Connect with us

Science & Technology

Future Goals in the AI Race: Explainable AI and Transfer Learning

Published

on

Recent years have seen breakthroughs in neural network technology: computers can now beat any living person at the most complex game invented by humankind, as well as imitate human voices and faces (both real and non-existent) in a deceptively realistic manner. Is this a victory for artificial intelligence over human intelligence? And if not, what else do researchers and developers need to achieve to make the winners in the AI race the “kings of the world?”

Background

Over the last 60 years, artificial intelligence (AI) has been the subject of much discussion among researchers representing different approaches and schools of thought. One of the crucial reasons for this is that there is no unified definition of what constitutes AI, with differences persisting even now. This means that any objective assessment of the current state and prospects of AI, and its crucial areas of research, in particular, will be intricately linked with the subjective philosophical views of researchers and the practical experience of developers.

In recent years, the term “general intelligence,” meaning the ability to solve cognitive problems in general terms, adapting to the environment through learning, minimizing risks and optimizing the losses in achieving goals, has gained currency among researchers and developers. This led to the concept of artificial general intelligence (AGI), potentially vested not in a human, but a cybernetic system of sufficient computational power. Many refer to this kind of intelligence as “strong AI,” as opposed to “weak AI,” which has become a mundane topic in recent years.

As applied AI technology has developed over the last 60 years, we can see how many practical applications – knowledge bases, expert systems, image recognition systems, prediction systems, tracking and control systems for various technological processes – are no longer viewed as examples of AI and have become part of “ordinary technology.” The bar for what constitutes AI rises accordingly, and today it is the hypothetical “general intelligence,” human-level intelligence or “strong AI,” that is assumed to be the “real thing” in most discussions. Technologies that are already being used are broken down into knowledge engineering, data science or specific areas of “narrow AI” that combine elements of different AI approaches with specialized humanities or mathematical disciplines, such as stock market or weather forecasting, speech and text recognition and language processing.

Different schools of research, each working within their own paradigms, also have differing interpretations of the spheres of application, goals, definitions and prospects of AI, and are often dismissive of alternative approaches. However, there has been a kind of synergistic convergence of various approaches in recent years, and researchers and developers are increasingly turning to hybrid models and methodologies, coming up with different combinations.

Since the dawn of AI, two approaches to AI have been the most popular. The first, “symbolic” approach, assumes that the roots of AI lie in philosophy, logic and mathematics and operate according to logical rules, sign and symbolic systems, interpreted in terms of the conscious human cognitive process. The second approach (biological in nature), referred to as connectionist, neural-network, neuromorphic, associative or subsymbolic, is based on reproducing the physical structures and processes of the human brain identified through neurophysiological research. The two approaches have evolved over 60 years, steadily becoming closer to each other. For instance, logical inference systems based on Boolean algebra have transformed into fuzzy logic or probabilistic programming, reproducing network architectures akin to neural networks that evolved within the neuromorphic approach. On the other hand, methods based on “artificial neural networks” are very far from reproducing the functions of actual biological neural networks and rely more on mathematical methods from linear algebra and tensor calculus.

Are There “Holes” in Neural Networks?

In the last decade, it was the connectionist, or subsymbolic, approach that brought about explosive progress in applying machine learning methods to a wide range of tasks. Examples include both traditional statistical methodologies, like logistical regression, and more recent achievements in artificial neural network modelling, like deep learning and reinforcement learning. The most significant breakthrough of the last decade was brought about not so much by new ideas as by the accumulation of a critical mass of tagged datasets, the low cost of storing massive volumes of training samples and, most importantly, the sharp decline of computational costs, including the possibility of using specialized, relatively cheap hardware for neural network modelling. The breakthrough was brought about by a combination of these factors that made it possible to train and configure neural network algorithms to make a quantitative leap, as well as to provide a cost-effective solution to a broad range of applied problems relating to recognition, classification and prediction. The biggest successes here have been brought about by systems based on “deep learning” networks that build on the idea of the “perceptron” suggested 60 years ago by Frank Rosenblatt. However, achievements in the use of neural networks also uncovered a range of problems that cannot be solved using existing neural network methods.

First, any classic neural network model, whatever amount of data it is trained on and however precise it is in its predictions, is still a black box that does not provide any explanation of why a given decision was made, let alone disclose the structure and content of the knowledge it has acquired in the course of its training. This rules out the use of neural networks in contexts where explainability is required for legal or security reasons. For example, a decision to refuse a loan or to carry out a dangerous surgical procedure needs to be justified for legal purposes, and in the event that a neural network launches a missile at a civilian plane, the causes of this decision need to be identifiable if we want to correct it and prevent future occurrences.

Second, attempts to understand the nature of modern neural networks have demonstrated their weak ability to generalize. Neural networks remember isolated, often random, details of the samples they were exposed to during training and make decisions based on those details and not on a real general grasp of the object represented in the sample set. For instance, a neural network that was trained to recognize elephants and whales using sets of standard photos will see a stranded whale as an elephant and an elephant splashing around in the surf as a whale. Neural networks are good at remembering situations in similar contexts, but they lack the capacity to understand situations and cannot extrapolate the accumulated knowledge to situations in unusual settings.

Third, neural network models are random, fragmentary and opaque, which allows hackers to find ways of compromising applications based on these models by means of adversarial attacks. For example, a security system trained to identify people in a video stream can be confused when it sees a person in unusually colourful clothing. If this person is shoplifting, the system may not be able to distinguish them from shelves containing equally colourful items. While the brain structures underlying human vision are prone to so-called optical illusions, this problem acquires a more dramatic scale with modern neural networks: there are known cases where replacing an image with noise leads to the recognition of an object that is not there, or replacing one pixel in an image makes the network mistake the object for something else.

Fourth, the inadequacy of the information capacity and parameters of the neural network to the image of the world it is shown during training and operation can lead to the practical problem of catastrophic forgetting. This is seen when a system that had first been trained to identify situations in a set of contexts and then fine-tuned to recognize them in a new set of contexts may lose the ability to recognize them in the old set. For instance, a neural machine vision system initially trained to recognize pedestrians in an urban environment may be unable to identify dogs and cows in a rural setting, but additional training to recognize cows and dogs can make the model forget how to identify pedestrians, or start confusing them with small roadside trees.

Growth Potential?

The expert community sees a number of fundamental problems that need to be solved before a “general,” or “strong,” AI is possible. In particular, as demonstrated by the biggest annual AI conference held in Macao, “explainable AI” and “transfer learning” are simply necessary in some cases, such as defence, security, healthcare and finance. Many leading researchers also think that mastering these two areas will be the key to creating a “general,” or “strong,” AI.

Explainable AI allows for human beings (the user of the AI system) to understand the reasons why a system makes decisions and approve them if they are correct, or rework or fine-tune the system if they are not. This can be achieved by presenting data in an appropriate (explainable) manner or by using methods that allow this knowledge to be extracted with regard to specific precedents or the subject area as a whole. In a broader sense, explainable AI also refers to the capacity of a system to store, or at least present its knowledge in a human-understandable and human-verifiable form. The latter can be crucial when the cost of an error is too high for it only to be explainable post factum. And here we come to the possibility of extracting knowledge from the system, either to verify it or to feed it into another system.

Transfer learning is the possibility of transferring knowledge between different AI systems, as well as between man and machine so that the knowledge possessed by a human expert or accumulated by an individual system can be fed into a different system for use and fine-tuning. Theoretically speaking, this is necessary because the transfer of knowledge is only fundamentally possible when universal laws and rules can be abstracted from the system’s individual experience. Practically speaking, it is the prerequisite for making AI applications that will not learn by trial and error or through the use of a “training set,” but can be initialized with a base of expert-derived knowledge and rules – when the cost of an error is too high or when the training sample is too small.

How to Get the Best of Both Worlds?

There is currently no consensus on how to make an artificial general intelligence that is capable of solving the abovementioned problems or is based on technologies that could solve them.

One of the most promising approaches is probabilistic programming, which is a modern development of symbolic AI. In probabilistic programming, knowledge takes the form of algorithms and source, and target data is not represented by values of variables but by a probabilistic distribution of all possible values. Alexei Potapov, a leading Russian expert on artificial general intelligence, thinks that this area is now in a state that deep learning technology was in about ten years ago, so we can expect breakthroughs in the coming years.

Another promising “symbolic” area is Evgenii Vityaev’s semantic probabilistic modelling, which makes it possible to build explainable predictive models based on information represented as semantic networks with probabilistic inference based on Pyotr Anokhin’s theory of functional systems.

One of the most widely discussed ways to achieve this is through so-called neuro-symbolic integration – an attempt to get the best of both worlds by combining the learning capabilities of subsymbolic deep neural networks (which have already proven their worth) with the explainability of symbolic probabilistic modelling and programming (which hold significant promise). In addition to the technological considerations mentioned above, this area merits close attention from a cognitive psychology standpoint. As viewed by Daniel Kahneman, human thought can be construed as the interaction of two distinct but complementary systems: System 1 thinking is fast, unconscious, intuitive, unexplainable thinking, whereas System 2 thinking is slow, conscious, logical and explainable. System 1 provides for the effective performance of run-of-the-mill tasks and the recognition of familiar situations. In contrast, System 2 processes new information and makes sure we can adapt to new conditions by controlling and adapting the learning process of the first system. Systems of the first kind, as represented by neural networks, are already reaching Gartner’s so-called plateau of productivity in a variety of applications. But working applications based on systems of the second kind – not to mention hybrid neuro-symbolic systems which the most prominent industry players have only started to explore – have yet to be created.

This year, Russian researchers, entrepreneurs and government officials who are interested in developing artificial general intelligence have a unique opportunity to attend the first AGI-2020 international conference in St. Petersburg in late June 2020, where they can learn about all the latest developments in the field from the world’s leading experts.

From our partner RIAC

Continue Reading
Comments

Science & Technology

Rachel Lyons: Shaping the future of humanity in space

Published

on

image source: Space for Humanity

Rachel Lyons is the executive director at Space for Humanity. Space for Humanity is a non profit organisation in the US which is cultivating a movement to expand access to space for all humanity. Rachel is working towards making space exploration more inclusive and accessible to people worldwide. Space for Humanity is advocating space inclusivity in the US and is working with space experts, astronauts and other prominent people in the space sector to bring about change. In this conversation with Modern Diplomacy, Rachel discusses more about her experience working in the space advocacy sector.

What role is Space for Humanity playing in the future of the world?

This is a big question. If you think about our world, and the systems that we have in place – the types of people they favor, the types of activities that get prioritized, it becomes clear that these systems were built with foundational values of money and power being the highest priorities. If our values shift to things like the preservation of life, love, and wellbeing of humans and our planet — and this is what S4H is working to fundamentally address — the structures that are built on top of it will also begin to shift. This is what we are working to address. A shift in perspective that will ultimately cause behavior, relationships, and systems to change accordingly.

Why is advocacy important in the space exploration sector? What are some things you want to change about how we explore space? 

Advocacy is important because it influences public opinion and policy. Very often, when I share the importance of space exploration, people question why we are going to space when we face so many challenges on our own planet. The reality is, the technological advancements in space have impacted the lives of people globally in positive ways, and culturally the impacts have been massive (for example, the EarthRise Image of our planet from a distance from the Apollo era is said to have sparked the modern environmental movement). It is important for people to know, we go to space not because we choose it over earth, but because we love earth.

How can countries increase collaboration for space exploration?

This is a big question – I can talk about it from the individual’s perspective. If you are a young person, and you’re interested in space, by joining and supporting organizations like Students for the Exploration and Development of Space and the Space Generation Advisory Council, you can meet like minded people that are just beginning their career. Starting off early, networking, learning about what people are working on can open up collaborative opportunities exponentially for your entire career, no matter where that takes people.

Will all countries get an equal opportunity to. Go to space first when Space for Humanity’s citizen flights start?

Yes – that is our mission. And, there are some restrictions that we need to be realistic about. For example, countries that have more access to the internet are more likely to hear about S4H’s mission. Additionally, because of guidelines and safely with the flight providers, people must speak english in order to fly, so that limits access to others. And, it is extremely important to us for our mission to be as accessible as possible.

Why do you think it’s necessary for people to go to space and see Earth from above?

The perspective shift. Seeing the earth from above — the beauty, fragility, and interconnectedness of everything on it, can change a person for the reason of their lives. This cognitive shift is called the Overview Effect and it has been widely studied. Many astronauts return to earth with a new care for our planet and new care for people. They see how special and finite our existence is. They see the miraculousness and meaninglessness of it all at once. This perspective is essential, given the global nature of our greatest challenges, and what we are currently facing.

How is Space for Humanity planning to increase operations and advocacy across the globe?

Keep sharing our mission! The majority of our online content is totally free. We have people from 100+ countries that have applied to our program, follow us on social media, and attend our events. We are working to bring more and more people from all over onto our leadership board as well. We are so excited to keep expanding, and having efforts across the globe is an essential part of our mission.

How do you plan to share Space for Humanity’s vision with the world?

So many ways. We’re already done it via social media, launch parties, webinars, in person events, at conferences, public events, and more. We will continue doing this – sharing our mission IS our mission. Creating a perspective shift, on earth or off of it, IS our mission. In future years, when we sponsor astronauts to go to space, they will return to earth and commit themselves to sharing our mission. This is how we will continue amplifying the message.

Do you see other organisations like Space for Humanity starting worldwide? With a similar model?

There are similar organizations, like the Space Generation Advisory Council, that is a global network of space professionals.

Then there’s the Space Frontier foundation, that hosts a yearly conference and is a space advocacy organization.

The Planetary Society does a really great job of sharing space globally as well.

Virgin Galactic is a commercial space flight organization, where people will soon be able to purchase tickets to go to space.

These all exist and are doing great work, and there is no other organization like Space for Humanity. There is no organization that is working to start a movement using the spaceflight perspective, by sponsoring people from all over the world to go to space.

Continue Reading

Science & Technology

Antivirals, Spaceflights, EdTech, and Hyperloops: 20 Markets That Will Transform Economies

Published

on

As the world grapples with the socio-economic consequences of the COVID-19 pandemic, there is increasing demand to shape a new economy that addresses broader societal and environmental challenges while generating economic growth. To achieve this, the world needs to set an ambitious agenda of technological and socio-institutional innovations to pilot new markets that can help solve these challenges.

The World Economic Forum highlights 20 markets that could transform our economies. Some will rely particularly on advances in technology (e.g. broad-spectrum antivirals, spaceflights), while others will require radically new social and institutional set-ups (e.g. skills capital, water rights, quality credits). Others will emerge from a combination of both elements (e.g. data, genes and DNA sequences). Each of these markets has potential benefits in multiple dimensions. For example, they could help societies to protect and empower people (e.g. precision medicines and orphan drugs, EdTech and reskilling services), advance knowledge and understanding (e.g. artificial intelligence, spaceflights, satellite services), or protect the environment (e.g. greenhouse gas allowances, reforestation services, hydrogen).

“While protecting people remains the priority at present, now is also the time to plan a post-pandemic transformation of our economies. We must ensure that new economic activities do not only generate growth but also provide solutions to the problems that our societies are facing, said” says Saadia Zahidi, Managing Director, World Economic Forum. “The future of our economies, societies and the planet depend on developing these new, inclusive and sustainable markets.”

Creating these markets will require close collaboration between the public and the private sectors to:

  • Invent new products that can be sustainably produced
  • Nurture a set of companies to produce new products and bring them to market
  • Foster enough demand to sustain a commercially viable market
  • Establish clear standards that all actors can rely on and the market can converge on
  • Create alignment within society on how to value the new product
  • Develop the legal frameworks to identify, hold and exchange the new product
  • Build the necessary infrastructure to exchange, distribute and store the new product

Coalitions of actors at country and global level can come together to pursue the establishment of these conditions. For optimal societal outcomes, these markets should be designed around fairer and more sustainable ways of producing and distributing value. Examples include more collaboration between the public and the private sectors, innovative models to finance research and development, and designing the public sector’s risk-taking into the new ventures. Public institutions have a key role to play in catalysing public-private collaborations and create the systemic conditions for selected markets to emerge.

A preliminary mapping of countries’ potential for breakthrough technological and socio-institutional innovation indicates that those with advanced technological capabilities, strong social capital and future-oriented institutions are likely to succeed in developing a broader set of new markets. In particular, the Netherlands, Luxembourg, Denmark, Germany and Norway have the highest potential for socio-institutional innovation, while Japan, Germany, the United States, the Republic of Korea and France have the highest potential to generate breakthrough technological development.

Most advanced economies also score highly across both these dimensions. A number of high-income economies from the Middle East (Bahrain, Saudi Arabia, United Arab Emirates) and East Asia (Indonesia, Malaysia) as well as a few small island states (Barbados, Cyprus, Malta, Mauritius, Seychelles) and emerging African countries (Kenya and Namibia) can rely on significant levels of social capital and future orientation of policy-makers but do not yet have a mature technological system. A smaller group of advanced economies (Czech Republic, Israel, Italy, Japan, Spain) as well as the BRICs and other emerging economies (Hungary, Poland) present solid technological systems but need development in the social and institutional fabric to deliver these markets.

The disruptions brought by the COVID-19 pandemic provide an opportunity to pilot breakthrough technological and socio-institutional innovations that can grow into entire new markets. Success will ultimately depend on how well multistakeholder actors work together to create the necessary conditions for a number of key new markets to emerge that will help make economies more inclusive and sustainable. Existing market structures are not neutral; high levels of concentration and market power in adjacent industries to the new markets might slow down or even curb the establishment of such new markets.

Continue Reading

Science & Technology

Light at the end of the tunnel: New technologies to fight the COVID-19 on transport

Published

on

Disinfection robots, thermometer robots, smart tunnels, automatic passenger counting, powerful ultraviolet lamps and other examples of how new technologies reshaped public transport amid the COVID-19 outbreak.

The coronavirus pandemic has led to significant changes in many areas of life in just a few months. As the coronavirus continued to spread around the world, governments in several countries took measures to restrict movement, and people themselves tried to avoid traveling on public transport. The demand for the services of transport operators has dropped drastically. So, according to the Moovit Public Transit Index, passenger traffic in public transport on April 15, 2020 decreased in Israel by 92.1%, in Rome – by 89.2%, in Madrid – by 88.1%, in New York-by 74.8% and has not yet recovered. City residents are afraid to use public transport actively again, and their fears are fully justified. High daily passenger traffic and high frequency of contact between passengers make public transport an ideal environment for the spread of infections. The problem of fighting the spread of infections while maintaining normal life activity is particularly acute for large cities, such as Moscow or Beijing, where daily passenger traffic reaches 19.4 and 12.3 million passengers respectively. The average density of passengers on a bus or in a traincar at the same time ranges from 2 to 5 people per square meter, while, according to World Health Organization (WHO) recommendations, in order to comply with safety standards, passengers must maintain a social distance of 1.5 meters. Furthermore, virus particles can remain for a long time on public surfaces inside a bus or a traincar. Handrails on public transport are usually made of plastic, on which the coronavirus can remain up to 3 days, according to the New England Journal of Medicine. By touching them passengers increase the risk of contagion.

The key task for transport operators is to make the usage of public transport safe. To help them solve this problem came technology -all kinds of robots are widely used among innovations. With their help, it is possible to carry out disinfection effectively and safely without the involvement of staff. The Hong Kong Metro, also known as the Mass Transit Railway (MTR), together with the biotechnology company Avalon Biomedical Management Limited, has developed a disinfection robot that can disinfect even the most inaccessible places of traincars and stations. In addition to disinfection, robots can cope with more complex tasks. So, in Ningbo Lishe International Airport was tested a 5G-supporting robot-thermometer, which can measure temperature at a distance of 5 meters up to 10 people simultaneously and also identify those who are not wearing a face mask. Another innovation in many transport operators is the sanitary gate. According to Giulio Barbieri, one of the manufacturers, this is a “a tested, safe, and effective method to sanitize people and objects in just 5 seconds, killing up to 99% of any pathogenic microbes on the surfaces, including COVID-19”For example, the technology was tested in the Moscow and Dubai metros. In Moscow the clothes of the employees entering the depot were processed using a disinfection tunnel; at the same time, the territory was manually disinfected, so that the entire depot was safer for the staff.

The process of digitalization of ticket systems, which began long before the pandemic, also had a positive effect. Thanks to the competent actions of transport operators, the number of contactless payments in public transport around the world increased by 187% in the period from April to June, as evidenced by a report from Visa. Following WHO recommendations, many transport operators have made it mandatory to wear masks and maintain social distance on public transport. A number of digital technologies have been developed to comply with these rules. In the Beijing metro, compliance with a mask regime is controlled by cameras with a facial recognition system that can identify people. In addition, in the Panama Metro, observance of social distance is monitored by sensors which determine the degree of capacity of train cars. The technology called Mastria, which aggregates information from train weight sensors, ticket machines, signalling, management systems, CCTV and mobile networks for the Panama metro was developed by Alstom (a french manufacturer specializing in the production of infrastructure for rail transport) and installed almost a year ago. In just three months, thanks to artificial neural networks, it was possible to reduce average waiting times at stations by 12%. This development became particularly relevant during the pandemic. The Moscow metro is planning to introduce a similar technology. To maintain the social distance digital displays with colored indicators that reflect the level of capacity of subway cars will be installed. In the Moscow metro a new generation of traincars with an automatic air disinfection system built into climate control systems helped to reduce the risk of infection. It makes it possible to disinfect the air without disrupting the train schedule and attracting employees. The Moscow metro rolling stock consists of more than 50% of train cars with built-in UV lamps, and this percentage is constantly growing. After evaluating the effectiveness of using UV lamps to disinfect public transport, the transport operator MTA New York City Transit, together with Columbia University, launched a pilot project worth 1 million dollars on the use of disinfecting lamps. During the first phase of the project, 150 autonomous lamps were purchased and installed to decontaminate wagons, stations and buses in New York, during the second phase it is planned to install equipment in commuter rails. To carry out disinfection measures, the New York City Subway took unprecedented measures – the closure of the subway from 1 to 5 a.m. daily.

The use of robots, disinfection tunnels, digital technologies, ultraviolet lamps, and intensive work of staff – all this helped to reduce the risk of the spread of coronavirus in public transport and made a significant contribution to fighting the global problem. According to the coronavirus distribution model, developed by Imperial College London at the beginning of the pandemic, if no action had been taken by mid-March there would have been over 500,000 deaths from COVID in the UK and over 2.2 million in the USA. At the moment, in the middle of October, there are about 43,000 deaths in the UK and about 214,000 in the USA. Of course, these are high rates, but they could have been much higher if the necessary measures were not taken in time. Technological innovations already available today will continue to be used, which will make the stay of passengers on public transport more comfortable and safer, reducing the risk of the spread of any infectious disease, especially during the flu and cold seasons.

Continue Reading

Publications

Latest

Trending