A joke of today's share.

after conquering go, Deepmind is still exploring the possibilities of artificial intelligence. Just in the last few days, they released a new video showing people a new challenge of AI: learning parkour.

of course, this is far from the cool difficulty of parkour videos. In fact, the researchers are testing whether their artificial intelligence can become self-taught and learn how to move forward in different terrain. They set up various model conditions in the virtual world and let artificial intelligence try to move forward in an environment full of obstacles for reinforcement learning (reinforcement learning).

and then we see a rising wonder.

(there is no doubt that the process is full of failures)

now AI's performance in sports still seems to make people laugh to the point of split, but "moving like a human" is a big challenge for computers and robots themselves, and it's actually a good attempt.

at present, artificial intelligence only controls the movement of models in the virtual world, but in the future, more sophisticated technology may be used in the real world to make robots better adapt to the environment.

Elegant and striking, nobody should be without our sweet 16 dresses. Enjoy the best online shopping experience ever, at your fingertips.

original video: preprint of https://www.youtube.com/watch?v=hx_bgoTF7bs

's corresponding paper: https://arxiv.org/abs/1707.02286

is interested in further understanding, please take a look at