And they looked absolutely ridiculous doing it:
Usain Bolt's 2009 world record for the 100-meter dash still remains unbroken at 9.58 seconds, but only if we're talking about humans. During the second annual World Humanoid Robot Games that kicked off in Beijing on Saturday, the Tiangong Ultra humanoid robot completed the 100-meter dash in 9.39 seconds during a preliminary heat. Developed by Beijing Humanoid Robot Innovation Center, the record-holding robot overtook Honor's Lightning robot, which completed the sprint in 9.47 seconds.
Despite the impressive finish times, don't expect these humanoid robots [1:23 Really funny. --Ed] to be used for any commercial purposes yet. After their sprints, they slammed directly into pads and some even required human assistance to be stretchered off the course. However, it's still a major improvement from last year's inaugural World Humanoid Robot Games, where the Tiangong Ultra won the 100-meter dash in 21.5 seconds.
Aside from this Robot Olympics event, Beijing hosted a half-marathon for Chinese companies to show off the capabilities of their humanoid robots. The event had its fair share of crashes too, but Honor's Lightning robot managed to secure the gold medal with a finish time of 50 minutes and 26 seconds, which also surpassed another human record. The World Humanoid Robot Games in Beijing kicked off August 22 and will run until the 26, hosting thousands of robot participants competing in more than just track and field events, including soccer, table tennis and boxing.
(Score: 5, Interesting) by bzipitidoo on Thursday August 27, @12:25AM (6 children)
Super simple sports that require little thought and judgment, and only one skill, always were going to be the first place robots would excel. Footraces are about as simple as it can get. Skiing on the other hand can be much more difficult. It's harder to ski than to run. Further, I guess current practice is to give the skiers lots of experience with a particular slope. Suppose they had to ski a slope they've never seen before? For footraces, one racetrack is much like another, all as level as possible. Not so ski slopes. I suspect robots have more work to do before they could beat humans at skiing a slope for the first time. But I think someday they will.
Another way to diversify are events such as the decathlon. A sport such as chess boxing is another direction, but it would have to be something other than chess, some mental game that computers haven't mastered. The day may come that there are no such games. Even generating a board game the contestants can't have practiced beforehand, like Fischer Random chess but an even larger space, maybe Fischer Random chess with some random assortment of fairy chess pieces, I suspect computers are already able to trounce humans at that.
(Score: 3, Interesting) by JoeMerchant on Thursday August 27, @01:02AM
Have you seen Boston Dynamics' demos? I could easily imagine them tearing up the moguls, and possibly optimizing a slalom run - especially if they design the body for aerodynamics.
🌻🌻🌻🌻✌️ [google.com]
(Score: 3, Interesting) by JoeMerchant on Thursday August 27, @01:07AM (3 children)
> some mental game that computers haven't mastered
You remember AlphaZero? That was an eternity ago in AI years.
https://deepmind.google/blog/muzero-mastering-go-chess-shogi-and-atari-without-rules/ [deepmind.google]
Apparently Nim is one that the AlphaZero struggles with:
https://wild.maths.org/play-win-nim [maths.org]
🌻🌻🌻🌻✌️ [google.com]
(Score: 3, Interesting) by bzipitidoo on Thursday August 27, @11:00AM (2 children)
I was thinking of Arimaa, a game designed to flummox AI. And it did. From 2002 to 2014, the world champ of Arimaa was human. In 2015, a program good enough to beat all the humans debuted.
Surprising that Nim is AI's weakness. Seems that neural nets aren't good at figuring out that a game has a forced win. How do they do at Connect 4, another game with a forced win?
(Score: 2) by sneftel on Friday August 28, @07:52AM (1 child)
Why would you assume that? And what makes you think that chess does not have a forced win?
(Score: 2) by bzipitidoo on Friday August 28, @10:27PM
I do not think that chess has a forced win, but it's not impossible that it does. The 3 possible outcomes for perfect play in chess are a draw, or a forced win for white, or a forced win for black. I'd put the odds of it being a draw at over 99.99%. Until chess is solved, we can't be certain. Chess is solvable, but solving it may be impractical, taking more computing resources than we can expect to ever have. Parts of chess are solved, namely all endgames with fewer than some small number of pieces -- 7, I think. In any case, over the past 29 years, computer chess engines have opened up a tremendous lead over humans. We don't even bother with man vs machine matchups any more, the engines are that dominant. 2005 was the last year that a human beat a good computer engine without a handicap.
I can only guess why neural nets are bad at Nim. It's actually simple to write a program to calculate the outcome of a Nim game, and what moves to play. Seems neural nets just can't grasp the key to Nim.
(Score: 5, Interesting) by sneftel on Thursday August 27, @08:50AM
Biomechanical research suggests it's easier to ski than run (see Hans-Christer Holmberg's work on the former). Yes, I know, it feels harder, but that's because you've been walking virtually your whole life, and skiing for perhaps a few thousand hours at most. But bipedal locomotion requires maintaining dynamic equilibrium in a complicated, chaotic system (even on a flat, featureless track!) whereas skiing involves more direct control of weight transfer. I don't know what the difficult bits of robot skiing would be, but I don't think it's correct to classify running as a "super simple sport" and skiing as not.
(Score: 4, Insightful) by jb on Thursday August 27, @07:42AM
Pretty sure using any kind of motor counts as cheating in a running race.
If it didn't then the world record would almost certainly be held by a dwarf (or perhaps by a child, if the rules allow children to compete against adults) given the obvious power to weight ratio advantage when aided by motors.
(Score: 2) by VLM on Thursday August 27, @03:17PM
I haven't had my breakfast tea yet, but if I did my math correctly... isn't 100 meters in less than 8 seconds about a third of a G constant horizontal acceleration?
We're getting near the limit of sneaker vs pavement dynamic friction, aren't we?
Like it would be "very impressive" to maintain 2G horizontal acceleration without putting sticky tar on the shoes, essentially impossible given normal track athletic materials, and 2G horizontal acceleration would cover 100 meters in like three seconds, I think. Maybe running on stairs (up or down...) people could maintain over 2G horizontal acceleration, which is strange to think about.
D=1/2*A*T*T so...
This feels about right for engineering back of envelope calculations but in the context of a running dude, or humanoid shape anyway, I have to be missing something. I think perhaps the dude does "a large percentage" of his total acceleration pushing off the chock/starting blocks. I think if I can deadlift / squat well over my body weight I could achieve well over 2G acceleration if we were having a 50 centimeter dash. I wonder if I could beat that Usain Bolt dude in a 50 centimeter dash; runners and cardio bunnies are not known for muscles. I'm not used to moving that fast for obvious weight lifting safety reasons, tendon strength may be the limiter.
Basic kinematics, plus or minus being half asleep, is one of those scenarios where a classic old-fashioned slide rule is intuitively helpful when you just slide the cursor along and think about it.
(Score: 1) by oldeschool on Thursday August 27, @04:39PM
A machine is faster than a human?
granted it's a bipedal machine, but come on...