IT iFly Tek CEO Liu said 's in the is the best proof of its to learn and . [Photo/]

– By 2020, China- will be smart to gain to the , China's exam.

“Our goal is to make our smart to enter first-class in three to five years,” Liu , of tech firm Co. Ltd., said at the 's in .

Liu has three : , and . have or the human brain in the first two . , , the to think and , is far more .

“We have found the only way to crack ,” Liu says. He and his hold that human is the of of , oral or . “So the key is and .”

's

“It is easy for a robot to sit the are in ,” Liu says. “But it is hard for them to 80 of human and for first-class .”

The A12 lab of of , he says, is to pass the exam. Japan's Todai robot has the goal of with other who want to enter the of Tokyo by 2021.

tests such as math, and are easy for AI they can be with , Liu says. Japan's have the level of an in exams in math and .

The arts – or more , (NLU) – are . “A in or a test is the most ,” Liu says. “NLU is what on.”

While the is set, the goal is not to beat the human brain, but to human .

“Most never all ,” Liu says. Fast find their time in class while slow feel if they keep up. must often spend time of .

” to the of , and , AI can meet each 's needs and the 's now.” Liu shows how can notes on a piece of and give it a score. It can also a 's in his or her so the can .

South China's has used 's in the oral test.

More than 80 and are from 's and Liu the in is huge.

's

As Liu the , two big on each side of the stage were his into text a .

For this like a long- , but Liu said it had posed many .

He the of the smart home. If you want to the by voice, AI must your voice by noise in the house. When you are with the on, the robot must the sound of the TV. If you want the air to hear you from 5 away, the AI must “grasp” far-field that it to pick up the sound from a and from all .

” lies in , which out the text, , and ,” he says.

As the words on the , five with the in . After a of 1,000 words, the robot won with an rate of 99.29 , to 80.84 for the best .

“With this , can be on film,” Liu says.

Asked if might human , Hu Yu, vice of , , “Rest easy – excel in and, of , they are of data. But they .”

In the , Hu says, AI will help with . As for , “we will .”

“A lift his son 'he' is too heavy.” Can a robot tell who is “he” ?

“This is very easy for , but too hard for a robot,” Hu says. To the to be able to the .

its “Hyper Brain ” last year, with the aim of AI . “This , on the , NLU, but by , it's of the whole world, the and human ,” Hu says.

“A is on the basis of ,” Hu says. “When you see the face of a cat, the of cat soon forms in your brain. You begin to think of its mew and how its fur feels.”

has for , such as and audio . When they have , it will be by a for and , so that we can make . The of the will be via a , a loop that all of a human's .

“We don't have to copy the brain,” Hu says. “Our job is to find the true and of brain.”

Such raise the of a -style of the .

Hu says in AI will lead away from such . “The more we learn from our brain, the we can from self-, our , good and evil. In the , will be able to more , but not self-.”