Specialists in the field of human-level artificial intelligence are currently facing the necessity of dealing with not only scientific, but also ethical issues. What are the challenges that scientists are confronted with? How can we safeguard humanity from potential threats? Professor Aleksey Samsonovich of the Institute of Intellectual Cybernetic Systems at the National Research Nuclear University MEPhI has shared his stance on these issues and more.
Sputnik: Aleksey Vladimirovich, we are now talking within the framework of the international conference titled “Biologically Inspired Cognitive Architectures”, BICA 2019. Which talking points are now taking centre stage in light of the creation of human-level artificial intelligence?
Aleksey Samsonovich: The conference has let us again come to realise the key challenges in the creation of next-gen artificial intelligence, which is set to become man’s equal partner and friend.
To start with, it may be considered a challenge to create human-level emotional intelligence capable of exploiting emotions when making decisions and generating behaviour.
More specifically, this ability will enable the machine to set up long-term social connections similar to relationships between people that are close to each other. It will also enable it to explain its moves and decisions, to pick goals and a way of learning.
Emotional intellect is a key to solve all the other challenges. It is not limited only to an ability to detect and express emotions, which most researchers working in the field are preoccupied with. Although both processes are critical, it is much more essential and difficult to understand how the two are interconnected – something that emotional intellect actually provides.
Secondly, an intelligent agent should grasp the context of what is going on. It means that based on the available information, he should be able to generate and constantly update the global scenario of the past and present, determine his own role in it, and then, based on this, set clear-cut goals and ways to achieve them. In scientific literature, this very capability is referred to as “narrative intellect” (one shouldn’t confuse it with programs that work in line with the scenario or narrative, pre-written by a person).
Sputnik: Is there also a third challenge?
Aleksey Samsonovich: This is the course to take to teach artificial intelligence, which, like humans, will work out the system of its values that will motivate it throughout learning, as well as to set goals, a roadmap for learning, along with choosing methods (such as experiments to check hypotheses, search for available resources and questions to humans) and, pushed by its own initiative, use these means to achieve its goals. Also, the machine should be capable of developing from a children’s level to an adult level and beyond, to super-intellect.
This third branch can be called human-level learning intelligence, in order to differentiate it from “deep learning” and other forms of machine learning that are popular today.
As far as I can see, human-level learning intelligence will be rendered possible based on emotional and narrative intelligence.
Sputnik: What kind of research are you currently conducting? How is your work going on virtual human assistants and emotional intelligence modelling?
Aleksey Samsonovich: We are modelling human emotional intelligence seeking to develop socially compatible virtual assistants based on our models.
We are looking into a great variety of concrete goals and branches of research. For instance, this is about the creation of a virtual partner in a video game, a virtual coordinator, or moderator; a virtual creative assistant to a composer, a choreographer, designer, solver of insight tasks; or a virtual listener (soon it will be a virtual interlocutor), as well as a virtual pet that will be comparable to a real one in terms of its social-emotional characteristics.
It is rather challenging to share the results in detail here, but they are available in open-source study papers. Despite the vast nature of our attempts and successes, we are in need of a breakthrough in one direction: in the explanation of the practical value of the general model of human emotional intelligence embodied in an intelligent agent.
Not long ago, our laboratory, BICA Lab, at the department of cybernetics at the National Research Nuclear University MEPhI, was empowered with brand-new research techniques, including an experimental platform based on virtual and mixed reality (with monitoring the direction that the subject’s eyes take), as well as electromyography and automated facial expression analysis, which enable us to register the emotions of the test subjects immersed in virtual surroundings.
A big share of our research is now conducted using these techniques. Some experiments also involve electroencephalography and functional magnetic resonance tomography (functional MRT).
Our immediate super-task is the improvement and empirical validation of social-emotional cognitive architecture “eBICA” and the development of catchy demonstrations based on it. This research is being conducted at the National Research Nuclear University MEPhI and financed by the Russian Science Foundation (grant number 18-11-00336).
Sputnik: Elon Musk’s start-up has recently announced research that may at some point transform humans into cyborgs capable of managing computer programmes with the power of thought. How promising do you think this branch of study is?
Aleksey Samsonovich: I think artificial intelligence should become, on the one hand, a full-fledged partner and friend for humans, and on the other hand, an extension of normal, or replacement of lost human capabilities, including means based on a direct interface with the human brain.
In this case, humans would literally be able to use a computer or microchip as part of their brains. Both are meant for good and must be fully controlled by their creator. I wish Elon Musk all possible success.
Sputnik: How great is the necessity to develop the ability to oppose artificial intelligence?
Aleksey Samsonovich: I think, in our era, humans will be in full control of AI, at least due to its dependence on human activity when it comes to its maintenance and provision. If there is anyone capable of making it rise against humans, these are only humans themselves. Then the question arises about the means to oppose those humans.
It is no secret that AI is mostly needed by the military. Today, like in the past, new assault and defensive methods and techniques are being created, but the distinguishing feature of our time is that both contain AI. In this respect, the importance of the development of assault means is really huge and some countries prioritise it from a financial point of view.
Sputnik: What awaits us after the creation of AI that surpasses human intelligence? Will we co-exist with machines or merge into some single entity?
Aleksey Samsonovich: AI will become (and is already becoming) an indispensable part of human civilisation. However, there could be two ways that this could unfold: AI potentially inheriting our values, hopes, and ideals, or replacing them with “machine values” that will spread not only on to machinery, but also human society. It depends on us which of the two scenarios will be fulfilled.
Sputnik: What does it mean to remain humans in the AI era?
Aleksey Samsonovich: To remain humans in the AI era means to take the first course of action.