Tag Archives: Zero

#437912 “Boston Dynamics Will Continue to ...

Last week’s announcement that Hyundai acquired Boston Dynamics from SoftBank left us with a lot of questions. We attempted to answer many of those questions ourselves, which is typically bad practice, but sometimes it’s the only option when news like that breaks.

Fortunately, yesterday we were able to speak with Michael Patrick Perry, vice president of business development at Boston Dynamics, who candidly answered our questions about Boston Dynamics’ new relationship with Hyundai and what the near future has in store.

IEEE Spectrum: Boston Dynamics is worth 1.1 billion dollars! Can you put that valuation into context for us?

Michael Patrick Perry: Since 2018, we’ve shifted to becoming a commercial organization. And that’s included a number of things, like taking our existing technology and bringing it to market for the first time. We’ve gone from zero to 400 Spot robots deployed, building out an ecosystem of software developers, sensor providers, and integrators. With that scale of deployment and looking at the pipeline of opportunities that we have lined up over the next year, I think people have started to believe that this isn’t just a one-off novelty—that there’s actual value that Spot is able to create. Secondly, with some of our efforts in the logistics market, we’re getting really strong signals both with our Pick product and also with some early discussions around Handle’s deployment in warehouses, which we think are going to be transformational for that industry.

So, the thing that’s really exciting is that two years ago, we were talking about this vision, and people said, “Wow, that sounds really cool, let’s see how you do.” And now we have the validation from the market saying both that this is actually useful, and that we’re able to execute. And that’s where I think we’re starting to see belief in the long-term viability of Boston Dynamics, not just as a cutting-edge research shop, but also as a business.

Photo: Boston Dynamics

Boston Dynamics says it has deployed 400 Spot robots, building out an “ecosystem of software developers, sensor providers, and integrators.”

How would you describe Hyundai’s overall vision for the future of robotics, and how do they want Boston Dynamics to fit into that vision?

In the immediate term, Hyundai’s focus is to continue our existing trajectories, with Spot, Handle, and Atlas. They believe in the work that we’ve done so far, and we think that combining with a partner that understands many of the industries in which we’re targeting, whether its manufacturing, construction, or logistics, can help us improve our products. And obviously as we start thinking about producing these robots at scale, Hyundai’s expertise in manufacturing is going to be really helpful for us.

Looking down the line, both Boston Dynamics and Hyundai believe in the value of smart mobility, and they’ve made a number of plays in that space. Whether it’s urban air mobility or autonomous driving, they’ve been really thinking about connecting the digital and the physical world through moving systems, whether that’s a car, a vertical takeoff and landing multi-rotor vehicle, or a robot. We are well positioned to take on robotics side of that while also connecting to some of these other autonomous services.

Can you tell us anything about the kind of robotics that the Hyundai Motor Group has going on right now?

So they’re working on a lot of really interesting stuff—exactly how that connects, you know, it’s early days, and we don’t have anything explicitly to share. But they’ve got a smart and talented robotics team that’s working in a variety of directions that shares overlap with us. Obviously, a lot of things related to autonomous driving shares some DNA with the work that we’re doing in autonomy for Spot and Handle, so it’s pretty exciting to see.

What are you most excited about here? How do you think this deal will benefit Boston Dynamics?

I think there are a number of things. One is that they have an expertise in hardware, in a way that’s unique. They understand and appreciate the complexity of creating large complex robotic systems. So I think there’s some shared understanding of what it takes to create a great hardware product. And then also they have the resources to help us actually build those products with them together—they have manufacturing resources and things like that.

“Robotics isn’t a short term game. We’ve scaled pretty rapidly but if you start looking at what the full potential of a company like Boston Dynamics is, it’s going to take years to realize, and I think Hyundai is committed to that long-term vision”

Another thing that’s exciting is that Hyundai has some pretty visionary bets for autonomous driving and unmanned aerial systems, and all of that fits very neatly into the connected vision of robotics that we were talking about before. Robotics isn’t a short term game. We’ve scaled pretty rapidly for a robotics company in terms of the scale of robots we’ve able to deploy in the field, but if you start looking at what the full potential of a company like Boston Dynamics is, it’s going to take years to realize, and I think Hyundai is committed to that long-term vision.

And when you’ve been talking with Hyundai, what are they most excited about?

I think they’re really excited about our existing products and our technology. Looking at some of the things that Spot, Pick, and Handle are able to do now, there are applications that many of Hyundai’s customers could benefit from in terms of mobility, remote sensing, and material handling. Looking down the line, Hyundai is also very interested in smart city technology, and mobile robotics is going to be a core piece of that.

We tend to focus on Spot and Handle and Atlas in terms of platform capabilities, but can you talk a bit about some of the component-level technology that’s unique to Boston Dynamics, and that could be of interest to Hyundai?

Creating very power-dense actuator design is something that we’ve been successful at for several years, starting back with BigDog and LS3. And Handle has some hydraulic actuators and valves that are pretty unique in terms of their design and capability. Fundamentally, we have a systems engineering approach that brings together both hardware and software internally. You’ll often see different groups that specialize in something, like great mechanical or electrical engineering groups, or great controls teams, but what I think makes Boston Dynamics so special is that we’re able to put everything on the table at once to create a system that’s incredibly capable. And that’s why with something like Spot, we’re able to produce it at scale, while also making it flexible enough for all the different applications that the robot is being used for right now.

It’s hard to talk specifics right now, but there are obviously other disciplines within mechanical engineering or electrical engineering or controls for robots or autonomous systems where some of our technology could be applied.

Photo: Boston Dynamics

Boston Dynamics is in the process of commercializing Handle, iterating on its design and planning to get box-moving robots on-site with customers in the next year or two.

While Boston Dynamics was part of Google, and then SoftBank, it seems like there’s been an effort to maintain independence. Is it going to be different with Hyundai? Will there be more direct integration or collaboration?

Obviously it’s early days, but right now, we have support to continue executing against all the plans that we have. That includes all the commercialization of Spot, as well as things for Atlas, which is really going to be pushing the capability of our team to expand into new areas. That’s going to be our immediate focus, and we don’t see anything that’s going to pull us away from that core focus in the near term.

As it stands right now, Boston Dynamics will continue to be Boston Dynamics under this new ownership.

How much of what you do at Boston Dynamics right now would you characterize as fundamental robotics research, and how much is commercialization? And how do you see that changing over the next couple of years?

We have been expanding our commercial team, but we certainly keep a lot of the core capabilities of fundamental robotics research. Some of it is very visible, like the new behavior development for Atlas where we’re pushing the limits of perception and path planning. But a lot of the stuff that we’re working on is a little bit under the hood, things that are less obvious—terrain handling, intervention handling, how to make safe faults, for example. Initially when Spot started slipping on things, it would flail around trying to get back up. We’ve had to figure out the right balance between the robot struggling to stand, and when it should decide to just lock its limbs and fall over because it’s safer to do that.

I’d say the other big thrust for us is manipulation. Our gripper for Spot is coming out early next year, and that’s going to unlock a new set of capabilities for us. We have years and years of locomotion experience, but the ability to manipulate is a space that’s still relatively new to us. So we’ve been ramping up a lot of work over the last several years trying to get to an early but still valuable iteration of the technology, and we’ll continue pushing on that as we start learning what’s most useful to our customers.

“I’d say the other big thrust for us is manipulation. Our gripper for Spot is coming out early next year, and that’s going to unlock a new set of capabilities for us. We have years and years of locomotion experience, but the ability to manipulate is a space that’s still relatively new to us”

Looking back, Spot as a commercial robot has a history that goes back to robots like LS3 and BigDog, which were very ambitious projects funded by agencies like DARPA without much in the way of commercial expectations. Do you think these very early stage, very expensive, very technical projects are still things that Boston Dynamics can take on?

Yes—I would point to a lot of the things we do with Atlas as an example of that. While we don’t have immediate plans to commercialize Atlas, we can point to technologies that come out of Atlas that have enabled some of our commercial efforts over time. There’s not necessarily a clear roadmap of how every piece of Atlas research is going to feed over into a commercial product; it’s more like, this is a really hard fundamental robotics challenge, so let’s tackle it and learn things that we can then benefit from across the company.

And fundamentally, our team loves doing cool stuff with robots, and you’ll continue seeing that in the months to come.

Photo: Boston Dynamics

Spot’s arm with gripper is coming out early next year, and Boston Dynamics says that’s going to “unlock a new set of capabilities for us.”

What would it take to commercialize Atlas? And are you getting closer with Handle?

We’re in the process of commercializing Handle. We’re at a relatively early stage, but we have a plan to get the first versions for box moving on-site with customers in the next year or two. Last year, we did some on-site deployments as proof-of-concept trials, and using the feedback from that, we did a new design pass on the robot, and we’re looking at increasing our manufacturing capability. That’s all in progress.

For Atlas, it’s like the Formula 1 of robots—you’re not going to take a Formula 1 car and try to make it less capable so that you can drive it on the road. We’re still trying to see what are some applications that would necessitate an energy and computationally intensive humanoid robot as opposed to something that’s more inherently stable. Trying to understand that application space is something that we’re interested in, and then down the line, we could look at creating new morphologies to help address specific applications. In many ways, Handle is the first version of that, where we said, “Atlas is good at moving boxes but it’s very complicated and expensive, so let’s create a simpler and smaller design that can achieve some of the same things.”

The press release mentioned a mobile robot for warehouses that will be introduced next year—is that Handle?

Yes, that’s the work that we’re doing on Handle.

As we start thinking about a whole robotic solution for the warehouse, we have to look beyond a high power, low footprint, dynamic platform like Handle and also consider things that are a little less exciting on video. We need a vision system that can look at a messy stack of boxes and figure out how to pick them up, we need an interface between a robot and an order building system—things where people might question why Boston Dynamics is focusing on them because it doesn’t fit in with our crazy backflipping robots, but it’s really incumbent on us to create that full end-to-end solution.

Are you confident that under Hyundai’s ownership, Boston Dynamics will be able to continue taking the risks required to remain on the cutting edge of robotics?

I think we will continue to push the envelope of what robots are capable of, and I think in the near term, you’ll be able to see that realized in our products and the research that we’re pushing forward with. 2021 is going to be a great year for us. Continue reading

Posted in Human Robots

#437896 Solar-based Electronic Skin Generates ...

Replicating the human sense of touch is complicated—electronic skins need to be flexible, stretchable, and sensitive to temperature, pressure and texture; they need to be able to read biological data and provide electronic readouts. Therefore, how to power electronic skin for continuous, real-time use is a big challenge.

To address this, researchers from Glasgow University have developed an energy-generating e-skin made out of miniaturized solar cells, without dedicated touch sensors. The solar cells not only generate their own power—and some surplus—but also provide tactile capabilities for touch and proximity sensing. An early-view paper of their findings was published in IEEE Transactions on Robotics.

When exposed to a light source, the solar cells on the s-skin generate energy. If a cell is shadowed by an approaching object, the intensity of the light, and therefore the energy generated, reduces, dropping to zero when the cell makes contact with the object, confirming touch. In proximity mode, the light intensity tells you how far the object is with respect to the cell. “In real time, you can then compare the light intensity…and after calibration find out the distances,” says Ravinder Dahiya of the Bendable Electronics and Sensing Technologies (BEST) Group, James Watt School of Engineering, University of Glasgow, where the study was carried out. The team used infra-red LEDs with the solar cells for proximity sensing for better results.

To demonstrate their concept, the researchers wrapped a generic 3D-printed robotic hand in their solar skin, which was then recorded interacting with its environment. The proof-of-concept tests showed an energy surplus of 383.3 mW from the palm of the robotic arm. “The eSkin could generate more than 100 W if present over the whole body area,” they reported in their paper.

“If you look at autonomous, battery-powered robots, putting an electronic skin [that] is consuming energy is a big problem because then it leads to reduced operational time,” says Dahiya. “On the other hand, if you have a skin which generates energy, then…it improves the operational time because you can continue to charge [during operation].” In essence, he says, they turned a challenge—how to power the large surface area of the skin—into an opportunity—by turning it into an energy-generating resource.

Dahiya envisages numerous applications for BEST’s innovative e-skin, given its material-integrated sensing capabilities, apart from the obvious use in robotics. For instance, in prosthetics: “[As] we are using [a] solar cell as a touch sensor itself…we are also [making it] less bulkier than other electronic skins.” This, he adds, will help create prosthetics that are of optimal weight and size, thus making it easier for prosthetics users. “If you look at electronic skin research, the the real action starts after it makes contact… Solar skin is a step ahead, because it will start to work when the object is approaching…[and] have more time to prepare for action.” This could effectively reduce the time lag that is often seen in brain–computer interfaces.

There are also possibilities in the automation sector, particularly in electrical and interactive vehicles. A car covered with solar e-skin, because of its proximity-sensing capabilities, would be able to “see” an approaching obstacle or a person. It isn’t “seeing” in the biological sense, Dahiya clarifies, but from the point of view of a machine. This can be integrated with other objects, not just cars, for a variety of uses. “Gestures can be recognized as well…[which] could be used for gesture-based control…in gaming or in other sectors.”

In the lab, tests were conducted with a single source of white light at 650 lux, but Dahiya feels there are interesting possibilities if they could work with multiple light sources that the e-skin could differentiate between. “We are exploring different AI techniques [for that],” he says, “processing the data in an innovative way [so] that we can identify the the directions of the light sources as well as the object.”

The BEST team’s achievement brings us closer to a flexible, self-powered, cost-effective electronic skin that can touch as well as “see.” At the moment, however, there are still some challenges. One of them is flexibility. In their prototype, they used commercial solar cells made of amorphous silicon, each 1cm x 1cm. “They are not flexible, but they are integrated on a flexible substrate,” Dahiya says. “We are currently exploring nanowire-based solar cells…[with which] we we hope to achieve good performance in terms of energy as well as sensing functionality.” Another shortcoming is what Dahiya calls “the integration challenge”—how to make the solar skin work with different materials. Continue reading

Posted in Human Robots

#437624 AI-Powered Drone Learns Extreme ...

Quadrotors are among the most agile and dynamic machines ever created. In the hands of a skilled human pilot, they can do some astonishing series of maneuvers. And while autonomous flying robots have been getting better at flying dynamically in real-world environments, they still haven’t demonstrated the same level of agility of manually piloted ones.

Now researchers from the Robotics and Perception Group at the University of Zurich and ETH Zurich, in collaboration with Intel, have developed a neural network training method that “enables an autonomous quadrotor to fly extreme acrobatic maneuvers with only onboard sensing and computation.” Extreme.

There are two notable things here: First, the quadrotor can do these extreme acrobatics outdoors without any kind of external camera or motion-tracking system to help it out (all sensing and computing is onboard). Second, all of the AI training is done in simulation, without the need for an additional simulation-to-real-world (what researchers call “sim-to-real”) transfer step. Usually, a sim-to-real transfer step means putting your quadrotor into one of those aforementioned external tracking systems, so that it doesn’t completely bork itself while trying to reconcile the differences between the simulated world and the real world, where, as the researchers wrote in a paper describing their system, “even tiny mistakes can result in catastrophic outcomes.”

To enable “zero-shot” sim-to-real transfer, the neural net training in simulation uses an expert controller that knows exactly what’s going on to teach a “student controller” that has much less perfect knowledge. That is, the simulated sensory input that the student ends up using as it learns to follow the expert has been abstracted to present the kind of imperfect, imprecise data it’s going to encounter in the real world. This can involve things like abstracting away the image part of the simulation until you’d have no way of telling the difference between abstracted simulation and abstracted reality, which is what allows the system to make that sim-to-real leap.

The simulation environment that the researchers used was Gazebo, slightly modified to better simulate quadrotor physics. Meanwhile, over in reality, a custom 1.5-kilogram quadrotor with a 4:1 thrust to weight ratio performed the physical experiments, using only a Nvidia Jetson TX2 computing board and an Intel RealSense T265, a dual fisheye camera module optimized for V-SLAM. To challenge the learning system, it was trained to perform three acrobatic maneuvers plus a combo of all of them:

Image: University of Zurich/ETH Zurich/Intel

Reference trajectories for acrobatic maneuvers. Top row, from left: Power Loop, Barrel Roll, and Matty Flip. Bottom row: Combo.

All of these maneuvers require high accelerations of up to 3 g’s and careful control, and the Matty Flip is particularly challenging, at least for humans, because the whole thing is done while the drone is flying backwards. Still, after just a few hours of training in simulation, the drone was totally real-world competent at these tricks, and could even extrapolate a little bit to perform maneuvers that it was not explicitly trained on, like doing multiple loops in a row. Where humans still have the advantage over drones is (as you might expect since we’re talking about robots) is quickly reacting to novel or unexpected situations. And when you’re doing this sort of thing outdoors, novel and unexpected situations are everywhere, from a gust of wind to a jealous bird.

For more details, we spoke with Antonio Loquercio from the University of Zurich’s Robotics and Perception Group.

IEEE Spectrum: Can you explain how the abstraction layer interfaces with the simulated sensors to enable effective sim-to-real transfer?

Antonio Loquercio: The abstraction layer applies a specific function to the raw sensor information. Exactly the same function is applied to the real and simulated sensors. The result of the function, which is “abstracted sensor measurements,” makes simulated and real observation of the same scene similar. For example, suppose we have a sequence of simulated and real images. We can very easily tell apart the real from the simulated ones given the difference in rendering. But if we apply the abstraction function of “feature tracks,” which are point correspondences in time, it becomes very difficult to tell which are the simulated and real feature tracks, since point correspondences are independent of the rendering. This applies for humans as well as for neural networks: Training policies on raw images gives low sim-to-real transfer (since images are too different between domains), while training on the abstracted images has high transfer abilities.

How useful is visual input from a camera like the Intel RealSense T265 for state estimation during such aggressive maneuvers? Would using an event camera substantially improve state estimation?

Our end-to-end controller does not require a state estimation module. It shares however some components with traditional state estimation pipelines, specifically the feature extractor and the inertial measurement unit (IMU) pre-processing and integration function. The input of the neural networks are feature tracks and integrated IMU measurements. When looking at images with low features (for example when the camera points to the sky), the neural net will mainly rely on IMU. When more features are available, the network uses to correct the accumulated drift from IMU. Overall, we noticed that for very short maneuvers IMU measurements were sufficient for the task. However, for longer ones, visual information was necessary to successfully address the IMU drift and complete the maneuver. Indeed, visual information reduces the odds of a crash by up to 30 percent in the longest maneuvers. We definitely think that event camera can improve even more the current approach since they could provide valuable visual information during high speed.

“The Matty Flip is probably one of the maneuvers that our approach can do very well … It is super challenging for humans, since they don’t see where they’re going and have problems in estimating their speed. For our approach the maneuver is no problem at all, since we can estimate forward velocities as well as backward velocities.”
—Antonio Loquercio, University of Zurich

You describe being able to train on “maneuvers that stretch the abilities of even expert human pilots.” What are some examples of acrobatics that your drones might be able to do that most human pilots would not be capable of?

The Matty Flip is probably one of the maneuvers that our approach can do very well, but human pilots find very challenging. It basically entails doing a high speed power loop by always looking backward. It is super challenging for humans, since they don’t see where they’re going and have problems in estimating their speed. For our approach the maneuver is no problem at all, since we can estimate forward velocities as well as backward velocities.

What are the limits to the performance of this system?

At the moment the main limitation is the maneuver duration. We never trained a controller that could perform maneuvers longer than 20 seconds. In the future, we plan to address this limitation and train general controllers which can fly in that agile way for significantly longer with relatively small drift. In this way, we could start being competitive against human pilots in drone racing competitions.

Can you talk about how the techniques developed here could be applied beyond drone acrobatics?

The current approach allows us to do acrobatics and agile flight in free space. We are now working to perform agile flight in cluttered environments, which requires a higher degree of understanding of the surrounding with respect to this project. Drone acrobatics is of course only an example application. We selected it because it makes a stress test of the controller performance. However, several other applications which require fast and agile flight can benefit from our approach. Examples are delivery (we want our Amazon packets always faster, don’t we?), search and rescue, or inspection. Going faster allows us to cover more space in less time, saving battery costs. Indeed, agile flight has very similar battery consumption of slow hovering for an autonomous drone.

“Deep Drone Acrobatics,” by Elia Kaufmann, Antonio Loquercio, René Ranftl, Matthias Müller, Vladlen Koltun, and Davide Scaramuzza from the Robotics and Perception Group at the University of Zurich and ETH Zurich, and Intel’s Intelligent Systems Lab, was presented at RSS 2020. Continue reading

Posted in Human Robots

#437603 Throwable Robot Car Always Lands on Four ...

Throwable or droppable robots seem like a great idea for a bunch of applications, including exploration and search and rescue. But such robots do come with some constraints—namely, if you’re going to throw or drop a robot, you should be prepared for that robot to not land the way you want it to land. While we’ve seen some creative approaches to this problem, or more straightforward self-righting devices, usually you’re in for significant trade-offs in complexity, mobility, and mass.

What would be ideal is a robot that can be relied upon to just always land the right way up. A robotic cat, of sorts. And while we’ve seen this with a tail, for wheeled vehicles, it turns out that a tail isn’t necessary: All it takes is some wheel spin.

The reason that AGRO (Agile Ground RObot), developed at the U.S. Military Academy at West Point, can do this is because each of its wheels is both independently driven and steerable. The wheels are essentially reaction wheels, which are a pretty common way to generate forces on all kinds of different robots, but typically you see such reaction wheels kludged onto these robots as sort of an afterthought—using the existing wheels of a wheeled robot is a more elegant way to do it.

Four steerable wheels with in-hub motors provide control in all three axes (yaw, pitch, and roll). You’ll notice that when the robot is tossed, the wheels all toe inwards (or outwards, I guess) by 45 degrees, positioning them orthogonal to the body of the robot. The front left and rear right wheels are spun together, as are the front right and rear left wheels. When one pair of wheels spins in the same direction, the body of the robot twists in the opposite way along an axis between those wheels, in a combination of pitch and roll. By combining different twisting torques from both pairs of wheels, pitch and roll along each axis can be adjusted independently. When the same pair of wheels spin in directions opposite to each other, the robot yaws, although yaw can also be derived by adjusting the ratio between pitch authority and roll authority. And lastly, if you want to sacrifice pitch control for more roll control (or vice versa) the wheel toe-in angle can be changed. Put all this together, and you get an enormous amount of mid-air control over your robot.

Image: Robotics Research Center/West Point

The AGRO robot features four steerable wheels with in-hub motors, which provide control in all three axes (yaw, pitch, and roll).

According to a paper that the West Point group will present at the 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), the overall objective here is for the robot to reach a state of zero pitch or roll by the time the robot impacts with the ground, to distribute the impact as much as possible. AGRO doesn’t yet have a suspension to make falling actually safe, so in the short term, it lands on a foam pad, but the mid-air adjustments it’s currently able to make result in a 20 percent reduction of impact force and a 100 percent reduction in being sideways or upside-down.

The toss that you see in the video isn’t the most aggressive, but lead author Daniel J. Gonzalez tells us that AGRO can do much better, theoretically stabilizing from an initial condition of 22.5 degrees pitch and 22.5 degrees roll in a mere 250 milliseconds, with room for improvement beyond that through optimizing the angles of individual wheels in real time. The limiting factor is really the amount of time that AGRO has between the point at which it’s released and the point at which it hits the ground, since more time in the air gives the robot more time to change its orientation.

Given enough height, the current generation of AGRO can recover from any initial orientation as long as it’s spinning at 66 rpm or less. And the only reason this is a limitation at all is because of the maximum rotation speed of the in-wheel hub motors, which can be boosted by increasing the battery voltage, as Gonzalez and his colleagues, Mark C. Lesak, Andres H. Rodriguez, Joseph A. Cymerman, and Christopher M. Korpela from the Robotics Research Center at West Point, describe in the IROS paper, “Dynamics and Aerial Attitude Control for Rapid Emergency Deployment of the Agile Ground Robot AGRO.”

Image: Robotics Research Center/West Point

AGRO 2 will include a new hybrid wheel-leg and non-pneumatic tire design that will allow it to hop up stairs and curbs.

While these particular experiments focus on a robot that’s being thrown, the concept is potentially effective (and useful) on any wheeled robot that’s likely to find itself in mid-air. You can imagine it improving the performance of robots doing all sorts of stunts, from driving off ramps or ledges to being dropped out of aircraft. And as it turns out, being able to self-stabilize during an airdrop is an important skill that some Humvees could use to keep themselves from getting tangled in their own parachute lines and avoid mishaps.

Before they move on to Humvees, though, the researchers are working on the next version of AGRO named AGRO 2. AGRO 2 will include a new hybrid wheel-leg and non-pneumatic tire design that will allow it to hop up stairs and curbs, which sounds like a lot of fun to us. Continue reading

Posted in Human Robots

#437303 The Deck Is Not Rigged: Poker and the ...

Tuomas Sandholm, a computer scientist at Carnegie Mellon University, is not a poker player—or much of a poker fan, in fact—but he is fascinated by the game for much the same reason as the great game theorist John von Neumann before him. Von Neumann, who died in 1957, viewed poker as the perfect model for human decision making, for finding the balance between skill and chance that accompanies our every choice. He saw poker as the ultimate strategic challenge, combining as it does not just the mathematical elements of a game like chess but the uniquely human, psychological angles that are more difficult to model precisely—a view shared years later by Sandholm in his research with artificial intelligence.

“Poker is the main benchmark and challenge program for games of imperfect information,” Sandholm told me on a warm spring afternoon in 2018, when we met in his offices in Pittsburgh. The game, it turns out, has become the gold standard for developing artificial intelligence.

Tall and thin, with wire-frame glasses and neat brow hair framing a friendly face, Sandholm is behind the creation of three computer programs designed to test their mettle against human poker players: Claudico, Libratus, and most recently, Pluribus. (When we met, Libratus was still a toddler and Pluribus didn’t yet exist.) The goal isn’t to solve poker, as such, but to create algorithms whose decision making prowess in poker’s world of imperfect information and stochastic situations—situations that are randomly determined and unable to be predicted—can then be applied to other stochastic realms, like the military, business, government, cybersecurity, even health care.

While the first program, Claudico, was summarily beaten by human poker players—“one broke-ass robot,” an observer called it—Libratus has triumphed in a series of one-on-one, or heads-up, matches against some of the best online players in the United States.

Libratus relies on three main modules. The first involves a basic blueprint strategy for the whole game, allowing it to reach a much faster equilibrium than its predecessor. It includes an algorithm called the Monte Carlo Counterfactual Regret Minimization, which evaluates all future actions to figure out which one would cause the least amount of regret. Regret, of course, is a human emotion. Regret for a computer simply means realizing that an action that wasn’t chosen would have yielded a better outcome than one that was. “Intuitively, regret represents how much the AI regrets having not chosen that action in the past,” says Sandholm. The higher the regret, the higher the chance of choosing that action next time.

It’s a useful way of thinking—but one that is incredibly difficult for the human mind to implement. We are notoriously bad at anticipating our future emotions. How much will we regret doing something? How much will we regret not doing something else? For us, it’s an emotionally laden calculus, and we typically fail to apply it in quite the right way. For a computer, it’s all about the computation of values. What does it regret not doing the most, the thing that would have yielded the highest possible expected value?

The second module is a sub-game solver that takes into account the mistakes the opponent has made so far and accounts for every hand she could possibly have. And finally, there is a self-improver. This is the area where data and machine learning come into play. It’s dangerous to try to exploit your opponent—it opens you up to the risk that you’ll get exploited right back, especially if you’re a computer program and your opponent is human. So instead of attempting to do that, the self-improver lets the opponent’s actions inform the areas where the program should focus. “That lets the opponent’s actions tell us where [they] think they’ve found holes in our strategy,” Sandholm explained. This allows the algorithm to develop a blueprint strategy to patch those holes.

It’s a very human-like adaptation, if you think about it. I’m not going to try to outmaneuver you head on. Instead, I’m going to see how you’re trying to outmaneuver me and respond accordingly. Sun-Tzu would surely approve. Watch how you’re perceived, not how you perceive yourself—because in the end, you’re playing against those who are doing the perceiving, and their opinion, right or not, is the only one that matters when you craft your strategy. Overnight, the algorithm patches up its overall approach according to the resulting analysis.

There’s one final thing Libratus is able to do: play in situations with unknown probabilities. There’s a concept in game theory known as the trembling hand: There are branches of the game tree that, under an optimal strategy, one should theoretically never get to; but with some probability, your all-too-human opponent’s hand trembles, they take a wrong action, and you’re suddenly in a totally unmapped part of the game. Before, that would spell disaster for the computer: An unmapped part of the tree means the program no longer knows how to respond. Now, there’s a contingency plan.

Of course, no algorithm is perfect. When Libratus is playing poker, it’s essentially working in a zero-sum environment. It wins, the opponent loses. The opponent wins, it loses. But while some real-life interactions really are zero-sum—cyber warfare comes to mind—many others are not nearly as straightforward: My win does not necessarily mean your loss. The pie is not fixed, and our interactions may be more positive-sum than not.

What’s more, real-life applications have to contend with something that a poker algorithm does not: the weights that are assigned to different elements of a decision. In poker, this is a simple value-maximizing process. But what is value in the human realm? Sandholm had to contend with this before, when he helped craft the world’s first kidney exchange. Do you want to be more efficient, giving the maximum number of kidneys as quickly as possible—or more fair, which may come at a cost to efficiency? Do you want as many lives as possible saved—or do some take priority at the cost of reaching more? Is there a preference for the length of the wait until a transplant? Do kids get preference? And on and on. It’s essential, Sandholm says, to separate means and the ends. To figure out the ends, a human has to decide what the goal is.

“The world will ultimately become a lot safer with the help of algorithms like Libratus,” Sandholm told me. I wasn’t sure what he meant. The last thing that most people would do is call poker, with its competition, its winners and losers, its quest to gain the maximum edge over your opponent, a haven of safety.

“Logic is good, and the AI is much better at strategic reasoning than humans can ever be,” he explained. “It’s taking out irrationality, emotionality. And it’s fairer. If you have an AI on your side, it can lift non-experts to the level of experts. Naïve negotiators will suddenly have a better weapon. We can start to close off the digital divide.”

It was an optimistic note to end on—a zero-sum, competitive game yielding a more ultimately fair and rational world.

I wanted to learn more, to see if it was really possible that mathematics and algorithms could ultimately be the future of more human, more psychological interactions. And so, later that day, I accompanied Nick Nystrom, the chief scientist of the Pittsburgh Supercomputing Center—the place that runs all of Sandholm’s poker-AI programs—to the actual processing center that make undertakings like Libratus possible.

A half-hour drive found us in a parking lot by a large glass building. I’d expected something more futuristic, not the same square, corporate glass squares I’ve seen countless times before. The inside, however, was more promising. First the security checkpoint. Then the ride in the elevator — down, not up, to roughly three stories below ground, where we found ourselves in a maze of corridors with card readers at every juncture to make sure you don’t slip through undetected. A red-lit panel formed the final barrier, leading to a small sliver of space between two sets of doors. I could hear a loud hum coming from the far side.

“Let me tell you what you’re going to see before we walk in,” Nystrom told me. “Once we get inside, it will be too loud to hear.”

I was about to witness the heart of the supercomputing center: 27 large containers, in neat rows, each housing multiple processors with speeds and abilities too great for my mind to wrap around. Inside, the temperature is by turns arctic and tropic, so-called “cold” rows alternating with “hot”—fans operate around the clock to cool the processors as they churn through millions of giga, mega, tera, peta and other ever-increasing scales of data bytes. In the cool rows, robotic-looking lights blink green and blue in orderly progression. In the hot rows, a jumble of multicolored wires crisscrosses in tangled skeins.

In the corners stood machines that had outlived their heyday. There was Sherlock, an old Cray model, that warmed my heart. There was a sad nameless computer, whose anonymity was partially compensated for by the Warhol soup cans adorning its cage (an homage to Warhol’s Pittsburghian origins).

And where does Libratus live, I asked? Which of these computers is Bridges, the computer that runs the AI Sandholm and I had been discussing?

Bridges, it turned out, isn’t a single computer. It’s a system with processing power beyond comprehension. It takes over two and a half petabytes to run Libratus. A single petabyte is a million gigabytes: You could watch over 13 years of HD video, store 10 billion photos, catalog the contents of the entire Library of Congress word for word. That’s a whole lot of computing power. And that’s only to succeed at heads-up poker, in limited circumstances.

Yet despite the breathtaking computing power at its disposal, Libratus is still severely limited. Yes, it beat its opponents where Claudico failed. But the poker professionals weren’t allowed to use many of the tools of their trade, including the opponent analysis software that they depend on in actual online games. And humans tire. Libratus can churn for a two-week marathon, where the human mind falters.

But there’s still much it can’t do: play more opponents, play live, or win every time. There’s more humanity in poker than Libratus has yet conquered. “There’s this belief that it’s all about statistics and correlations. And we actually don’t believe that,” Nystrom explained as we left Bridges behind. “Once in a while correlations are good, but in general, they can also be really misleading.”

Two years later, the Sandholm lab will produce Pluribus. Pluribus will be able to play against five players—and will run on a single computer. Much of the human edge will have evaporated in a short, very short time. The algorithms have improved, as have the computers. AI, it seems, has gained by leaps and bounds.

So does that mean that, ultimately, the algorithmic can indeed beat out the human, that computation can untangle the web of human interaction by discerning “the little tactics of deception, of asking yourself what is the other man going to think I mean to do,” as von Neumann put it?

Long before I’d spoken to Sandholm, I’d met Kevin Slavin, a polymath of sorts whose past careers have including founding a game design company and an interactive art space and launching the Playful Systems group at MIT’s Media Lab. Slavin has a decidedly different view from the creators of Pluribus. “On the one hand, [von Neumann] was a genius,” Kevin Slavin reflects. “But the presumptuousness of it.”

Slavin is firmly on the side of the gambler, who recognizes uncertainty for what it is and thus is able to take calculated risks when necessary, all the while tampering confidence at the outcome. The most you can do is put yourself in the path of luck—but to think you can guess with certainty the actual outcome is a presumptuousness the true poker player foregoes. For Slavin, the wonder of computers is “That they can generate this fabulous, complex randomness.” His opinion of the algorithmic assaults on chance? “This is their moment,” he said. “But it’s the exact opposite of what’s really beautiful about a computer, which is that it can do something that’s actually unpredictable. That, to me, is the magic.”

Will they actually succeed in making the unpredictable predictable, though? That’s what I want to know. Because everything I’ve seen tells me that absolute success is impossible. The deck is not rigged.

“It’s an unbelievable amount of work to get there. What do you get at the end? Let’s say they’re successful. Then we live in a world where there’s no God, agency, or luck,” Slavin responded.

“I don’t want to live there,’’ he added “I just don’t want to live there.”

Luckily, it seems that for now, he won’t have to. There are more things in life than are yet written in the algorithms. We have no reliable lie detection software—whether in the face, the skin, or the brain. In a recent test of bluffing in poker, computer face recognition failed miserably. We can get at discomfort, but we can’t get at the reasons for that discomfort: lying, fatigue, stress—they all look much the same. And humans, of course, can also mimic stress where none exists, complicating the picture even further.

Pluribus may turn out to be powerful, but von Neumann’s challenge still stands: The true nature of games, the most human of the human, remains to be conquered.

This article was originally published on Undark. Read the original article.

Image Credit: José Pablo Iglesias / Unsplash Continue reading

Posted in Human Robots