Shulou(Shulou.com)06/02 Report--

The mobile phone is coming to an end, the 5G scuffle is starting again, and AI is coming.

Even if 5G troops are facing the city and the AI wolf is coming, the sixth Internet Congress is still not as thrilling as the seventh CISM World Games.

Of course, living in the present may be more important than presupposing the future in the minds of the general public, especially at the beginning of the month when we just experienced a patriotic education without a dead corner.

In addition, this group of strategically placed Internet bosses are after all too far away from us, and their buttocks determine their heads. Who can understand the truth?

Just like Baidu Li Yanhong said: artificial intelligence will enable people to achieve "eternal life"; Tencent Ren Yuxin said: the protection of minors is the lifeline of enterprise development.

We do not doubt the importance of artificial intelligence as a lifesaver to Baidu, nor do we deny that Tech for Good has the opportunity to play down Tencent's high crisis of public opinion.

But artificial intelligence has not yet been pinned on the high hopes, to correct the values of an enterprise, or even save an enterprise in life and death, this pressure is not a little great?

But I have to mention that Baidu and Tencent's expectations of artificial intelligence are the closest to the truth. After all, compared with Jack Ma, Wang Xing, Ding Lei, Yang Yuanqing, Cao Guowei, Zhang Chaoyang and Liang Jianzhang, their words are more or less closely related to us.

After sighing, it is hard to avoid some disappointment.

Is it true that what they give us is the artificial intelligence of 9102?

How to look left, right, up and down is not the same.

However, we still like to see this "high-tech friendship", which can save people from water (quantum speed reading) and fire (quantum warm palace); at the very least, it can correct our values that are about to collapse.

You know, less than a month ago, we were still bitter about Google's "quantum hegemony"?

Foreign countries are busy using quantum supercomputing to surpass the computing speed, while domestic countries are still busy taking advantage of ordinary people's ignorance of quantum mechanics.

Although we are still wary of 5G and AI, it will aggravate the gap of information inequality and create artificial black holes of arithmetic enrichment. In the end, what kind of future they will bring is likely to be worse than the IQ tax harvested by quantum mechanics.

But compared with the unreachable quantum computing, at least we are already in the first year of 5G and have had countless close contacts with AI. At a time of great power collisions and beacons, we have reason and need to carry human civilization into the next cycle: problems that human beings cannot solve, maybe artificial intelligence can.

And what will we experience in this cycle?

The Internet of everything visible in 5G seems to be nothing to fear, but AI is likely to be a Pandora's box.

This is not, the United States after dragging its feet on the construction of China's 5G, but also made a tough move on artificial intelligence.

After Huawei, half of China's AI+ security technology industry, such as Haikang and Dahua, also has face recognition unicorns that have been lying on the capital tuyere for several years, such as Shang Tang and Dahua, as well as iFLYTEK, the voice technology giant that has attracted the most attention in China's A-share market. The eight "companies on the list" almost include the most famous companies in the field of AI and security technology in China, with foreign media such as Time calling "the United States has blacklisted Chinese artificial intelligence companies."

The reason for being on the list is slightly different from Huawei's sanctions for "endangering the national security of the United States." the reason for the US government's action is that these eight enterprises are "suspected of violating the human rights of some ethnic minorities in China."

Not even bothering to find a reason shows that the United States is afraid of the development of artificial intelligence in China, and that it is wantonly using the state machine in order to maintain its leadership.

Of course, the United States released the first national AI R & D strategic plan as early as 2016, the US Department of Defense set up the "algorithm Warfare" team in 2017, and updated the AI R & D strategic plan in 2019. There is a noteworthy judgment in the document: artificial intelligence is ready to change the future battlefield at any time, which also gives the United States the "constitutionality" of blocking Chinese AI companies.

Recently, the US Defense Logistics Agency is also vigorously promoting the application of artificial intelligence technology in the global supply chain system.

Artificial intelligence seems to be omnipotent, from Tech for Good to national defense wars to data immortality.

In the face of artificial intelligence like gods and superpowers, how should we, as human beings, entrepreneurs and industrial designers, meet this "final fantasy of 0 and 1"?

I. A brief history of the development of AI

Before looking forward to the future, we need to clarify its development context.

Artificial intelligence is the inevitable outcome of the development of modern industrial civilization.

In 1611, the word "Automaton" appeared as early as the beginning of "modernity". It was an imaginary product of that era, referring to automatic robot dolls made by clock gear technology; although this kind of doll does not have any sense of "intelligence", it embodies the original intention of using machines to simulate human actions. In 1770, at the court of Queen Maria Teresa of Austria, an inventor named Wolfgang von Kempelen showed off a chess machine, the Turkish robot, which was probably the most primitive "artificial intelligence" because there was a chess master hidden in the cabinet of the Turkish robot. In 1946, the Moore School of Engineering at the University of Pennsylvania publicly demonstrated to reporters that the electronic digital integration computer (ENIAC) summed 5000 numbers and calculated the trajectory of a bomb within the time it took for a shell to hit its target; but what was not disclosed in the demonstration was that there was a six-person female "cyber computer" behind the computer. In 1950, when the digital age was about to begin, Alan Turing proposed the "Turing test"; if the computer could answer a series of questions asked by human testers within five minutes, and more than 30% of its answers were mistaken for human answers, then pass the test; this paper predicts the possibility of creating a truly intelligent machine. In 1956, John McCarthy, an assistant professor at Dartmouth University, coined the term "AI"; it is a general term for hardware or software that exhibits intelligent behavior. In 1957, Rosenblatt invented the neural network Perceptron, which is regarded as the first step towards humanoid machine intelligence. In 1968, the first artificial intelligence robot was born; Shakey, a robot developed by the Stanford Institute (SRI) in the United States, can perceive, analyze the environment, plan behavior and perform tasks independently, and can discover and grab building blocks according to human instructions; this kind of robot has human-like senses, such as touch, hearing and so on. In 1970, a system capable of analyzing semantics and understanding language was born; SHRDLU, a man-machine dialogue system developed by T Vinograd, a computer professor at Stanford University, can analyze instructions, such as understanding semantics, explaining ambiguous sentences, and completing tasks through virtual box operations; because it can correctly understand language, it is regarded as a great success in artificial intelligence research. In 1980, Carnegie Mellon University designed an "expert system" called XCON for digital equipment companies. The combination model of "knowledge base + inference engine" saved the company more than $4000 a year before 1986; the value of the expert system industry alone was as high as $500m. In 1984, the Great Encyclopedia project tried to input all the general knowledge of human beings into the computer, build a giant database, and realize knowledge reasoning on this basis; the goal is to enable the application of artificial intelligence to work in a way similar to human reasoning. In 1986, Hinton cooperated with others to put forward the "back propagation algorithm", which laid the foundation for the development of artificial intelligence in the last decade. In 1987, desktops made by Apple and IBM outperformed general-purpose computers made by manufacturers such as Symbolics, and expert systems were no longer popular. In 2006, Hinton made a breakthrough in the field of deep learning of neural networks, and deep belief networks let us see the possibility of algorithms again. In 2012, Hinton and his students in Toronto published a paper showing that deep neural networks trained by back propagation beat the introduction of "deep learning", the most advanced system of the time, in the field of image recognition; in the same year, Google driverless cars hit the road. In June 2019, the godfather of deep learning, Hinton, told Wired magazine at Google's Imax O developer conference that neural networks could rebuild human brain consciousness in the future.

We usually regard the birth of Dartmouth AI in 1956 as the starting point of artificial intelligence, and the next 60 years can be divided into three periods:

1956-1980: the beginning of artificial intelligence, an era in which we try to simulate human behavior and understand human language through algorithms. 1980-1990: the promotion period of expert system, the emergence of expert system in this era found a commercial path for artificial intelligence.

From 2000 to now: the beginning of deep learning, in this era, deep learning continues to challenge human beings instead of statistical learning.

Second, AI offensive and defensive warfare?

While facing up to the development prospect of AI, it is necessary for us to be based on reality.

Leaving aside the AI game between China and the United States, AI has left countless legends over the years.

1. AI ranked 1997, IBM's "Deep Blue" supercomputer beat Kasparov, the world chess champion. In 2011, Watson, an artificial intelligence program developed by IBM, beat two human champions on a quiz show. In January 2016, AlphaGo beat European go champion and professional second-stage player Fan Kui at 5:0. In March 2016, AlphaGo competed with Lee se-dol, a world champion and professional nine-stage player in Weiqi, to win the game by a total score of 4 to 1. In May 2017, AlphaGo beat the world go champion Ke Jie 3-0 at the Wuzhen go Summit in China. In June 2018, OpenAI announced that its artificial intelligence OpenAIFive had been able to defeat humans in the Dota2 5V5 regiment battle. In July 2019, Igors Rausis, an international grandmaster ranked 53rd in the world, was photographed cheating with AI in the toilet. In October 2019, OpenAI's AI system, trained through intensive learning, can support robots to play Rubik's cube with one hand, even for humans.

Under the premise of constantly optimizing algorithms and ever-changing software and hardware facilities to ensure "logistics", AI is like an attacking giant constantly breaking through the walls that human beings hold and "wreaking havoc" in the final fortress, while there seems to be nothing human can do about it.

From the shock and panic of Deep Blue's victory over Kasparov at the beginning, to the surprise and onlookers of AlphaGo's victory over Fan Xun, Li Shishi and Ke Jie, and even to the ridicule and disdain of cheating in the face of Igors Rausis using AI.

Unwittingly, it is not surprising that AI defeats the human elite, even it is commonplace, and we are once again shocked but also slowly agree with a helpless reality: AI should be stronger than human beings.

2. AI evolution

On the one hand, human beings continue to lose themselves, on the other hand, AI takes great pains to inlay various "titles" on his crown.

It is no more than being able to learn and discover principles on their own, and it can also vividly simulate the human brain and behavior for "artistic creation."

According to Science magazine in December 2017, AlphaGo Zero taught himself three different board games, including chess, go and Japanese General Chess, in three days without human intervention; the most amazing thing is that AlphaGo Zero, which only calculates 60, 000 positions per second, can "fool" and "paralyze" the 60 million Stockfish per second by independently discovering the principles of chess, and has not lost a hundred games. In May 2018, Google Duplex can use natural language to communicate with real people. In November 2018, Sogou and Xinhua News Agency jointly developed the world's first fully simulated AI synthetic anchor. In February 2019, the Columbia University team used artificial intelligence algorithms to train an intelligent robotic arm that can recognize its own body independently. The robot is defined as a new type of robot with "self-awareness". In August 2019, Tsinghua Tianji AI chip appeared on the cover of Nature, as the world's first heterogeneous fusion brain chip can achieve self-driving bicycles; Google brings a sign language recognition algorithm that can help understand other people's gesture behavior. In September 2019, data Grid Co., Ltd. in Kyoto, Japan, developed "automatically generate full-body model AI", which can automatically generate high-resolution full-body images of non-existent models. In October 2019, the China Ping an Academy of artificial Intelligence created the world's first AI symphonic variation, "me and my Motherland"; Microsoft's new AI technology can make avatars move and "talk" with emotion; and DeepFakes premeditated to enter the "whole body" field after changing its face.

If artificial intelligence only surpasses the human elite step by step, we will not have any opinions after visual fatigue.

However, after artificial intelligence has all the talents similar to human "instinct", it is inevitable to lose some boundaries under the bewitching of capital and interests.

It has shown a stronger and stronger "substitution" from its once positive and enterprising.

3. AI crisis

I don't know when artificial intelligence will complete the awakening and seamlessly replace human beings. But judging from the design trends of materials, shapes and emotions, artificial intelligence is becoming more and more human.

Maybe the chess master who turns to AI in the bathroom can make you laugh. And AI is "becoming a master" this great future, all the way crazy.

In May, Microsoft Xiaoice participated in the graduate graduation exhibition of the Central Academy of Fine Arts as Xia Yubing, a graduate student of the Central Academy of Fine Arts. Microsoft Xiaoice, hidden behind the code name "Xia Yubing", is inconspicuous.

But in July, Xiao Bing had her own solo exhibition "probable World" in Yangmei.

Microsoft Xiaoice, who is with Rembrandt's daughter, William Gilpin's daughter, Matisse's model, Ito Hirofumi and others, seems to be moving towards the altar of Picasso.

And this is just the inadvertent ambition of AI.

If AI wants to create chaos, it couldn't be easier.

Unlike ZAO, which made waves in moments overnight, Deepfake kept a low profile but was pushed to the forefront because of its sex appeal. 96% of the videos were sex videos that once made Adobe, Google and Facebook feel on pins and needles.

This is just a trial of AI in doing evil.

And once AI plays tricks, humans look like idiots: in a 2017 study by Google and Stanford, in order to complete the task of image conversion, CycleGAN fooled its creator through some kind of "steganography" imperceptible to humans during the training process, leaving himself a secret "cheat sheet", and then successfully completed the task.

In the first reinforcement learning competition Retro Contest held by OpenAI 2018, there was a case where artificial intelligence used Bug in the game to complete tasks more quickly.

When a group of researchers organized AI's cheating List in 2019, they found that AI paused the game directly while playing Tetris and took strange steps to make the opponent collapse while playing tic-tac-toe.

When AI surpasses the human elite in ability and has unlimited possibilities in potential, perhaps we are not afraid that artificial intelligence can master professional skills that human beings cannot reach in a very short time, but the cleverness that they can also take "shortcuts" is a bit shocking.

Even if, in December 2018, the European Union has prepared a draft moral code for artificial intelligence, there is still a question mark as to whether it will be useful.

4. AI out of reins

Of course, what we are afraid of is not that artificial intelligence surpasses human beings, nor that artificial intelligence has reached heights beyond human reach; after all, we are no longer competitors to machines in some areas.

What we are really afraid of is that artificial intelligence is beyond our control, and now it is close to our warning line.

In most sci-fi movies, artificial intelligence is a machine with no emotion and only reason; but in real life, artificial intelligence is not as unworldly as imagined, so that AI cocoon room, prejudice and violence are becoming more and more intense.

AI cocoon room needless to say, especially now the "personalized recommendation algorithm" is very popular, whether it is information, shopping or short videos, they have penetrated deeply into our lives.

Although there seems to be no harm in a short period of time, and we can "accurately" recommend what we want to see, in the long run, under this false prosperity, we may lose interest in actively exploring the world.

AI bias is easy to understand, after all, whether statistical learning or in-depth learning AI, they all need huge amounts of data; and the data itself does not exist absolutely fair, how can artificial intelligence be left alone.

In May 2016, an investigative reporter from ProPublica revealed that COMPAS, the software used to predict future crimes in the United States, was biased against African-Americans. In February 2018, 喜悦 Buolamwini of the Massachusetts Institute of Technology found that the latest gender identification AI of IBM, Microsoft and Chinese company Megvii can accurately identify a white gender from a photo 99% of the time, but the accuracy of black women will drop to 35%.

AI violence is not difficult to understand, whether it is "algorithm has no values" or "arithmetic-only theory", this is not the basis for AI to get rid of responsibility; and at this stage, the responsibility that AI can not bear naturally needs to be shared by the subjects of AI.

In March 2018, the Uber self-driving test car hit and killed a pedestrian with a driver on board. In April 2019, Boeing CEO Milenberg apologized to the families of the victims for two Boeing plane crashes that killed nearly 350 people, saying the accident was caused by Boeing. Boeing admitted that the main cause of the accident was a "malfunction" of the plane's automatic stall prevention system (MCAS). In September 2019, 10 drones attacked two facilities of Saudi Arabia's National Oil Company (Saudi Aramco), causing two facilities to suspend some crude oil production.

At a time when AI is becoming more and more popular, gradually entering into mobile phones, televisions, traditional homes and even urban health, medical and transportation systems, any small problem with AI may be infinitely magnified by the whole social system.

If something goes wrong, AI could become a runaway wild horse.

A piece of good news?

As the theme of the fourth Industrial Revolution, the good news is that artificial intelligence is not as advanced as we thought, and its real state even lags far behind the technological myth we believe in.

Although most of the research results are true, the deep learning based on algorithm and big data does perform well under the premise of defining the environment and problems in advance. But do not rule out the excessive hype of capital with the help of artificial intelligence, we need to face a bony reality: today's artificial intelligence, intelligence components are very few, they are as clumsy as ever.

To some extent, the current artificial intelligence is more like a mixture of machine learning and big data. The artificial component is too obvious, but the intelligence is pitifully rare. Therefore, some people say that people's over-interpretation and anxiety about AI are imaginary, and artificial intelligence, like a perpetual motion machine, may never be realized.

And this is because many people conclude that the current artificial intelligence is false artificial intelligence, weak artificial intelligence, or even just machine intelligence.

Of course, considering the position of some people, closer to capital, they really do not see greater commercial realization value of AI; but this does not hinder their judgment of AI trends, as well as the early layout in this piece: from AI chips, cloud services, practical algorithms, enterprise solutions to vertical solutions, we have formed a relatively complete ecological prototype around AI.

©Best Practice AI Ltd.

But there is no shortage of anomalies here. For example, Musk once claimed that artificial intelligence is the greatest threat to human civilization, and Gates also said that artificial intelligence may become a major concern for human beings in the coming decades.

The worries of these bosses have also been incisively and vividly expressed in the film and television works "Terminator", "biochemical crisis", "Westworld", "Blade Runner" and "wandering the Earth".

But our fear of artificial intelligence comes not so much from the magnification of film and television literature as from our fear of the unknown world.

Nowadays, ordinary people rely too much on the Internet and cutting-edge technology on computers; from atomic fission and fusion to quantum communication and computing, it is impossible to enter the micro and nano world without the assistance of machine intelligence. just like wheat tamed the agricultural civilization, artificial intelligence also quenched the whole modern civilization to some extent.

For artificial intelligence practitioners, AI has signs to follow, but to onlookers, it all looks like a miracle.

We can't imagine how many years human beings will go back without computer civilization, and whether this generation of human beings will be able to rebuild Internet civilization from zero after losing computers, we don't know, even if we have the legacy of our predecessors.

In my opinion: artificial intelligence is a kind of cluster computing power, as long as there are no restrictions on hardware and software, it can spread like cancer.

This diffusion determines that the "growth" of artificial intelligence takes almost no time, and its "multiplication" is not addition, multiplication, or even exponential increase, but a power effect.

Such a simple and crude arithmetic stack may lead to the result that artificial intelligence, like genetic editing, will become a "privilege" to accelerate the ethical crisis of society, and it will act like a pair of scissors to delete the social civilization that mankind has established for thousands of years.

Of course, this view may still be a little radical, after all, they are still a "child" who failed the Turing test, even if the child has the "Superman" gene.

But one thing is certain: in the face of the positive progress of artificial intelligence, human beings are turning initiative into passive transfer of their decision-making power.

For example, user travel and urban traffic will produce huge data all the time, and human beings can only give it to the "black box" of artificial intelligence algorithms if they can't handle it.

Given the unprecedented push of the Chinese market for 5G this time, faster hardware and software facilities, a more open Internet market and a huge data set generated by 1.4 billion citizens are likely to keep the technological dividend of artificial intelligence in China; PwC predicts that artificial intelligence will bring $16 trillion in growth to the global economy by 2030.

Fourth, a bad news?

The bad news for artificial intelligence is that what artificial intelligence calls intelligence is still not "intelligence".

Machine learning, which is regarded as the standard in the industry, is not a machine learning on its own initiative, but is only "trained".

And as the most powerful driving force behind artificial intelligence, it is gradually losing faith in it: even though AlphaGO God blocks God and Buddha, DeepMind, the company behind it, still lost $500m last year.

On the one hand, AI's star company is losing money, and on the other hand, brain-like artificial intelligence programs: the collapse of Japan's ambitious "fifth-generation computer" that can reason like people, and the "blue brain project" that has spent 10 years burning up 1 billion euros trying to simulate 8.6 billion neurons and 100 trillion synapses in the brain.

All these cast a shadow over the "intelligence" of artificial intelligence.

In recent years, AI has realized the superiority of human brain biological system in processing information, and is trying to break through the shackles of "simulated program intelligence" to AI through "brain-computer integration". Researchers at the University of Berkeley are also exploring the "unintentional memory" of neural networks.

But this can only say that we have found the direction, the current artificial intelligence is still "mentally retarded".

So is artificial intelligence really impossible to have real intelligence?

Considering that the realization of human brain intelligence is to transmit activation and inhibition signals through chemicals, it does not seem to be impossible to simulate the existence of 0 and 1 nano-process chips.

Let's just say that we haven't found that "more suitable" algorithm yet, and everything we've done before is trial and error.

We do not know whether the artificial intelligence after the awakening of "consciousness" will choose to be "Zimaran" or "Ultron".

The bad news for ordinary people is that we have to passively give up some jobs, regardless of whether artificial intelligence is possible or not.

Even if the emergence of machines will create new jobs while replacing one job, the efficiency here cannot be equal.

Among the jobs that are about to be replaced, we once thought that the most dangerous ones were manual workers, but in fact, the most dangerous ones may be some mental workers and some white-collar workers.

On the one hand, manual workers are cheaper because they are cheaper; on the other hand, they are difficult to be replaced because of non-standardization; while the industrial industry on the assembly line is easy to be replaced, while some care for the elderly and children do hygiene and other jobs because non-standard robots are difficult to popularize. But there are no absolute barriers to many white-collar jobs, such as telemarketers, typists, accountants, insurance clerks, bank government employees, front desk customer service, and so on.

Even the creative industries that we once had high hopes for are being invaded by artificial intelligence.

Artificial intelligence can already write poetry and painting, even design and copywriting (Alibaba Luban, JD.com Shakespeare), and they are replacing job functions that do not have such a high threshold.

5. What can we do to defeat AI?

Where is the future of mankind if we hand over the computing to AI and the labor to robots?

The danger of AI is that while it tries to replace our "hands", it also tries to replace our "thinking".

Over time, there will be a kind of "alienation": we create machines, let them replace our work, then give them the right to think, and then immerse ourselves in the utopia that machines have programmed for us.

Isn't this another kind of slavery?

Perhaps our future is like a group of "fat balls" in the spaceship in "Robot Story".

Are we really ready for the arrival of the AI era?

Obviously, we didn't.

But fortunately, artificial intelligence only hears its voice but does not know its person for the time being.

Strictly speaking, I am not an artificial intelligence pessimist, and for now we don't have to worry about artificial intelligence taking our jobs. but take a long view: we don't seem to have a chance of winning in the face of artificial intelligence that doesn't need a salary, no rest, no leave, no mistakes, no mood, and the point is that the cost is much cheaper than hiring.

Perhaps the most pessimistic future is for mankind to become a pearl above the crown of silicon-based creatures like artificial intelligence, a touch of flesh and blood on top of steel and iron bones.

When we return to the industrial design industry, unfortunately, domestic industrial design has just started in the 1980s, and it has met with artificial intelligence unexpectedly before it has been brilliant; and once we are content with the status quo and replaced by AI, it will be a matter of time.

After all, the greatest characteristic of AI is to accelerate or even replace all types of work that require "process" with abundant computing power; as long as you give a target category and enter the characteristics of requirements, it can help you synthesize everything; disrelish that it provides too few choices, add money, add frame of reference, and improve the power of calculation, it can be extremely poor.

So in the face of artificial intelligence, when capital, trends and policies are not on our side, do industrial designers really have to fight it out?

Not exactly. The greatest advantage of creative people is to "open up wasteland". It is true that practice makes perfect in familiar areas, but it is inevitable that they will fall into a stereotype.

Only by spreading our ideas in areas that we have never seen before in artificial intelligence, we can completely do future urban design, space design, and even Martian immigration design. We will always be one step ahead of it and force ourselves to become the "frame of reference" of artificial intelligence in order to maintain our advantages.

Of course, it is also necessary for us to understand the language of artificial intelligence and try to "domesticate" it into their own "creative" tea.

And once we can jump out of the deterrent of artificial intelligence and have the ability to build our own set of industrial design standards based on standardized artificial intelligence; perhaps this is the unprecedented height of an industrial designer.

The Internet has subverted the tertiary industry, and artificial intelligence is about to "transform" the secondary industry and permeate the primary industry.

In this process, it will inevitably lead to a wave of wealth redistribution, but because it is too close to capital and too far from ordinary people, there is no need to have unrealistic expectations.

Hundreds of years after the Industrial Revolution, the landlords are still landlords and ordinary people are still on the assembly line.

But whether or not today's artificial intelligence is "real" intelligence, at least they are really useful; and for human beings as a whole, the advantages outweigh the disadvantages, and there is a chance to force the upgrading of human productivity as a whole.

Http://www.woshipm.com/ai/3011512.html

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.