Stanford University has published a report with predictions about the future of our future until 2030. Alexander Krainov, head of the computer vision and machine intelligence technology service at Yandex, singled out the most important things from him especially for Afisha Daily.
In 2014, Stanford University launched a 100-year study on artificial intelligence, in which scientists are going to evaluate how the introduction of new technologies in AI and machine learning affects society. It is planned to release reports on AI-related topics throughout the project. The first report in this series was recently published, which predicts what the future holds for us right up to 2030. Obviously, the future will not be the same for people in different countries, and the researchers are looking at some abstract North American city. Russia has its own specifics, and hence its own nuances of the development of the technological future. Let's try to figure out what the report tells us and how relevant it is for us.
Transport
Stanford researchers believe that it is in the field of transportation that the most noticeable changes associated with artificial intelligence will occur in the near future. We are talking about both built-in driver assistants (all kinds of sensors that analyze the state of the car and the situation on the roads) and unmanned vehicles, and traffic control systems, improved by big data analytics and machine learning, will save cities from traffic jams. In Holland, for example, they even thought about "smart roads", which would be stuffed with all kinds of sensors and help drivers assess the condition of the road "on the go."
With them, however, everything is also not so transparent: the problem here is not only the safety of driving, which will ensure the autopilot, but also in legal matters. Who will be to blame if a car without a pilot hits a person or crashes into another car? The driver can always be fined or revoked, but how can you fine the car? Another aspect is the attitude towards self-driving cars in society. Any incident involving the autopilot triggers a wave of discussion and provides arguments to opponents of innovations. The happy owners of high-tech cars like Tesla themselves do not help the situation - they sleep with the autopilot switched on and ignore the car's recommendations to take control.
Promotional video:
In addition to difficulties with the legislation, the use of such technologies in Russia can be complicated by the fact that our overall road situation is more complicated. This also applies to the quality of the road surface, and weather, and driving habits. All this will require a higher level of development in tuning algorithms. And our traffic laws are more conservative than the states or in Europe, and can take longer to change.
Health care
Healthcare is one of the most promising areas of AI adoption. The authors of the report agree with this, but note that it is also one of the most difficult industries. The price of error here is the patient's life, and any health data is very sensitive. Therefore, ethical issues in health care are especially acute. Both the bureaucracy and the outdated mechanisms of the work of medical institutions hinder - it will take a very long time to overcome these obstacles. But all this does not prevent technologies from actively developing, and new technology companies are entering the industry, including in Russia.
Mass collection of medical data (a necessary foundation for AI training) became possible some time ago, during the boom in sports apps and activity trackers, but large analytics still have not reached it for a number of reasons, including legal restrictions and privacy issues. The same goes for image recognition - for example, X-ray images that are already taken and stored digitally. With telemedicine, things are better - projects are being launched, including government ones, to introduce high-tech tools for doctors, such as remote participation of a surgeon in an operation using HD broadcasting. In the near future, it can be expected that machine intelligence will be able to analyze a mass of data on different patients and their treatment histories to highlight similar cases.give recommendations and thereby save the therapist's time. The trend here is not much different from other industries - all automated work, relying on the knowledge base in the human head and on comparing data, will in the future be replaced by AI. True, for a long time the final decision will still be for the person.
In Russia, the telemedicine sector has been looked at for a long time and intently, there is a state program for its implementation, the first stage of which will begin in 2017. While this program has nothing to do with AI, but it can indirectly contribute to the beginning of the introduction of AI in telemedicine - from automatic processing of textual information such as prescriptions for drugs to the analysis of images from patient records. Moreover, we are already working on the recognition of pathologies in images using neural networks and there is an obvious demand for access to highly qualified medical services in remote settlements.
Education
In the foreseeable future, robots will not replace teachers - this applies both to the United States and even more so to Russia, where the teacher has always been perceived as an educator as well. Researchers in the Stanford report pay attention not so much to how artificial intelligence will be implemented in education, but to questions about new technologies that help teachers and at some level replace them, for example, when passing educational online programs. The researchers cite the Carnegie Cognitive Tutor as an example, which helps students learn math: the system can adapt to the needs of each student - and, depending on them, changes tips and feedback on the passage of the lesson.
Intelligent training systems are also developing, which are widely used in the United States to train various specialists - from programmers to engineers. When a virtual adaptive environment is formed to solve specific problems in real life, AI helps in it to adjust the process to the student's actions. This is, for example, the Sherlock system, which was invented back in 1989 and is used to train technicians in the US Air Force. You can also note the significant progress of online translators, which is happening thanks to the use of machine intelligence. This makes educational literature in other languages more accessible.
Safety
Machine intelligence, which is already actively used in the security field, will be used more actively in the future. Researchers speculate that artificial intelligence will be able to help identify lies during interrogation. And analyzing large data sets of crimes, including the history of crimes in a specific area, video recordings and movements of suspects, can help predict where the next crime may occur - much like in the TV series Suspect. Also, don't forget about cybersecurity. Machine intelligence systems are already helping to detect financial crime based on suspicious activity on someone's credit card - such systems will become even more effective in the future.
Of course, the use of AI for tracking systems is of concern to people. But you can look at it differently, asking this question: what is better - if you are being watched through the camera by a "soulless algorithm" or a very specific person? Perhaps, in the first case, privacy is violated much less. AI aims to keep track of only dangerous patterns, and it just excludes constant human monitoring. Imagine an oil pipe that must be constantly monitored so that intruders do not illegally join it. You can set up cameras and occasionally launch patrols along the pipe, or you can launch a drone and use a trained system to analyze the terrain for the appearance of foreign objects nearby, for example, cars or groups of people. Yandex Data Factory and Accenture have a similar project - the system monitors long-distance objects such as power lines, oil pipelines and gas pipelines, which would be too expensive for people to patrol, and is able to detect suspicious activity - for example, unauthorized cars, groups of people, etc. …
Entertainment
Artificial intelligence has been used in entertainment for a long time - for example, in games, computer enemies build their behavior based on the actions of the player, which is an excellent example of artificial intelligence. On social media, recommendation algorithms also use AI, and the Facebook news feed is a classic example. They talk about the use of machine intelligence technologies in their blog: this includes translation of posts, and smart search, and adaptation of the feed to the interests of a particular user on the fly (depending, for example, on what he likes and what links he opens). However, all this is a relatively simple level of using complex technology and in the future, according to researchers, the degree of personalization of content will be much higher than today.
AI is also entering the field of art - more and more there are quite successful examples of composing texts and melodies by programs. For example, this year, Yandex enthusiasts made the Neural Defense project, where the neural network wrote lyrics in the style of Yegor Letov. Now these are nothing more than entertaining experiments, but in the future it is easy to imagine how neural networks paint pictures or create new pieces of music, and those that are almost guaranteed to become hits: after all, a neural network is able to identify the necessary conditions for a composition to become a hit.
There are no differences between the development of entertainment technologies in the United States and Russia. Here we are not lagging behind the West, and on the whole we are in for about the same prospects and problems. But making a recommendation system or entertainment bot based on English is easier - more data, and the language itself is more formalized. The Russian language is very difficult, which slows down the process somewhat.
Should you be afraid of unemployment?
One of the biggest fears of artificial intelligence is that it will take jobs away from humans. This is not to say that this fear is completely unfounded. Researchers at Stanford believe that while machine intelligence will indeed replace many people in a wide variety of industries, it will create many new jobs at the same time, but it is difficult to say which ones. In addition, AI will not replace the work of millions of people at once - this process will be extended in time and will be gradual in the sense that AI will first come to the aid of a human employee and only then will it be able to replace him. This will make the process of reducing the employment of people in some professions smooth and painless.