What artificial intelligence has achieved in a year. Artificial intelligence: what they promise us and what we risk

Technology and media, 13 Sep, 09:00

AI researchers from Russia lost to competitors in terms of the number of publications ... development artificial intellect(artificial intelligence, AI) - are significantly inferior to world competitors in terms of the number of scientific publications devoted to this technology. Such a conclusion is given in the almanac " Artificial intelligence... own juice. As the head of the data analysis department and artificial intellect HSE University Sergey Kuznetsov, scientific publications are an indicator of academic development...

Finance, 10 Sep, 16:02

Putin proposed to use artificial intelligence in public administration ... the head of Sberbank German Gref suggested using the bank's achievements in the field artificial intellect to solve problems in the field of public administration. The transcript of the meeting is published on ... . In February, RBC sources reported that Putin was exploring the possibility of introducing artificial intellect into the Russian economy, public administration, banks, industry and agriculture... Facebook unveiled the results of work on a device for reading minds The technology allows decoding human thoughts and translating them into text. To do this, the test subject is placed in the brain with special electrodes, but in the future, developers want to abandon this approach and make the process non-invasive. Facebook has conducted the first tests of a technology that allows you to transform thoughts into text. The results of my research... Sberbank implemented artificial intelligence in a mobile app ... adapts to the client according to his preferences thanks to the integrated artificial intelligence. The new version of the application is available for iPhone and Android users from... » through the official mobile phone stores. The main update is to introduce artificial intellect: the neural network will analyze the client's preferences for more than... Musk's company announced the first steps to connect the brain to the computer ... rather strange, but eventually we will achieve symbiosis with artificial intellect", - said the day before at a press conference in San Francisco Musk. However... Putin admonished the Cabinet of Ministers and state-owned companies with the words "there is no time to sniff" ... TASS, including on joint work in the field of development artificial intellect with Sberbank and on the development of the "Quantum Communications" direction with Russian Railways... Roscosmos announced plans to explore the Moon with the help of robots ... Roscosmos "expects to explore the moon using robots with elements artificial intellect. It is reported by TASS with reference to the press service of the state corporation... American billionaire makes biggest donation to Oxford ... 2024 Institute of Ethics in artificial intellect. The new division will study ethical issues related to development artificial intellect and other computer technologies. Schwartzman told the BBC he donated the money to Oxford University because artificial intelligence- main... Putin instructed to conclude agreements on high technologies with state-owned companies ... was made as a result of a meeting on the development of technologies in the field artificial intellect which took place on May 30th. The head of state proposed to conclude agreements with ... last week, Putin announced the creation of a development strategy in Russia artificial intellect. He noted that as part of the strategy, the authorities will develop a step-by-step plan... Putin announced the creation of a strategy for the development of artificial intelligence ... sphere artificial intellect, then she will become the "ruler of the world." Then the Russian president noted that the struggle for technological leadership, including in the field artificial intellect, has already become a field for global competition. In 2017, Putin said that development artificial intellect brings not only... SberCloud will develop artificial intelligence with American Nvidia ... services based artificial intellect. In particular, companies will develop an implementation strategy artificial intellect into the Sberbank ecosystem, finalize the SberCloud solution in the artificial intellect AI Cloud, which ... a single management portal will be able to work with all major areas of technology artificial intellect: speech analysis and synthesis, computer vision, character and text recognition ... Sberbank gave a forecast for the damage to the global economy from cyber attacks in 2019 ... . In particular, it is necessary to determine who is responsible for decisions taken by artificial intellect, determine the procedure for investigating crimes in the virtual world, develop the foundations for ... two to five years to create a federal agency for robotics in Russia and artificial intellect. Natalia Demchenko IT executives saw in AI the danger of "forgetting how to think" ... and 71% of the general public are concerned that the emergence of a full-fledged artificial intellect(AI) will lead to a decrease in mental abilities in humans. With results...

Finance, Feb 26, 11:59 am

Gref admitted the loss of billions of rubles due to artificial intelligence ... bugs work artificial intellect improved, Gref continued. “We learned [from every mistake], inserted all sorts of filters to calibrate, verify or validate the system artificial intellect", - he said. Responding to a request from RBC about the losses from the introduction artificial intellect, the press service of Sberbank specified ...

Economics, Feb 01, 16:48

Putin will hold a meeting on the introduction of artificial intelligence in Russia ... develop approaches to a national development strategy artificial intellect and submit relevant proposals. Intelligence on the map​ Support for neurotechnologies and artificial intellect in Russia is provided for by the national program "Digital... in various working groups, including those dedicated to artificial intelligence. According to the representative of MTS, artificial intelligence- one of the key directions of the company's development. However, neither... Soros called Xi Jinping the most dangerous opponent of democracy ... about the serious danger of leadership using authoritarian regimes such as China, artificial intellect and called Chinese President Xi Jinping the most dangerous opponent of the democratic ... World Economic Forum in Davos, Bloomberg reports. "Tools of control developed artificial intellect give an inherent advantage to totalitarian regimes over open societies,” the agency cites... Fakes and robots: what will be the main technology trends of 2019 ... for artificial intellect Telenor predicts that governments and private companies around the world will start setting frameworks for governance next year. artificial intellect, to adopt codes of conduct to promote high ethical standards in this area. Systems artificial intellect should be non-discriminatory, transparent... Microsoft confirms readiness to sell its technology to the Pentagon ... creating reliable protection", including for the assembly of weapons equipped with artificial intellect. This was stated by Microsoft President Brad Smith, according to The ... artificial intellect, which can be used by the military. However, at the end of July... Computer-generated painting sold at auction for $432,500 Created artificial intellect the painting was sold for $432.5 thousand at Christies auction. Oh... give out in him a man of the church. This portrait was the first created artificial intellect painting sold at auction. The Parisian team participated in its creation...

Own business, 13 September 2018, 14:06

The root of the trouble: how a $ 1 billion startup prepared a revolution in auto insurance ... the only American startup that insures customers by calculating risk using artificial intellect. Founded in 2015, New York-based Lemonade deals with real estate insurance... , co-founder of MoneyRock artificial intellect, are not insurers, but rather fintech startups and some banks ... China launches 'rumour debunker' online ... has integrated over 40 local rumor debunking platforms and uses artificial intelligence to identify rumors. It operates under the control of 27 government departments... Amnesty International called on humanity to ban combat robots ... the years have moved from the realm of science fiction to reality. "The emergence of endowed artificial intellect drones, automatic guns that can choose their own targets - testifies ... the next five years to spend almost $ 900 million on the introduction of technologies artificial intellect. Artificial the US military intends to use the mind precisely for those functions on ... The media learned about plans to entrust the examination of the exam to artificial intelligence ... Federal Institute of Pedagogical Measurements Oksana Reshetnikova. According to her, development artificial intellect by that time should have reached a level that allows you to evaluate detailed answers ... any artificial intelligence", he said. According to him, if the answer to the exam involves unambiguously correct answers or solutions, then the option with artificial intellect ... Artificial intelligence Microsoft named the second finalist of the World Cup-18 ... France. The machine named the England team as the second participant in the final. At the rate artificial intellect Bing by Microsoft, France will be the finalists of the World Cup... on social networks. Before the start of the World Cup, he also made his predictions artificial intelligence developed by the investment bank Goldman Sachs. However, his predictions turned out to be wrong - so... Google has developed an algorithm to predict the death of hospitalized ... diagrams (saved in .pdf format) that were previously unavailable for artificial intellect. Google technology can be especially useful in radiology, cardiology, ophthalmology and... . In early June, Google imposed a ban on development in the field artificial intellect for military purposes. The company's CEO Sundar Pichai noted that modern ... Google banned the development of artificial intelligence for the military ... participate in a project to create artificial intellect for military drones. Sundar Pichai recalled that technology is already artificial intellect and machine learning have a big... keeping people safe, promised Sundar Pichai. He especially emphasized that artificial intelligence must remain under the control of the people. “Our AI technologies will be located... Rosatom will offer startups up to ₽3 billion for smart cities and 3D printing ... Rosatom" will invest 3 billion rubles. in artificial intelligence and digital solutions. In 2017, the president instructed state-owned companies to support ... those associated with promising sectors of the Russian and world economy. Key directions: artificial intelligence and other digital solutions in industry and service applications, renewable... Rostelecom recorded a drop in investments in Russian startups ... global digitalization trends. The company named mobile communication as the most important directions, artificial intelligence and e-commerce. The rating was compiled on the basis of an analysis of 6 million ... % of revenue for investments in high technologies, such as virtual networks, artificial intelligence, industrial internet, security, biometrics. Senior Telecommunications Analyst, Uralsib Capital... Head of artificial intelligence at Google moved to Apple Head of Development in the field artificial intellect At Google, John Giannandrea joined Apple and took over... artificial intellect, says CBNC. John Giannandrea joined Google in 2010, under his leadership in the development of artificial intellect implemented...

Technology and media, 13 Mar 2018, 09:38

Yandex allowed users and companies to create chatbots for Alisa Yandex launches the Yandex.Dialogues platform, which allows you to register and develop your own chatbots, or services for the voice assistant of the Alisa company. As a representative of Yandex told RBC, so far the platform is in beta mode, in the future the functionality of Dialogues will be extended to other Yandex products. With the help of Yandex...

Technology and media, 01 Mar 2018, 13:18

Putin urged to make Russia one of the world's centers of "big data" ... , digital maritime and air navigation. Logistics needs to be organized with the help of artificial intellect Putin noted. Earlier, the President announced the need to prepare a plan for the modernization of... In China began the development of artificial intelligence for the needs of the army ... by 2023, a technology development park will be created in Beijing artificial intellect. The area of ​​the park, located in the western district of Mentugu, will be 54.87 ... enterprises, their annual profit is estimated at $ 7.7 billion. In addition to artificial intellect in the park will be engaged in the development of ultra-high-speed transmission of large amounts of data, cloud ... Sberbank held a business breakfast in Davos with the participation of the robot Sophia ... million net profit using one of the methodologies artificial intellect in risk management. Total through the use artificial intellect and data analysis, the bank earned $2 ... through the use of artificial intellect and data analysis in risk management and sales. The topic of Sberbank's business breakfast was "Technology artificial intellect changing companies and... China announced the creation of a park for the development of artificial intelligence ... In 2023, Beijing will set up a technology development park artificial intellect. This is reported by Xinhua News Agency, citing local authorities ... for the integration of science and education in the country. In the park besides artificial artificial Unlimited rationality: how robots will make deals ... depend on decisions made by computers. Therefore, the development of a code of ethics for artificial intellect should become a key problem for both governments and IT ... and entrepreneurs have already drawn attention to the unwillingness of humanity to offer for artificial intellect(AI) a clear code of ethics, which would be, in particular, the basis ...

Kai Fu Lee, chairman of SinovationVentures, believes that AI is "aimed at large-scale job cuts" while concentrating wealth in the hands of companies that develop or adopt AI. Others believe that such fears were present during the emergence of all world-changing technologies up to the printing press in the 15th century.

The Economist reassures readers that "AI is creating demand for jobs" and a growing number of people around the world are "providing digital services online." Which companies and countries will thrive in the AI ​​era? Which segments will disappear, change, be created? How will the nature of work change?

Warfare

Proponents of combat drones argue that such weapons can hit targets with much greater accuracy than humans; and the larger the role they play in the theater of operations, the less often technicians will use them to harm.

But what if such weapons become independent and work on their own, without human intervention? Will the removal of people from the list of military personnel lead to an even more severe and unstoppable arms race?

An open letter published during the International Joint Conference on Artificial Intelligence in 2015 warned that autonomous weapons "do not require expensive or hard-to-find raw materials, so will become ubiquitous and cheap for all significant military forces to mass-produce." Will the era with automated weapons be more peaceful or more militant?

The RAND researchers are calling for an analytical framework and an international effort dedicated to the use of long-range combat drones in counter-terrorism and targeted assassinations.

Making decisions

Politicians are constantly faced with a huge number of choices and motives - in the days of social networks there are many more than twenty years ago. Such information overload does not allow coping with the situation during a crisis, not to mention multiple crises.

Recently, there was a proposal to pass "all decisions that the president makes through the computer - not to make the final choice, but to help the leader in the face of a person."

But while AI is now for the most part sinless, the RAND study highlights the risks of algorithmic bias in news filtering, influencing criminal justice, and even social security benefits and visas. What decisions should be left to AI? What should remain in the hands of man? In the hands of a team of people?

Creation

The world is accustomed to AI that can perform breathtaking computational feats and defeat humans at popular board games (it's been just over 20 years since IBM's DeepBlue supercomputer famously defeated chess grandmaster Garry Kasparov). How will he continue to progress in the creative space of people?

Artificial intelligence researcher Jesse Engel believes it will "redefine the creative process... by augmenting it with smart tools that provide new opportunities for expression." Others are not so optimistic. Journalist Adrienne Lafrance notes that AI can already "flirt", "write novels" and "forge famous paintings with astonishing accuracy". What does it mean to be creative? Moreover, what does it mean to be human?

Discussions of AI often go to extremes, whether it's the promise of a utopia free of human suffering or the danger of a dystopia where robots enslave their human creators. A more balanced and thorough analysis is needed to help shape policies to mitigate risks and maximize benefits. It is necessary to take certain steps to overcome the fears that AI will work for the state and society.

How can AI affect the national interests of a country? What types of AI, if any, can be considered strategic technologies based on government criteria? Where should market forces play, and where should politics? While AI remains largely science fiction, these questions are becoming more and more important.

Despite the scale of the event, it could be seen that AI, like machine learning, occupies a strong position in the field of mobile development and is not going to retreat. Here is a brief overview of the new products in this niche.

Artificial intelligence

  • Motorola+Alexa

Motorola has partnered with Amazon and wants to integrate its Alexa voice control device with its products. This is just the tip of the iceberg: Motorola plans to have the AI ​​assistant perform some useful actions on its own based on the owner's personal data (for example, book a table in a restaurant if such a need was mentioned in the email).

  • line inc. + Clova

Meanwhile, the Asian company Line Inc., the creator of the messenger of the same name, is working on the creation of an assistant Clova, which in the future may compete with Alexa. Apparently, it will be capable of quite complex user interaction: after creating the original application, developers will most likely integrate Clova with other applications and services, giving the user the possibility of ubiquitous voice control.

  • LG vs Sony + Google Assistant

Through close collaboration with Google, LG is also keeping up with the AI ​​boom. The G6 was the first phone since the Pixel to feature Google Assistant. But now Google says the assistant is coming. For example, on the XZ Premium smartphone announced by Sony, which now has every chance to rise in the ranking among new products. Meanwhile, HTC is working on its Sense Companion and the Samsung Galaxy S8 is expected to feature Bixby.

Machine learning

  • Huawei

The algorithms underlying the Huawei P10 software should calculate the user's habits and daily routine while the device is running, and based on this data, allocate memory and other resources, thus gradually improving the UX. The ultimate goal is to make the phone run faster and faster over time.

  • Netflix

The company focuses on improving the quality of the video you watch on mobile devices with an average or low Internet connection speed. The algorithms used must learn to understand what exactly makes the video quality and enjoyable to watch. Added to this is a special coding scheme that compresses individual scenes of a movie / series in different ways, depending on what happens in them.

  • RoboCar

And finally, one of the exhibits that attracts the most attention: the autonomous racing car RoboCar was presented at the exhibition. It perfectly combines the sci-fi design of the body and the filling of a complex of radars, lidars, ultrasonic and high-speed sensors, each of which transmits a signal to the NVIDIA “brain”, which interprets all the information it receives on the fly.

Making predictions is a thankless task, especially since progress in the field of modern machine learning methods is ahead of the wildest expectations in recent years. But some areas in the field of training deep neural networks, in which we can expect significant progress in the near future, I dare to name.

Firstly, this is the development of the ideas of neural network reinforcement learning, which will allow the development of new self-learning algorithms for agents interacting with the environment. These can be both robots and programs operating in virtual space, for example, playing intellectual games such as Go (already done) or Starcraft (in progress). The main goal here, of course, will be the creation of an algorithm that can “on the fly” adapt to a new complex game or environment.

Secondly, it is the development of new learning methods on the fly and meta-learning. The first allows the computer to grasp new concepts and meanings from a few examples, similar to how a person does, and unlike modern neural networks, which learn a new concept after showing thousands and tens of thousands of examples.

The second allows the neural network itself to select the parameters of its training method. Now the quality and speed of training neural networks significantly depends on the setting of a number of parameters (usually called hyper-parameters to distinguish them from the network weights, which, in fact, are adjusted during training), as well as on the architecture of the network itself. Now they are determined by a person or semi-automatic procedures that are far from optimal. Because of this, neural networks learn longer and worse than they could.

The works that appeared in 2016 show that this work can, in principle, be entrusted to an auxiliary neural network. As we all remember from high school, the sign of the end of the industrial revolution is the moment when "machines begin to make machines." Perhaps in the future, the moment when neural networks begin to train neural networks will be the same important milestone, and there is reason to believe that this will happen as early as 2017.

Thirdly, neural networks will learn to speak with a person (both in terms of generating replica texts and in the sense of synthesizing speech indistinguishable from human), generate photorealistic pictures and video sequences according to a text description, and write large meaningful texts. This will be our near future thanks to the rapid progress in the field of so-called. generative deep learning models. Of course, this will lead to the creation of new businesses, the emergence of new types of goods and services, as well as an increase in labor productivity in traditional sectors of the economy, such as mobile operators or banks, which will be able to abandon expensive and inefficient call centers.

Solving all these problems will be an important step towards the Holy Grail of machine learning - the creation of artificial intelligence. AI, of course, will not appear next year, but in 5-10 years it will undoubtedly be developed. Moreover, the already existing elements of artificial intelligence will help scientists in creating a full-fledged AI and, thereby, will speed up work in this direction. The creation of AI will be the most important achievement of mankind and will provide it with a powerful civilizational leap forward.

It is important to note that the rapid progress in the field of AI has become possible largely due to the fact that these developments are carried out openly and any person with the minimum necessary training (for example, a graduate of the HSE Faculty of Computer Science) can take part in them: even large IT -corporations - leaders in the field of deep learning, there are no secrets (except for near-term commercial ones), software implementations of most methods are available, as well as a mathematical description of algorithms that turn out to be surprisingly not so complicated, given the global nature of the tasks they solve. This compares favorably with machine learning, say, from the space or nuclear programs of the mid-20th century.

Additional materials

Lecture by Dmitry Vetrov on machine learning (in order to understand what it is in general)