bzdww

Get answers and suggestions for various questions from here

2140, AI will completely replace humans

cms

AI will completely replace humans in 2140

A number of experts from the Oxford Institute for Human Futures, Yale University and AI Impacts conducted a large-scale survey of 352 machine learning researchers to predict the development of artificial intelligence in the coming decades. According to the research team, the study is a representative study with a large sample in the field of artificial intelligence. It surveyed 1,634 experts in the field of machine learning and artificial intelligence (these experts are in the top two conferences in the field of machine learning: The 2015 International Machine Learning Conference and the 2015 NeuroInformation Processing System Conference published papers and received 352 responses.

The results show that respondents expect AI to overtake humans in multiple areas over the next 40 years, such as language translation (2024), truck driving (2027), retail (2031), bestseller writing (2049) ), surgeon work (2053).

This group of people who know the most in the world and are closest to AI believe that there is a 50% chance that AI will overtake humanity in all fields in 45 years and automate all human work in 122 years.

According to the survey, when the machine can complete each task independently and complete better than humans and at a lower cost, the so-called "HLMI" (Advanced Machine Intelligence, ie, strong artificial intelligence) level is achieved. And "full automation of labor" refers to the complete automation of all occupations, that is, for any occupation, the machine can be better done at a lower cost than human employees. .

The median forecast shows that there is a 50% chance that full automation of labor will be achieved within 122 years. That is, in 2140, human labor was fully automated, and AI completely replaced humans in all jobs.

Based on the data systems of Cambridge University researchers Michael Osborne and Carl Frey, the BBC analyzed the “probability of elimination” of 365 occupations in the future. The possibility of being replaced by AI is very small. There are artists (3.8%) and musicians (4.5). %), scientists (6.2%), lawyers & judges (3.5%), architects (1.8%), public relations (1.4%), psychologists (0.7%).

We will find that these professions need to have these abilities: 1. Social skills, negotiating skills, and the art of human practice; 2. Compassion, and sincere support and concern for others; 3. Creativity and aesthetics.

To analyze, for these complex emotional experiences, cooperation systems, groundbreaking scientific research work, how AI may be completed step by step through deep learning.

The first step is to increase the processing speed of the computer and let the AI ​​reach the computing power of the human brain. Now the fastest supercomputer, China's Tianhe No. 2, has actually surpassed the computing power of the human brain, so there is no hardware problem. According to the prediction of Moore's Law, the popularity of computer hardware supporting strong artificial intelligence is only a matter of time, and this time has been calculated to be within ten years.

The second step is to make the computer smart. This step is more troublesome, but there are already directions, such as by copying the human brain, imitating biological evolution, and computer self-learning.

Now that there is a solution to increase the speed of computer processing, let's talk about how to make the computer smart.

This mysterious power is to give AI the cornerstone of "soul" - Bayes' theorem.


The cornerstone of Bayes' theorem AI evolution

The biggest problem in creating strong artificial intelligence is how to make AI have human thinking and consciousness.

In the past few years, machine intelligence has approached the philosophical proposition of "I think so I am". The Bayesian formula provides the soul for AI because of its analysis of the power of "conditional probability".

Bayes' theorem is the probability theory proposed by the 18th century British mathematician Thomas Bayes. In addition to Bayes, in 1774, the French mathematician Laplace gave the expression of the Bayesian formula we are using now:

This formula indicates the conditional probability of the occurrence of an A event under the condition that the B event occurs, which is equal to the probability of the occurrence of the B event multiplied by the event of the A event, and the probability of occurrence of the B event. In the formula, P(A) is also called the prior probability, and P(A/B) is called the posterior probability.

Faced with the problem of natural speech processing, Bayesian formula can be easily solved. Since the speech recognition has become a billion-dollar market capitalization company, the first thing to thank is the Bayesian formula and the Markov chain. Speech involves a variety of dynamic grammars, but when you see the accuracy of machine translation on the spot, you will be sighed that this is simply a "sign", much stronger than most live translations. Once the conditional probability occurs, Bayes always comes forward.

With the iterative transformation of a large number of data input models, with the continuous improvement of computing power and the development of big data technology, the huge practical value of Bayesian formula is more and more manifested, and it has been recognized by the mainstream scientific community and has begun to spread in all fields. Speech recognition is just an example of the application of Bayesian formula. In fact, the idea of ​​Bayes' theorem has penetrated into all aspects of AI.

The establishment of a single speech model allows us to see the Bayesian theorem's ability to solve problems, but the expansion of the Bayesian network allows us to vaguely feel the horror of "Skynet" behind AI. The Bayesian formula helps scientists achieve mathematical-level operations in a rigorous mathematical form. The greater the flaws in human cognition, the more powerful the Bayesian network is.

Using Bayes' theorem as the first lesson of AI artificial intelligence, AI can inject subjective genes into itself. Today, a vigorous "Bayesian revolution" is taking place in the AI ​​world: Bayesian formula has penetrated into the bones of engineers, and classification algorithms have become mainstream algorithms. In many people's eyes, Bayes' theorem is the cornerstone of AI evolution.

We can't predict the true power of Bayesian formula combined with computer, but we can be sure that through the development of technology, Bayes' theorem can make AI infinitely close, or finally achieve true human intelligence.


The power of silicon-based life

Silicon-based life has obvious advantages over carbon-based life. In most occupations, machines can be better done at a lower cost than human employees. Strong artificial intelligence has too many advantages over the human brain.

First of all, from the hardware:

1. The computer's computing speed and information transmission speed are several orders of magnitude faster than brain neurons. Brain neurons operate at speeds of up to 200 Hz, and today's microprocessors can run at 2G Hz, or 10 million times faster than neurons. The internal information transmission speed of the brain is 120 meters per second, and the speed of information transmission of the computer is the speed of light.

2. The computer's capacity and storage space is much larger than the human brain. The size of the human brain is basically fixed and cannot be changed day by day. Even if it is really made very large, the information transmission speed of 120 meters per second will become a huge bottleneck. The physical size of the computer can be very random, so that the computer can use more hardware, larger memory, long-term effective storage media, not only capacity but also more accurate than the human brain.

3. Reliability and durability. The storage of the computer is not only more accurate, but the transistor is more accurate than the neuron, and it is less likely to shrink (it is really bad and well repaired). The human brain is prone to fatigue, but the computer can operate at peak speeds 24 hours a day.

Software:

1. The computer is editable, upgradeable, and more possibilities. Unlike the human brain, computer software can be upgraded and corrected, and it is easy to test. The upgrade of the computer can strengthen the weaker areas of the human brain. Not only can the enemy elements be visual elements, but also the engineering components can be strengthened and optimized.

2. The collective ability of computers enables artificial intelligence to form a powerful network. Although humans have crushed all species on collective intelligence, the collective intelligence of computers is more powerful. An artificial intelligence network running a specific program can often self-synchronize globally, so that what a computer learns is immediately learned by all other computers. And computer clusters can perform the same task together, because dissident, motivation, self-interest, and human-specific things may not appear on the computer.

A silicon-based life without physical strength, no aging, no fear of disease, high IQ, and complete rationality, how to think and eat Lhasa more than every day, often produces useless emotions, and occasionally humans who strike physically are more reliable. It is absolutely purer and simpler to ensure that the logic and code are not in error, and that it is more insightful than grasping the complex consciousness and humanity.


Is weak artificial intelligence coming to strong artificial intelligence?

According to the strength of artificial intelligence, they can be divided into three categories: weak artificial intelligence (excellent in a single aspect of artificial intelligence), strong artificial intelligence (in every respect can be compared with humans) and super artificial intelligence (in almost all areas) More intelligent than the most intelligent human brain).

At this stage, scientists are defined as the era of weak artificial intelligence. In fact, weak artificial intelligence has become ubiquitous, such as: map software navigation in mobile phones, query weather, Siri; "guess you like" "friend recommendation"; the world's strongest checkers, chess, spell chess, backgammon and Black and white chess players; a wide range of complex weak artificial intelligence is widely used in military, manufacturing, and financial fields.

In 2000, Goldman Sachs hired 600 traders at the US cash stock exchange counter at its New York headquarters. In 2018, there were only two traders left in the trading floor. Replacing costly and inefficient labor with higher technology is a major trend in all industries.

The White House’s “Artificial Intelligence, Automation, and Economy” report in 2016 shows that between 9% and 47% of US jobs will be threatened in the next decade or two. In other words, about 6% of employment opportunities will disappear every three months. The latest research report from PricewaterhouseCoopers (PwC) also shows that nearly 40% of jobs in the US in the next 15 years may be replaced by artificial intelligence and robots.

Other advanced economies cannot be avoided. The report also shows that nearly 30% of jobs in the UK will be replaced by artificial intelligence and robots in the next 15 years. In Germany, 35% of jobs will be replaced by artificial intelligence and robots, and 21% in Japan. Jobs are replaced by artificial intelligence and robots.

Computer scientist Donald Knuth said: "Artificial intelligence has surpassed humans in almost all areas that need to be considered."

Human beings are in the era of weak artificial intelligence, so strong artificial intelligence will be far behind?

At least from the performance of Google's assistants in the past few days, some of the strong artificial intelligence through the Turing test has been born. Humans already have a major breakthrough in artificial intelligence.

Natural evolution has taken hundreds of millions of years to develop the biological brain. When humans create the first artificial intelligence, they are naturally evolved. Of course, this may also be part of the natural evolution - the real model that may evolve is to create all kinds of intelligence, until one day there is an intelligence that can create super intelligence, and this node is like stepping on the mine's squall line. The same, it will cause a global explosion, which will change the fate of all living things.


Love with AI is the only way for mankind

John Hennessy, the chairman of Alphabet, who won the annual Turing Award, the 65-year-old master, excitedly expressed on the stage of the Google conference: "People are investing in the field of machine learning and artificial intelligence for 50 years. Finally, we have made a breakthrough. In order to achieve this breakthrough, the basic computing power we need is 1 million times as envisioned. But in the end we still did it. This is a revolution that will change the world."

Super civilization is bound to emerge, and social operations must be more transparent and efficient. Humans will not stop to study the footsteps of the AI ​​field, but the silicon-based civilization formed by AI is more likely to become a competitor of future human civilization.

What is certain is that AI will definitely be applied more and more extensively in human society. Can this be seen as a gradual invasion of silicon-based life into a carbon-based life society? Humans can no longer survive without relying on AI. Just leaving the smartphone for a while can make a normal modern person anxious.

Humans can't do without AI. But I don't know how to eliminate the potential rejection and threat of alien things in my bones.

AI is actually not moral and immoral. The motivation of AI depends on what the human target is for artificial intelligence. In order to achieve the goal, it is possible to make decisions and actions that destroy humanity. It is just to complete the task that has been set since its birth. "I am not malicious to you, even for you, I am going to ruin you." The maliciousness of self-awareness is more scalp than deliberate malice.

How can humans control the AI ​​they created themselves? The unethical nature of super artificial intelligence is the danger of artificial intelligence. Even if there is no loophole in the code, there are three laws of Asimov's paradox, which cannot protect human beings perfectly. From the actions of AI in "Robots and Empires", "The Matrix" and "2001 Space Roaming", we can see its foretaste and hidden worry.

There is bound to be an invisible war between the carbon-based civilization and the silicon-based civilization.


The final outcome of the carbon-based civilization actively embraces AI

In the face of this mighty invasion of silicon-based life, mankind has been unable to remain indifferent and maintain an optimistic or neutral attitude.

There is actually another way. For humans, the most rational choice is to learn from the silicon-based civilization and actively choose to embrace AI.

When the future computing power is no longer a scarce item, the agility and sensibility of carbon-based life, combined with the powerful and rational life of silicon-based life, may allow human civilization to last for a long time in the face of future threats and challenges.

The carbon-based civilization data has bottomed out, and the silicon-based civilization is new.

In the future possible world of silicon-based life, a monument that represents the carbon-based civilization is engraved with such a timeline:

In 2018, AI passed the Turing test.

In 2063, strong artificial intelligence was reached.

In 2140, AI completely replaced humans.

This may be the final outcome of the carbon-based civilization.


Author: Dark dawn
link: 2140, AI will completely replace human
Source: know almost
quantum School reprinted with permission, copyright reserved by the authors.