“The pace of human creation of technology is accelerating, and the power of technology is growing at an exponential rate. Exponential growth is confusing. It starts with a very small growth, and then explodes at an incredible rate – if one does not pay close attention to its development trend, this growth will be completely unexpected.”

Ray Kurzweil, known as “the legal heir of thomasedison” in Inc. magazine, wrote in the book singularity approaching. The world’s leading inventor with 13 honorary doctorates has described the future of Artificial Intelligence Society for people.

Ray Kurzweil believes that due to the existence of Moore’s law, technology will grow exponentially, and human society will reach the singularity of artificial intelligence in 2045. Secondly, human beings based on biological morphology are essentially just an algorithm system under a set of highly complex neural networks, which will be replaced by more advanced algorithm systems in the future.

“Blind optimism may be the deadliest weapon of mass destruction.” Piero scaruffe said: “artificial intelligence is not a new concept. It originated in 1956 or more years ago. However, in the past, artificial intelligence did not get rapid development because the computer processing system was not powerful enough.”

Judging from the degree of application of artificial intelligence in reality, the current progress in the field of AI driverless seems to confirm Piero scarufi’s point of view. It is not difficult to find that the travel field has always been the forefront of the application of advanced technologies, whether it is the improvement of the steam engine or the invention of the internal combustion engine.

Tracing back to the origin, the technical basis for the outbreak of driverless technology in recent years also comes from the revolutionary achievements made by Hinton in the field of deep learning in 2006. Therefore, the deep learning algorithm based on neural network can be deeply applied in computer vision, speech recognition and computer behavior decision-making, thus forming the technical basis of driverless software. In the realization of driverless engineering application, There are no major technical barriers. Therefore, the ceiling of unmanned driving still lies in the limitations of AI technology based on deep learning.

On the other hand, L4 level automatic driving based on AI technology has begun to enter the commercialization stage. At present, Google waymo, Tesla autopilot, baidu Apollo and general cruise have all realized L4 level automatic driving.

Driverless Achilles heel

In 2016, in the United States, an autonomous Tesla models collided with a white trailer truck, resulting in the death of the driver. This is the first case of driverless car accident.

Later, according to the environmental analysis of the accident site, some professionals pointed out that under the direct light, the image recognition system relying on the camera failed to detect the white truck passing through the road ahead in time. At the same time, due to the low position of the millimeter wave radar and the general vertical angle of view of the millimeter wave radar within ± 5 °, when Tesla was close to the side of the trailer truck, the radar beam passed through the truck from the lower side, resulting in missed detection, Thus causing accidents. After the accident, Tesla improved the driverless system and revised the interpretation of autopilot on the official website.

In fact, the safety issue is indeed the Achilles’ heel of the full implementation of driverless technology. The driverless system based on AI technology with deep learning algorithm as the core has not really solved the problem of driving safety caused by “computer understanding deviation”.

From the perspective of AI technology evolution, the “intelligence” with deep learning algorithm as the core is not really intelligence, but the achievement of statistical “optimal solution” based on big data and deep learning algorithm under the principle of “dynamic programming”. Therefore, in order to solve the safety problem of driverless driving at present, the possibility of “unsafe” must be reduced to a red line lower than the probability of human car accidents under this framework, so as to have the “bottom line” for driverless driving into thousands of households.

In May this year, at the 6th China robot summit held in Ningbo, zhengnanning, an academician of the Chinese Academy of engineering, delivered a speech on the theme of intuitive AI and unmanned driving. Academician zhengnanning proposed that it is impossible to establish a scene model covering all under the algorithm model, but “to construct a human like autonomous driving based on cognition, so that AI autonomous driving has a human like decision-making mechanism, it can cope with the changes of traffic scenes with high dynamics and strong randomness.

In Xiaobian’s opinion, the establishment of algorithm model based on human thinking decision-making mechanism makes AI have human like “consciousness”, which can not be achieved with current technical conditions. On the one hand, human decision-making is often achieved through their own experience, rather than forming a single decision-making mechanism in the driving scene. On the other hand, perceptual factors often dominate the decision-making process of most people, Algorithmic decision-making is 100% rational decision-making, and in some specific cases, rational decision-making is often not the “optimal choice”.

In the film “mechanical public enemy” (also known as “I, robot”), Dale Spooner, starring Will Smith, fell into the water with a little girl in a car accident. After calculation, the artificial intelligence robot chose Dale Spooner with higher productivity and gave up the little girl’s life. If similar events occur in reality, it is obvious that human rescuers will give priority to saving the girl, Because this is the “optimal solution” under the constraint of human nature.

“Singularity” of AI driving technology under “Ai safety trap”

In the future, driverless driving will be fully applied to the field of travel at some time in the future. At that time, the existing traffic rules and even the road shape may have new changes. From the initial application of unmanned driving to the advent of the era of unmanned driving, people will be in a mixed travel era of “human +ai driving” for a long time. In this process, the corresponding laws and regulations must also adapt to it.

If safety is the “ticket” for AI to land driverless, then the adaptation of driverless to the existing traffic system and rules is a direct “game” between AI and human beings.

In essence, the evolution of AI driverless is a process in which human beings gradually hand over the travel part to AI on the premise of improving the convenience and safety. In this process, human beings retain the dominant power in the travel field and hand over the travel safety and control power to AI, so as to realize the liberation of human beings.

In this process, human beings, as one of the players in the game, have very contradictory psychology. On the one hand, people hope to liberate manpower through AI to obtain the “comfort” of travel experience. On the other hand, people worry that under the existing technical conditions, AI decisions will bring security risks and moral risks. Therefore, the landing of driverless is not only the landing of technology, but also the systematic adaptation of public recognition and driverless traffic regulations.

At the decision-making level, AI based on in-depth learning will not have a “human like” decision-making model for a long time. Therefore, people can expect that AI driverless is essentially a traffic aid under low safety risk. In this sense, the progress of AI driverless will increase human drivers’ falling into the “Ai safety trap”: on the one hand, the “non-human” AI can not really guarantee the safety of drivers, On the other hand, the increasingly advanced AI driverless technology will increase the “inertia” of drivers, resulting in potential safety risks.

In Xiaobian’s opinion, the key for unmanned driving to cross the “Ai safety trap” lies in whether it can accurately judge the singularity of AI unmanned driving technology evolution, and the principle of judging whether unmanned driving has reached the technological singularity can be considered from two aspects:

AI is fully capable of analysis and decision-making as a “person” (that is, artificial intelligence to realize independent thinking);

The accident rate of AI unmanned driving based on deep learning in actual road driving is far lower than that of human driving.

Secondly, from the perspective of reality, software programs are an indispensable component of AI technology. In the networked state, AI that obtains vehicle control is also more vulnerable to attacks by network hackers. Therefore, in addition to driving safety, network security is also a problem that needs to be solved for the real landing of unmanned driving.

So, how long will it take for a real driverless landing?

From the perspective of the development of AI technology, since the breakthrough in the field of deep learning was made in 2006, deep learning based on neural network has developed rapidly. Big data, deep learning algorithm and computing power have become the three core technologies in the field of AI. At present, the computing power in the three elements of AI technology still relies on powerful computers as logistics support. However, with the failure of Moore’s law, the traditional semiconductor industry is gradually facing a technical bottleneck, AI technology progress may face new stagnation.

The failure of Moore’s Law means that under the existing size, computer computing power is also facing a physical bottleneck, and the growth of AI technology needs the support of a large number of computing power. Therefore, it can be predicted that the growth of AI technology will fall into a new difficult period. At the same time, the stagnation of AI technology development will further limit the application of AI technology in the field of unmanned driving.

Under the existing AI technology and its growth space, in the future, the landing of unmanned driving will inevitably be divided into two stages, namely, the commercial landing in closed scenes and the commercial landing as a driving assistant function. There is still a long way to go to truly realize intelligent unmanned driving.

Ray Kurzweil’s “singularity approaching” makes people sigh that the era of artificial intelligence seems close at hand. However, as he wrote in his book, “people always overestimate the goals that can be achieved in the short term, but tend to underestimate the goals that will take a long time to achieve.” Perhaps, we still know little about the profound impact of real artificial intelligence on human society, but people should also give more rational cognition to the practical application of AI today, which is also the key to the long-term prosperity of AI technology.

Responsible editor: CT

Leave a Reply

Your email address will not be published.