerating.

[00:03:35]Second crisis environmental.

[00:03:38]The fact that we know how to stop climate change. The technology exists.

[00:03:44]We need to adapt and build infrastructure quickly enough. And for this you need someone that manufacture, assemble and install the the technology.

[00:03:56]And so in this picture, we illustrate that families stopped their plan to install solar panel on the roof because it just became too expensive when the labor is missing.

[00:04:12]The third crisis, it's about economy.

[00:04:15]It's linked to supply chain organization worldwide. So, I'm sure you remember COVID.

[00:04:24]Um we discovered that among the most powerful nations on Earth, some were struggling to manufacture their own masks. And we are just talking about a piece of paper and elastic.

[00:04:41]Another example is the Suez Canal blockage. Uh one ship stopped the whole the world economy for 2 weeks.

[00:04:49]And And then maybe recently you tried to purchase a new cell phone or a new smartphone. The price uh go went up and that's because of a single short point in uh Asia for the manufacturing of RAM. So, most of the goods that the world consume today are manufactured in just a few places, which create massive dependencies and geopolitical tension. And we are in a in a era where we need to avoid this at all cost.

[00:05:25]So, the few examples that I mentioned, they might feel unrelated, but they are not.

[00:05:34]Three crises, demographic, environmental, economic, and they converge on one bottleneck, which is labor, its availability and its cost. And this is accelerating.

[00:05:48]So, if you have to remember of only one statistic, it would be the number of children per family.

[00:05:56]Across every continent, it's going down.

[00:06:00]And we are almost below the replacement rate.

[00:06:05]That's the number at which the global population of the world just decrease.

[00:06:10]So, this is a a normal phenomenon.

[00:06:12]Just societies evolved.

[00:06:15]But for your information, the US, Europe, and China I were well below this rate already.

[00:06:24]As a result, 85 million workers will be short in less than 4 years from now.

[00:06:29]And this is the equivalent of 8.5 trillion in economic value just gone in less than 4 years.

[00:06:36]And and this can be combined GDP of the UK and Japan.

[00:06:42]So, three crises and one solution.

[00:06:48]We need to build this next generation of robots as quickly as possible because they will address the labor shortage and fill this gap.

[00:06:58]So, what are the challenges?

[00:07:02]First challenge, it's building the robot brain.

[00:07:06]This means that we need the ability I mean, those robots need the ability to learn as quickly as possible, as efficiently as possible, the task that we do.

[00:07:20]And so, at Uma, we took a unique approach.

[00:07:24]Even before scaling data and computes, we came back to the fundamentals and develop and focused on a radically efficient learning approach.

[00:07:37]We call it real-time learning.

[00:07:41]And it is inspired by how you and I we learn.

[00:07:45]Do you remember the first time you learn how to tie your shoes?

[00:07:50]I'm sure that uh your parents teach it teach you how to do it.

[00:07:56]Actually, they just show you the task.

[00:07:58]Then you fell loads.

[00:07:59]You try again.

[00:08:01]You practiced. And at some point, your hands just knew.

[00:08:05]And that's real-time learning.

[00:08:07]So now, please have a look to this video.

[00:08:17]So this it's our AI controlling two robot arms and automating the task of picking and scanning it and dropping various items. Four items that are extremely challenging uh with today's technology in terms of robotics.

[00:08:35]So the first item is uh cables that are deformable. Then you have shrimp quite slippery. You have mouse pads super thin. And then you have those bags that need to be handled with care because they are fragile.

[00:08:49]So you see the the AI retry retry and then it get it.

[00:08:54]Here, we are messing with it.

[00:09:01]It's amazing to watch, no?

[00:09:04]And finally, it succeed.

[00:09:07]We really tried hard to mess with it and uh still it succeeded.

[00:09:12]And this is how you go from one robot doing one task to billions of robots doing almost anything.

[00:09:22]So now, second challenge, building the body of the robot.

[00:09:28]At iTuring, our conviction is that to learn like us, to automate the task that we do, these robots need to be humanoids.

[00:09:38]They need a body to navigate our space and hence to handle our tools.

[00:09:47]And so, today I'm extremely excited to unveil our North Star design.

[00:10:05]We design it to be approachable, calm, and competent.

[00:10:11]A robot you would feel at ease having around in your workplace, but also at home. And here you can see it wearing clothes like futuristic looking kind of clothes.

[00:10:24]We design it with three safety design principle.

[00:10:29]Safety is fundamental. It at the core of what we do at Uma. And it means caring for people and looking after them. And first safety principle, it is lightweight.

[00:10:45]Less mass, less kinetic energy.

[00:10:48]We baked safety into the physics.

[00:10:51]Second, we borrow from aerospace redundance, independence, layers of safety that reinforce each other in terms of software systems.

[00:11:04]Third, our own world models is dedicated to safety.

[00:11:10]Before any action, our world model predict the outcome and then it's just enable the movements only when the path is clear.

[00:11:24]And I'm sure that the companies that win at robotics are not the one that ship the most units today.

[00:11:33]They are the one that earn the trust on the long term. And at UMA, we are building the safest, but also the more capable, most capable robots.

[00:11:44]Starting with our very first prototype developing our lab to validate the fundamentals.

[00:11:54]We call it version zero.

[00:12:10]AI, software, hardware, a small team, 9 months, designed and assembled in Paris. And so, I often get asked, "Why did you start UMA in Europe?" It's because it's the best market.

[00:12:28]That's where the situation is the most critical.

[00:12:32]Highest demands in terms of automation, highest cost of labor.

[00:12:36]That's why we started UMA in Europe.

[00:12:41]And North America and Asia will soon follow.

[00:12:45]In terms of go-to-market, three steps. First, logistics. That's where we have controlled environments.

[00:12:54]Uh we can deploy hundreds of robots at a single site, and demand is immediate immediate.

[00:13:00]Second step, manufacturing. That's where the requirement in terms of dexterity is higher.

[00:13:07]And third, homes, where this technology will increase the quality of life for everyone. So, now, we know what to build, how to build it, and where to deploy.

[00:13:24]But how do we build these robots quickly enough to address the crisis I talked about? Let me tell you about smartphones.

[00:13:35]From the first iPhone to as many as people on Earth, it took the industry only 12 years.

[00:13:45]But, a smartphone cannot build another smartphone. A robot can. So, each generation of human robots is going to build the next one, more capable, safer, and at a speed which is exponential.

[00:14:03]Robots building robots, that's how we go quickly enough. So, this next generation of robots, it is about making sure that the work that holds our society and our life together still gets done, and done for every one of us.

[00:14:24]Every generation faces challenges that shape the future, and this is our turn.

[00:14:31]And at Uma, we build this future with a deep sense of responsibility, humanistic values, and passion.

[00:14:41]Thank you.