Foreign media reported that after decades of development, artificial intelligence has changed from a concept in science fiction to the mainstream of today’s technological development. Apple’s voice control system Siri, Amazon’s smart assistant Alexa, IBM’s Watson supercomputer, Google Brain, etc., intelligent machines that can understand the world and give efficient feedback seem to be in sight.
Powerful Internet access and computer algorithms have enabled machines to make a leap of progress in understanding the human voice and video information, and are gradually moving towards true artificial intelligence. AI (Artificial Intelligence), is artificial intelligence products that can realize automatic responses far away from us? How to keep them constantly updating and upgrading themselves but still keeping them within the controllable range of human beings?
This problem brings general worries and anxiety to people. “I think the development of artificial intelligence will eventually lead to the extinction of the human race,” physicist Stephen Hawking gave this view in an interview with the BBC. Elon Musk, the founder of Tesla, called AI “the biggest potential threat to mankind.” Bill Gates, former president of Microsoft, expressed similar concerns.
How much worry is necessary? How urgent is the problem? We have gathered a number of experts in the fields of industry, scientific research, and policy decision-making to assess the development risks of artificial intelligence. Participating in this topic discussion included one of the founders of Skype and a think tank member of the Institute of Potential Risks for Human Future Development-Jaan Tallinn, and Guruduth S. Banavar, an expert in charge of computer cognition research at IBM’s Watson Institute for Supercomputers. Professor of the Department of Computer Science at the University of Padua, the Radcliffe Institute for Advanced Study of Harvard University, and the president of the International Association for Artificial Intelligence Research-Francesca Ross.
The following is the edited content of the expert’s conversation.
Table of Contents
Influenced by science fiction and movies, the threat posed by artificial intelligence is mainly the conflict between humans and machines. This view is actually wrong. Artificial intelligence can solve problems better than humans in certain areas, such as big data screening. However, in some areas where humans are good at, machines perform very badly. Such as common sense reasoning, asking wonderful questions, and “jumping out” to think about questions. The combination of humans and machines is the foundation of our computer cognitive ability and a revolutionary idea for solving complex problems in various fields.
Artificial intelligence-based systems have provided convenience to our lives in many ways: stock automatic trading systems, automatic flight control systems, industrial flow robots, error detection, and search engines. In the past 5-10 years, Computer algorithms and advanced computer architecture have allowed us to develop many new applications.
However, it is necessary to realize that computer algorithms have been applied to the extreme. To achieve greater progress, you need to introduce more complex symbol systems-this is a very high requirement. Today, our neuroscience and cognitive science researches have only scratched the surface for the study of human intelligence.
My personal opinion is that artificial intelligence machines that can reach the human level and realize general perception and speculation are just a topic of entertainment.
Today’s artificial intelligence is unlikely to pose a threat to mankind. If we want to discuss the long-term impact of artificial intelligence, (that is, when artificial intelligence can perform strategy, science, and artificial intelligence research better than humans), we must involve the issue of controlling super intelligence.
The most advanced robots now also have an “off button” that allows humans to achieve complete control. However, the prerequisite for the “shutdown button” to work is that it is outside the robot’s work area. For example, a computer that can play chess is limited to playing games with people, and it does not perceive that its human opponent can unplug its power to terminate the game.
However, super-intelligent machines, through intentional or unintentional programming, can fully understand the surrounding world and predict the consequences of someone who can press their “shutdown button”, and it is difficult for humans to control it.
At the moment this is only theoretical speculation. But it needs to be carefully recognized that super-intelligent machines may only be controlled by the laws of physics and programs set by people for early intelligent robots.
The early programs were designed based on our very limited knowledge of physics. Imagine that if we are faced with an intelligent robot and the instruction we give is to “create a greater future for mankind”, we had better be very sure of how it will react.
In many specific fields, the performance of artificial intelligence has surpassed that of humans, including some fields involving precise judgment. At present, artificial intelligence does not threaten humans, but humans have been affected by machine judgments, such as online automated trading systems, health check systems, and unmanned vehicles, and unmanned weapons that will soon be popularized.
We need to clearly understand and analyze the potential threats of artificial intelligence in specific fields, not only to make them work, but also to ensure that they can be safe, friendly, and operate in accordance with human values. This is not a simple subject, because even humans cannot act in accordance with the principles set by themselves most of the time.
For example, in automated trading systems, a wrong decision made by a machine will cause economic losses to many people. The mistakes of driver less cars will also threaten the safety of human life.
Any risk has two sides: the risk of doing it and the risk of not doing it. We have already made many wrong decisions based on incomplete information and unprofessional professionals, resulting in thousands of deaths, millions of dollars in losses, and slow development in core areas such as medical care. Based on the application experience of IBM Watson supercomputers in many fields, I believe that in specific fields, artificial intelligence can effectively reduce risks.
However, I have a cautious view of the field of comprehensive artificial intelligence. I think we have always underestimated the complexity of this issue. We have too many unknowns.
IBM’s computer cognition system, like other modern computer systems, is based on cloud computing structure, algorithms, and large amounts of data. In order to prevent violations, the operation of these systems can be recorded, tracked, and reviewed. These cognitive systems do not operate autonomously, so their instructions, data, and structure require us to take certain protective measures to prevent man-made external attacks. Anyone who accesses and updates the system is controllable.
We mainly protect data through complex encryption and digital signatures, protect computing code through vulnerability scanning and other technical means and protect computer structures through isolation and other means.
These policies to support the safe operation of artificial intelligence are designed based on an in-depth analysis of risks. These policies need to be appointed by security and transaction, and the committees that manage sales and public safety (SEC), the Food and Drug Administration (FDA), and the National Institute of Standards and Technology (NIST) are the institutions responsible for setting safety standards in their respective fields. Review.
Can it be responsible for the medical injuries caused by its own mistakes?
The Watson supercomputer will not give a diagnostic opinion. It provides a variety of treatment suggestions to doctors by processing large amounts of patient medical data. The doctor will consider Watson’s recommendations and comprehensively analyze other factors to make a final treatment judgment.
Doctors are also making mistakes all the time, mainly because they cannot fully grasp all the information about a certain disease. A system like Watson Computer can help them reduce mistakes.
There was a study comparing doctors and medical automatic diagnosis systems. The conclusion is that most doctors lost to computers.
I think it is very beneficial for Watson Computer to provide more comprehensive medical advice. I very much hope that it will have the opportunity to diagnose me.
Technological progress does make some jobs disappear. An Uber driver once excitedly explained to me how he entered the destination on the tablet to get the itinerary. I told him that in a few years, it might be these small technological objects that brought me to my destination. The driver said happily that he would be more trouble-free, just sit in the driver’s seat and enjoy it. Sitting in the back row, I couldn’t help shaking my head. In my opinion, the biggest impact of driver less cars on mankind is not convenience, but lost jobs.
In the long run, in addition to work, we should also imagine what else our society can revolve around.
Throughout history, we have been building tools to help us accomplish things we couldn’t do. Each generation of tools will trigger our thinking about nature and working methods. Productivity increased, jobs were redefined, new occupations were born, and some occupations disappeared. The computer cognition system has expanded human thinking ability and has a stronger potential for substitution.
The key to the problem is that an organization should be established to quickly train professionals so that they can develop computer cognitive systems as assistants to assist them in their respective work. These skills can enhance human professionalism and set a new standard for “experts”.
There is a lot of uncertainty about when full artificial intelligence will be realized or even if it can be realized. (Although as a physicist, I think everything is possible!) But uncertainty is not an excuse for not thinking about control issues. For the issue of how to control, targeted research has just begun, and it may take decades to produce results.
I believe that we can completely minimize the adverse effects of artificial intelligence applied in specific fields on humans. We need to set human value standards for machines, set work principles and priorities, such as conflict resolution capabilities consistent with humans. The completion of the artificial intelligence design work in a specific field can prepare for the design of a comprehensive artificial intelligence that does not harm humans.
In the early 1990s, when computers entered the medical field, social activists in many countries called for the protection of patients’ rights. Ten years later, the medical information confidentiality agreement was born. Similarly, artificial intelligence in specific fields is also facing a serious problem, which rights should not be violated, and what responsibilities should be borne after infringement?
Selling ordinary cars now has to pass a series of safety tests, and selling driver less cars in the future will also require new safety test standards. What needs to be carefully tested is not only the car operating procedures but also the emergency judgment procedures designed on the basis of human ethics and moral principles.
What are the rights of drivers, passengers, and pedestrians in the world of driver less cars? Can consumers not provide all the information to investment advisors and smart assistants? Who should be responsible for investment advice? Is it an investment consultant, an investment service organizer, an intelligent assistant designer, or a data custodian? There are still many problems that human society with intelligent machines will face and solve in the future.
In the ever-evolving landscape of logistics and supply chain management, the role of technology has… Read More
A Mid Cap fund invests in equity and equity-related instruments of mid-cap companies. According to… Read More
In an undeniably serious and cutthroat business climate, organisations persistently look for techniques to further… Read More
The former pioneer of BlackBerry smartphones, converted to embedded and cybersecurity, announces its intention to… Read More
Sextoys have long been shrouded in stigma and secrecy, creating unnecessary barriers to open discussions… Read More