Regarding the rumor of Jay Chou's new song, this one is really outrageous

Picture source: Jay Chou MV

Jay Chou has released a new album. But as the popularity of the new song rises, the rumors have begun to emerge endlessly. Some people say: New songs sound like the AI ​​who knows Jay Chou most.

AI? When is it so powerful? Xiaolu flipped through the book about AI and finally found the answer in this "How to Create a Trusted AI".

Can AI create music? The answer is actually no, because between ideal and reality, there is a large pit called "AI Hong ditch". In this regard, we take the driverless technology as an example.

Recommended reading

01

Trust AI with thought

The first pit, we call it "trust pit". On the surface, the behavior of machines is often similar to human behavior, so we will think that the machine also has the same ideological mechanism as humans, and it is not the case.

In 2016, a Tesla owner completely entrusted his life to the automatic cruise system. It is said that he watched the "Harry Potter" movie while letting the system carry him on the road.

After driving for tens of thousands of kilometers, the vehicle encountered a situation other than expected: a white cargo truck crossed the highway, Tesla drilled directly to the truck, and the owner died on the spot.

The accident survey showed that the vehicle seemed to send several alarms several times to the owner. He asked him to put his hands on the steering wheel, but the owner seemed to be absent -minded.

The truth behind the accident is very clear: something seems to have wisdom at a certain moment, it does not mean that it does, and it does not mean that it can deal with all the situation like humans.

Picture source: Jay Chou MV

02

Indications of AI can solve all the problems of all problems

The second pit, we call it "Unreal Progressive Pit". Mistakenly thought that AI solved the simple problem, which was equivalent to making progress on the problem. In fact, for a complex technology project, it takes 10%to complete the previous 90%work, and it takes 90%to complete the last 10%.

Computer security experts have shown that many non -autonomous cars we are driving today are increasingly controlled by software. Therefore, the connection between them (including Bluetooth, mobile phone network and the Internet) with wireless networks is easily attacked by hackers. Because future self -driving cars will be completely controlled by software, they are more likely to be hacked malicious attacks.

Another progress problem is "artificial interference". People will inevitably want to make some pranks for fully autonomous cars to explore their weaknesses. For example, walking back and forth in front of the car pretend to cross the road to prevent the car from moving forward.

Picture source: Jay Chou MV

03

Consciously fall into the "robust pit"

The third pit is our so -called "robust pit". As the director of the Human and Automation Laboratory of Duke University said, the problem is not how far a certain autonomous car can run without an accident, but how strong the car's adaptability is.

In order to allow the vehicle to drive reliably in all circumstances, the driver (probably AI) needs to know the motivation, goals, and even emotions of other drivers, people and animals who are riding bicycles, pedestrians and animals.

Take a look at a responsible situation and instantly judge who may cross the road, rush across the streets to catch up with the bus, suddenly turn to a signal light, or stop on the pedestrian horizontal track to answer calls and organize schoolbags ... This is most human drivers. The second day, but the autonomous car does not have these.

Autonomous driving is still unable to reach the ability to leave people, and art in art can also confuse audiovisuals without reaching humans. When human beings create music, the most important thing is their own understanding of music. The foundation of understanding is meaning, and significance is the real obstacle of artificial intelligence, so AI actually has no "understanding" at all. Their only ability is "seeing a lot of knowledge".

Picture source: Jay Chou MV

04

AI cannot replace humans

If you have to say it, there is only one kind of "AI Divine Comedy". Its idea is to find a relatively fixed template through data analysis and learning, and then use the template to "create".

For example, through big data, collecting target users' preferences for music, the characteristics of often listening to songs, and curls to evaluate, and finally generate the user's favorite music. Not surprisingly, the user likes it very much, because each notes are carefully made according to the preference of people.

However, such works are only cold data. Without human hot emotional investment, they can't escape the change of music development. In the end, they can only be submerged in the torrent of the times and cannot become classics.

It is undeniable that the structure of the structure of music may be better in AI, but the emotional revealing and the "information" in music are realized by human subconscious. The huge operation function of the subconscious AI is completely incomparable. In this regard, when the human brain creates truly good music, the brain computing is much higher than the computer.

Therefore, from the perspective of the development of AI, the music obtained by the AI ​​algorithm can only be a tool, and cannot replace human creation.

At present, the successful AI music belongs to the narrow pop music, the same online song, and the real pop music cannot be done at all. I believe that the real composer will not use this tool but every notes, but every notes, but every notes, but every notes, but every notes, but every notes, but every notes, each notes, but every notes, but every notes, but every notes, each notes. Will be grasped in person.

Therefore, Zhou Dong's talented creative singers will inevitably be dismissive of this.Of course, if you want to understand whether AI can create music, what risks exist in AI, is AI really trustworthy? Then you must see this "How to Create a Trusty AI", this book outlines the futureThe best roadmap for artificial intelligence has made a clear and objective assessment of the current status of artificial intelligence.

In the era when it must be called AI, the book "How to Create the Trusted AI" is undoubtedly a sober agent.Forgot to learn deep learning, return to common sense reasoning, a more exciting AI road is in -depth understanding.

- END -