Chapter 798 Turing Test
Hearing this, the big bosses who were originally calm finally showed solemnity on their faces.
Those who can sit here are smart people.
They can all hear the authenticity of big data applications from Xu Liang’s words.
“However, big data and cloud computing are only the foundation, and what really brings about changes in the industry is artificial intelligence.
I believe many people have heard of the ‘Turing Test’.
Let a machine and a person sit behind the scenes, and let a referee communicate with the person behind the scenes and the machine at the same time. If the referee cannot judge whether the person he is communicating with is a human or a machine.
This means that this machine has the same intelligence as humans.
This is the entire content of the famous ‘Turing Test’.
Computer scientists believe that if a computer achieves five things, it can be considered to have the kind of intelligence Turing said.
First, speech recognition.
Second, machine translation.
Third, automatic summarization or writing of text.
Fourth, the chess champion who defeated humans.
Fifth, answer questions automatically.
Regarding how to achieve these five things, the academic community is divided into traditional artificial intelligence methods and other modern methods.
So what are the traditional artificial intelligence methods?
To put it simply, it is to first understand how humans generate intelligence, and then let the computer do what humans think.
This method is also called the ‘bird flying method’.
Just like humans observed the flight of birds and invented airplanes.
Observe the parade of fish invented like a submarine.
Invent through simulation.
But later years of research proved that this method is very unrealistic.
Because a machine is always a machine and can never think like a human.
Scientists had to find another way.
By the 1970s, everyone began to try another development path for machine intelligence.
That is, using data-driven and supercomputing methods to realize artificial intelligence.
This method is also called machine learning or knowledge discovery, which is the modern artificial intelligence development method we mentioned before.
The first person to make achievements in this area was Fred Jalnick, a professor at Cornell University in the United States in 1972.
He is not an artificial intelligence expert, he is an excellent communications expert.
He believes that the human brain is a source of information. From thinking to finding the right sentence and then speaking it through pronunciation, it is an encoding process.
It is a problem of information dissemination through a long channel through the medium (sound channel, air, etc.) to the listener's ears.
Finally, the listener understands it, which is a decoding process.
In other words, he believes that artificial intelligence speech recognition is a typical communication problem.
It can be solved by solving communication problems.
To this end, Jarinick used two mathematical models, namely Markov models, to describe the information source and channel respectively.
After finding the mathematical model, the next step is to use statistical methods to 'train' the parameters of the model, which today is machine learning.
Through this method, the speech recognition rate of artificial intelligence has increased from about 70% in the past to 90%.
At the same time, the scale of speech recognition has increased from a few hundred words to more than 20,000 words, which can be called a revolutionary development.
The most important thing is that Jarinick’s research came to a conclusion.
That is:
As the amount of data continues to increase, the system will become better and better.
Therefore, international artificial intelligence research is divided into two groups.
One group is the flying bird group that imitates humans, and the other group is the data-driven group.
The reason why the latter has not developed rapidly is mainly because it is very difficult to obtain data.
First, there was no machine-readable data at the time.
Secondly, different versions of many literary pearls are scattered in different countries, and their translations often do not correspond one-to-one.
Of course there are many other reasons that I won’t go into detail one by one.
However, this difficulty has changed in the Internet era.
Its emergence allows research institutions to easily obtain global machine-readable data.
Moreover, with the development of the Internet, the amount of data is increasing several times or even more than ten times every year.
With the support of huge data, the error rate of speech recognition was reduced by half in the ten years from 1994 to 2004.
And the accuracy of machine translation has doubled.
20% of the contribution comes from method improvements, and 80% comes from the increase in data volume.
Then there is the Global Machine Translation System Competition held in the United States in February this year.
Hongmeng and Google achieved a BLEU score of more than 50% through data-driven methods.
5% ahead of top research institutions such as the famous University of Southern California and IBM Watson Labs, which have been studying machine translation for decades.
In the past, it took 5 to 10 years to improve these five percentage points.
In Chinese to English translation, Hongmeng’s score is 17% ahead of the third place, and Google, which also uses data-driven methods, is 15% ahead of the second place. This gap has exceeded the gap of a generation. level.
Hongmeng and Google are both new companies that have been established for less than ten years.
The foundation in artificial intelligence research and development is definitely not as deep as that of Southern California and Watson Labs.
But we surpassed them.
Is the reason why we are better than them?
No.
So how did the gap arise?
Very simple.
As the two largest search companies in the world, Hongmeng Bing and Google both have huge search databases. And we are digitizing all pictures, books, newspapers and periodicals around the world every year.
This gives us the largest database in the world.
Although the University of Southern California and IBM Watson Labs have more talents than us, their research foundations are deeper than ours.
But it is far inferior to Bing and Google in terms of data volume.
So, they fell behind.
The results of this competition have had a huge impact in the field of artificial intelligence.
Judging from the news we have received, most scientific research institutions around the world have abandoned the "bird flying school" method of transmission and switched to data-driven methods.
In other words, 2005 will become a watershed in the field of global artificial intelligence.
Starting from this year, Niao Fei Pai will be completely abandoned, and data-driven will become the only mainstream.
I believe that as the amount of data continues to accumulate, artificial intelligence will become more and more ‘intelligent’ and ‘practical’.
It will have a profound impact on all aspects of society. ”
Xu Liang, who has completely entered his own rhythm, no longer needs manuscripts.
At this moment, he completely let go of the identities of both parties.
Completely treating the people in the audience as the audience.
And they were completely attracted by the content of Xu Liang’s words.
“The future of agriculture will completely get rid of the agricultural model that China has used for thousands of years, which has consumed a lot of manpower and material resources and intensive farming.
It will be replaced by intelligent agricultural factories.
In this factory, a large number of radio frequency chips are installed to collect all data such as temperature, humidity, soil fertility, etc., and integrate them into the artificial intelligence brain
Then the ‘intelligent brain’ injects water and fertilizer according to the needs of the crops through drip irrigation.
Use 10% or even less water and fertilizer to grow double or even more agricultural output.
In the past, we might need twenty farmers to cultivate one hundred acres of land.
In the era of intelligent agriculture, only one person is responsible for managing and maintaining the 'artificial intelligence brain', which can manage thousands or even thousands of acres of agricultural land.
Efficiency and output are increased hundreds of times.
If we can build more nuclear power plants, solar power, wind power and hydropower in the future, we can bring down the price of energy.
Then we can make agriculture develop three-dimensionally.
Really get rid of the restrictions of the natural environment on agriculture. "
Xu Liangshun mentioned 'three-dimensional agriculture'. Before he was reborn, China established a 'three-dimensional agriculture factory' in the western region with low energy prices due to the skyrocketing solar power generation.
However, because even if energy prices fall, the investment is still relatively large
Therefore, it can only be used to grow high-value cash crops.
There is no basis for large-scale promotion.
So he was not prepared to say more.
“The future industry will use intelligence and big data systems to help workers, or even replace workers, to achieve comprehensive intelligence in the manufacturing industry.
Unmanned factories, unmanned assembly plants, will More and more.
The price of industrial products will drop several times.
Nowadays a mobile phone costs thousands of dollars.
In the future, mobile phones will not only have richer functions and more advanced performance, but will not even require you to spend money. China Unicom and China Mobile will give them to you, because the income from phone calls and Internet fees far exceeds the value of a mobile phone. ”
Looking at the suspicious looks in the audience, Xu Liang didn’t explain much.
Time will tell everything.
“When big data and artificial intelligence enter all aspects of industrial manufacturing and sales, not only will the number of workers gradually decrease, but the entire manufacturing industry will be reshuffled.
The low level of companies that rely solely on reducing workers Competition will no longer have advantages in manufacturing.
Competition in the future will be at the level of intelligence in the entire process from design to sales.
In other words, China will be. The last country that has and can develop its demographic dividend
in ten or twenty years.
A large population will no longer be an advantage. "
After finishing speaking categorically, Xu Liang continued.
"Intelligent medical treatment in the future.
No matter in any country, the biggest bottlenecks encountered in medical care are mainly reflected in several aspects.
First, the cost of medical care is getting higher and higher.
Nowadays, if you go to the hospital, just a physical examination will cost hundreds or even thousands of yuan;
If you go to the hospital and go through a series of procedures such as blood tests, urine tests, and MRI, it will cost thousands or even tens of thousands of yuan.
For ordinary people, this is a very large expenditure.
So the situation of being unable to afford medical treatment will become more and more serious.
Second, medical resources are imbalanced.
The medical resources in first-tier cities far exceed those in third- and fourth-tier cities, and even less so in ordinary counties.
Until now, there are no tertiary hospitals in more than a thousand cities and counties across the country.
Finally, and most importantly, many diseases cannot be cured.
Such as cancer, Parkinson's disease and Alzheimer's disease.
Despite the efforts of doctors and scientists around the world for many years, and the investment of large amounts of money by countries and R&D institutions around the world, the treatment of diseases such as cancer has been making slow progress over the past few years.
But we can use big data and artificial intelligence to solve the above problems.
(End of this chapter)