The Documentary of artificial intelligence
AI is not a brand new thing, it has been developed for decades! Below we introduce the three most representative development stages.
The picture above shows some milestone events within the field of AI from 1950 to 2017. In summary, it'll be divided into 3 major stages:
- The first wave (non-intelligent dialogue robot)
- 1950s to 1960s
- In October 1950, Turing proposed the concept of artificial intelligence (AI) and also proposed the Turing test to test AI.
Within a few years after the Turing test was put forward, people saw the "dawn" of computers passing the Turing test.
In 1966, the psychotherapy robot ELIZA was born
People in that era spoke highly of him, and some patients even liked chatting with robots. But his realization logic is very simple, it is a limited dialogue library, when the patient utters a certain keyword, the robot will reply to the specific word.
The first wave did not use any brand-new technology, but used some techniques to make the computer look like a real person, and the computer itself is not intelligent.
The second wave (voice recognition).
[1980s to 1990s]
In the second wave, voice recognitions is the most representative breakthroughs. The core reason for the breakthrough is to abandon the idea of the semiotic school and change to a statistical idea to solve practical problems.
The biggest breakthrough of the second wave is to change the thinking, abandon the thinking of the semiotic school, and instead use statistical thinking to solve the problem.
The third wave (deep learning + big data)
Early 21st century
Early006 was a watershed in the history of deep learning. Jeffrey Hinton published "A Fast Learning Algorithm for Deep Belief Networks" this year, and other important deep learning academic articles were also published this year, and several major breakthroughs were made in the basic theoretical level.
The main reason why the third wave will come is that two conditions have matured:
- After 2000, the Internet industry developed rapidly and formed massive amounts of data. At the same time, the cost of data storage has dropped rapidly. Makes the storage and analysis of massive data possible.
- The continuous maturity of GPUs provides necessary computing power support, improves the availability of algorithms, and reduces the cost of computing power.
Deep learning is the current mainstream technology
After various conditions have matured, deep learning has exerted its powerful capabilities. Continuously set new records in the fields of speech recognition, image recognition, NLP and so on. Let AI products truly reach the stage of usability (for example, the error rate of speech recognition is only 6%, the accuracy rate of face recognition exceeds humans, and BERT exceeds humans in 11 performance...).
The third wave came, mainly because the conditions for big data and computing power are available, so that deep learning can exert great power, and the performance of AI has surpassed human beings and can reach the "usable" stage, not just scientific research.
The difference between the 3 waves of artificial intelligence
The first two crazes were dominated by academic research, and the third craze was dominated by real business needs.
Most of the first two crazes were at the level of market propaganda, while the third craze was at the level of business models.
The first two upsurges were mostly about persuading the government and investors to invest in the academic community. The third upsurge was mostly about investors actively investing money in academic projects and entrepreneurial projects in hot areas.
Questions were raised when the first two crazes were more frequent, and problems were solved when the third craze was more frequent.
What can't artificial intelligence do?
3 levels of artificial intelligence.When exploring the boundaries of AI, we can first simply and crudely divide AI into 3 levels:
- Weak artificial intelligence
- Strong artificial intelligence
- Super artificial intelligence
3 levels of artificial intelligence: weak artificial intelligence, strong artificial intelligence, and super artificial intelligence
Weak artificial intelligence
Weak artificial intelligence is also called restricted field artificial intelligence (Narrow AI) or applied artificial intelligence (Applied AI), which refers to artificial intelligence that focuses on and can only solve problems in a specific field.
For example: AlphaGo, Siri, FaceID...
Strong artificial intelligence
Also known as Artificial General Intelligence (Artificial General Intelligence) or Full AI (Full AI), it refers to artificial intelligence that can do all the tasks of humans.
Strong artificial intelligence has the following capabilities:
Ability to reason, use strategies, solve problems, and make decisions when there are uncertainties
The ability to express knowledge, including the ability to express common sense knowledge
- Planning ability
- Learning ability
- Ability of communication using natural language
- Ability to integrate the above capabilities to achieve the goals set.
Super artificial intelligence
Assuming that computer programs, through continuous development, can be smarter than the world's smartest and most gifted humans, then the resulting artificial intelligence system can be called super artificial intelligence.
We are currently at the stage of weak artificial intelligence. Strong artificial intelligence has not yet been realized (even far away), and super artificial intelligence is even invisible. Therefore, the "specified field" is still an insurmountable boundary for AI.
What are the capabilities of artificial intelligence?
Do all math problems in the world have clear answers?
If there is a clear answer, can the answer be calculated in a limited number of steps?
For those mathematical problems that may be calculated in finite steps, can there be an illusory machine that allows him to keep moving, and finally when the machine stops, the mathematical problem will be solved?
Turing really designed a set of methods, and later generations called it a Turing machine. All computers today, including new computers being designed around the world, have not exceeded the scope of Turing machines in terms of their ability to solve problems.
(Everyone is from the earth, why is the gap so big??)
Through the above three questions, Turing has drawn a boundary. This boundary is not only applicable to today's AI, but also to future AI .
Capability boundaries of artificial intelligence
There are many problems in the world, only a small part of them are mathematical problems. In math problems, only a small part is solvable. Among the solvable problems, only part of it can be solved by an ideal Turing machine.In the latter part (the part that the Turing machine can solve), there is only part that can be solved by today's computersThe problems that AI can solve are only part of the problems that computers can solve.Worried that artificial intelligence is too powerful? You think too much! Insome specific scenarios, AI can perform well, but in most scenarios, AI is useless.
0 Comments