1 Overview of Artificial Intelligence
1 Overview of Artificial Intelligence
1 Overview of Artificial Intelligence
From May 23 to 27, 2017, AlphaGo Zero won the world championship Ke Jie
with a total score of 3:0 in Wuzhen, China. In the team Go match held
on 26 May, AlphaGo defeated five world champion players, Chen Yaoye,
Tang Weixing, Zhou Ruiyang, Shi Yue, and Mi Yuting.
The workshop pointed that every aspect of learning or any other feature of
intelligence should in principle be so precisely described that a machine
can be made to simulate it.
1970, Marvin Minsky: "In from three to eight years we will have a
machine with the general intelligence of an average human being."
From left: Gordon Moore (Moore's law), John McCarthy (doctor of mathematics,
creator of the Lisp programming language), Marvin Minsky (developer of Robot
C, the world's earliest robot that can simulate human activity), Oliver
Gordon Selfridge (founder of pattern recognition and developer of the first
workable artificial intelligence program), and Ray Solomonoff (inventor of
algorithmic probability)
Many theories emerged during AI development: symbolicism, connectionism, and
actionism. All these have facilitated the development of AI.
These symbolists first adopted the term "artificial intelligence" in 1956.
Later, the heuristic algorithm, expert system, and knowledge engineering
theory and technology made considerable progress in the 1980s. Symbolicism
has stood out for a long time and contributed greatly to the development of
AI, especially the expert system. It was critical for applying AI to
engineering applications. Symbolicism remained to be a mainstream even other
AI theories emerged.
Infrastructure layer: As AI develops, the infrastructure development each
time greatly promoted the evolution of algorithms and technologies.
Computers grew up in the 1970s and gained popularity in the 1980s. The 1990s
witnessed the improvement of computers' computing speed and storage
capacity, as well as the rise of the Internet. All these contributed a lot
to AI's development. In the 21st century, the emergence of large-scale
clusters, the accumulation of big data, and the rise of GPU and
heterogeneous/low power consumption chips enhanced the computing power. As a
result, deep learning came into being, pushing AI to a new height. Mass
training data is an important driving force for the development of AI.
Disadvantages:
Deep neural networks are complex and require large-scale data training.
Deep learning is only suitable for perception and is not suitable for
cognition and inference.
Today, the development of AI has broken through a certain threshold. Today,
AI improves the performance and efficiency in different vertical fields. The
accuracy of computer vision, speech recognition, and NLP is also improved.
AI application scenarios gradually play an important role in the real
business world.
Signal processing:
Human speech detection: effectively detects the time when the human
speech starts and ends, and distinguishes the human speech from the
background sound.
Speech enhancement: further enhances the speech area and suppresses the
noise area, effectively reducing the attenuation of far-field speeches.
Handheld devices are used in the near-field communication (NFC)
scenario, while vehicle-mounted and smart home devices are used in far-
field scenarios.
Uber's vehicles are equipped with HD laser radars, which could detect
pedestrians at such a distance if the hardware functions properly. As
Zhou Guang told China Makter (ID: xjbmaker), this accident was very
likely caused by the malfunction of the vehicle's decision-making
module. For example, it might think that the pedestrian would not enter
the lane directly. However, human behavior is hard to accurately
predict. That is why the accident occurred.
Huang Yubin, who agreed with Zhou Guang, said that there might be
nothing wrong with the perception module. He thought that the decision-
making module might fail to judge that a pedestrian would show up
suddenly. Huang Yubin thought that Uber's overspeed was the strangest
thing. "A self-driving car would not speed up on its own. It does what
people have written in programs." Huang Yubin thought that Uber's
overspeed was likely to be a choice forced by road conditions. "There
is a saying that it is unsafe for you to drive at a speed of 35 miles
per hour when others are driving at the speed of 40 miles per hour."
He Tao, co-founder and vice president of XPENG Motors, thought that the
algorithms of current self-driving vehicles were not mature enough.
Pedestrians' illegal behavior is a major cause of car accidents.
However, predicting possible emergencies and solving them in algorithms
is one of the missions of automatic driving. If a self-driving car
fails to avoid pedestrians who violate the traffic rules, how can it
deal with more complex scenarios in China?
In 2013, the U.S. National Highway Traffic Safety Administration (NHTSA)
issued the Federal AutomatedVehicles Policy, which stipulated the
responsibilities in case of self-driving test accidents. That is, after a
vehicle is transformed into a self-driving vehicle by a third party, the
original manufacturer is not responsible for any personal injury, death, or
property damage caused by the self-driving test, unless there are proofs
that the vehicle has defects before being transformed. For example, Google
takes the full responsibility for the security when it uses Mercedes-Benz
for self-driving test.
Although AI has entered a new stage of learning (early stage), there is no
need to worry about the so-called "human-computer war" robot control,
because AlphaGo does not know that he is playing chess, I don't have to
invent a game of chess.
Analysts said in a report that a machine might be able to make a
conversation with human beings like a spouse by 2020.
You think your opponent is a competitor, but your opponent is an era. --Li
Shanzhao
Answer: 1.B, 2.B
Answer: 3. A 4.C