Name: 谷歌的AlphaGo是如何擊敗職業巴杜克棋手的？ (How Google's AlphaGo beat a professional Baduk player)
Uploaded: 2021-01-14T06:00:47.000Z
Duration: 2 min 11 s
Description: 【看影片學英語】數萬部 YouTube 影片，搭配英漢字典即點即查，輕鬆掌握單字發音與用法，長久累積看電影不必再看字幕。

Those that tipped Professional Baduk Player and fellow human Lee Se-dol for victory over

AlphaGo earlier this week... may have underestimated Google's computer program.

Kim Ji-yeon explains why AlphaGo's victories are no fluke.

It seems as if Google's computer program AlphaGo has the capability to learn just like a human.

That's because it's based on an artificial intelligence program called "deep learning"...

an information processing technology that imitates human nerve cells.

But unlike humans, the program never ceases its continuous learning of Baduk, or Go...

not even for a second. AlphaGo is an improved version of previous

computer programs, which was only capable of preprogrammed moves.

"Deep learning is not a new concept in the IT world, but it only has become reality now

due to significant improvements in hardware including faster arithmetic abilities of CPUs,

faster internet connectivity, and widespread use of cloud technology... which allows thousands

of computers to come together to process and calculate the most optimal move."

AlphaGo is supposed to imitate how humans learn...but it may actually surpass human

intelligence in terms of higher learning capacity. AlphaGo practices Baduk some 3-thousand times

a day, a million times in a month. That would take a professional Baduk player

three games a day for a thousand years... to accomplish the same feat.

In fact, it continues to learn even while AlphaGo is playing against Professional Baduk

"The improved version of AI is on par with human intelligence, while earlier versions

were limited to certain features. In that sense, AlphaGo is an improved AI version that's

AlphaGo has been programmed to learn 160-thousand Baduk game sets and learned to play a million

Baduk game sets in a month on its own. Google's DeepMind, which developed AlphaGo,

says the ultimate goal for the program is to solve real world problems in the near future...

building on the learning process gained from playing Baduk.