Enhanced human intelligence (AI) vs AII and CII. Reach all who used to?

Published on June 17, 2016

Enhanced human intelligence (AI) vs AII and CII. Reach all who used to?

    In the light of recent trends aimed at the all-round development of the power of AI, for some reason, most followers of this unconditionally promising direction are consciously or not consciously or not, but do not consider artificially improved human intelligence (AI) as an alternative. The question is far from idle and some researchers are convinced of the problem: a technologically “improved” human brain will be able to compete, and in many cases surpass the AI ​​in its capabilities. How far is this opinion far from the truth, and what is this, actually: enhanced / improved human intellect?

    image

    I would like to continue the discussion, based on the opinion of the problem researcher, futurologist, popular blogger and co-founder of the Singularity Summit forum Mikhail Anisimov, who responded to thematic questions in this regard to the correspondent of the Gizmodo Internet portal .

    The dilemma in question looks something like this: man, as a species, has developed his brain and ability to think over thousands of years. And on this thorny path in trying to overcome their cognitive limitations, people have used and continue to use all possible means. Why are attempts to re-develop general artificial intelligence (AII) or even artificial super-intelligence (ICI) to take priority over attempts to unlock the potential of the human brain as fully as possible? After all, so far all attempts to create artificial intelligence with an artificial neural network somehow come down to more or less successful attempts to repeat the processes occurring in human neural networks. On the other hand, is it not easier and faster to reach the last goal and to endow a person with improved / enhanced intelligence (CI), using the capabilities of modern genetics, cybernetics, external interfaces and other factors stimulating the achievement of the result? Further conversation in the question-answer mode will take place in this context.

    Q. When we talk about enhancing (improving) the intellect, what should we really have in mind? Is it about creating new Einsteins? Or does the question actually look much bigger?

    A. The real goal in the development of IA is rather the creation of “super-Einsteins”, people by orders of magnitude and qualitatively smarter than any person who ever lived on Earth. And to achieve this goal, we will have to take a few key steps.

    Step one: create a direct neural connection with information. Conventionally, we can assume that we are talking about some "telepathic search engine Google."

    Step two: the development and improvement of the brain-computer interface, enhancing the work of the brain department responsible for visual perception. This will not only enhance our spatial visualization capabilities, and, as a result, significantly improve the efficiency of information processing. Try to imagine that you are able, almost instantly and without perceptible voltage, to create in your brain some kind of complex detailed scheme of an object or to remember once seen in the smallest details. Similar augmentations should be required for other areas of the sensory cortex, which are responsible for touch and hearing.

    Step three: augmentation of the prefrontal cortex. Researchers of IA are often comparing this area with the Holy Grail, because it is here that the efficiency of perception increases, and sensory information is combined into concepts.

    The final “product” of such a multi-step transformation will be a person who will be capable of many incredible intellectual feats from the point of view of an ordinary person. Among them, for example, will be the ability to control other people by the power of reason, a clear understanding of development trends and an accurate forecast of results in financial markets, the production of ideas that, when implemented at the level of technology, will be able to change the world almost “overnight”. Of course, now all this is balancing on the verge of paranoia. But wouldn't it be about the same associations that would have caused our current achievements in people of the Bronze Age? But today to realize this opportunity is really real.

    AT.What are the prospects for the development of AI, for example, in the coming decades? Do you really believe that the human brain can compete with the machine?

    In reality, it is difficult to imagine a competitive alternative to the human brain today. The brain is a “product of evolution” by the
    age of seven million years. During all this time, he was exposed to a fine natural setting and optimization, which, taking into account the existing restrictions, were very successful. Therefore, attempts to intensively disperse the capabilities of the brain, as a rule, result in the failure of its mechanisms. The simplest confirmation of this is the effects that develop with amphetamine addiction.


    Limitless Movie Trailer

    Available chemicals and their compounds today do not have the capacity to transform human abilities in the direction of the development of AI. I would like to express the opinion that all the arguments in favor of the effectiveness of the existing “brain accelerators” actually look more than debatable and doubtful. Achieving real breakthroughs will require the use of invasive technologies and the introduction of brain implants that physically interact with millions of neurons. This need, in turn, will require the presence of millions of microscopic electrodes and the creation of a unique control system that allows you to synchronize all the individual elements. At the same time, the most advanced development of the computer interface — the brain will be able to boast of a force of 1000 and a few such connections. This means that if you follow the path of development of IA, then the scale of modern developments should be increased at least 1000 times. Only in this case it will be possible to talk about a qualitative fundamental shift. Even if we assume that the development of technology will occur exponentially, then to achieve the "profitable" results will take at least 15 - 20 years.

    The development of IA is impossible without the development of nanoproducts And today, the results of the work of leading experts working on the brain-computer interfaces, including, for example, Ed Boyden from MIT, are entirely dependent on the progress made in this direction. Given the practical lack of production technologies that actually work at the level of atomic precision, nano-level self-assembly becomes the most likely way of developing brain-computer interfaces with millions of electrodes.

    Q. To what extent can the likelihood of adverse psychological effects develop in people with significantly improved intelligence? And is it appropriate in this context to even mention them as people?

    One of the most likely side effects will be madness. The human brain is an unprecedentedly precisely calibrated machine. And any fundamental change in the mode of its work will inevitably lead to what we used to call "madness." On the other hand, there are many different types of madness, significantly more than the behavioral patterns associated in our understanding with common sense. The fact that from the point of view of the external environment is generally accepted as a deviation from the norm for the owner of such a deviation is quite understandable from the point of view of his logic.

    On the other hand, even while maintaining full responsibility, various side effects can occur: convulsions, the state of information overload, a feeling of complete and deep alienation from society. People with high IQ, as a rule, feel a certain estrangement from the world, and if you imagine the level of opportunities for a person with an ID who realizes that they are equal to the level of development, possibly on a planet, then the level of alienation will increase many times.

    Most of the very smart people are not too eager to communicate and have fun, like, say, an American scientist Richard Feynman. As Hemingway once admitted: "An intelligent person is sometimes forced to get drunk in order to make contact with fools." And what will happen when alcohol is no longer able to maintain camaraderie and mutual affection? The total absence of an alternative in the ability to show a feeling of empathy can lead to the loss of the latter, and as a result - to mental disorders.

    Q. So what comes first? Strengthened or artificial intelligence?

    A. It is very difficult to answer this question unequivocally. There is a great temptation to wish for enhanced intelligence to appear before the AI. Much of this is due to works of art and games, in which the owner of the UI becomes the main character. But the wishes are few: despite certain mental priorities in the development of IA, the work on translating an idea into life is technologically unprecedentedly complex. My personal opinion: AI will come to us earlier, because its development is cheaper and cleaner.

    Both areas of research are extremely difficult to implement. Most likely, their arrival in the incarnation in which we imagine them will not take place before the 2060s, 2070s, and perhaps this will happen even later. But, ultimately, both directions of development will achieve their goals.

    AT.В чем вы видите преимущества и недостатки этих двух подходов в развитии интеллекта?

    A. The key advantage of the development path of AI is the disproportionately cheaper and more accessible development processes. On paper and in AI code has already been developed. It remains only to correctly assemble it. The most significant research of IA, in turn, is currently banned at the legislative level. This is due to the fact that serious UI research will require the use of the most complicated experiments on neurosurgery using brain implants. Brain implants, in turn, can fail, cause seizures, insanity and, in its extreme manifestation, even death. Here, researchers should be aware: a qualitative change in the abilities of human intelligence is no longer just a matter of a few drunk pills in the hope of gaining a couple of colorful superpowers for a while, it is a question,

    Most of the research in this area is under tight control and eats a lot of money. All experiments involving animals have always been and remain expensive. So, Theodore Berger developed an implant for the hippocampus to restore and improve memory performance over the years. In 2004, the first tests with the participation of living tissue were held, but since then information appears extremely sparingly and very rarely. Every few years, of course, some notes appear in the news, but the end result seems just as far from implementation as it was decades ago. And here, in comparison, the level of development that we have managed to achieve in the development of artificial intelligence cannot but impress!

    AT.In terms of predictability and control, does the AI ​​look safer than AI? What is more important to create first, AI or super power AI?

    Enhanced intelligence seems to be more unpredictable and uncontrollable in comparison with super-power AI. And in the long run, given the unpredictability factor, it actually looks even more dangerous. I recently wrote an article reflecting on global political transformation, the trigger of which is unlimited power, concentrated in the hands of a group of people endowed with IT or having exclusive access to molecular production. In it, I introduced the term “Maximilian”, meaning “best of the best”, to describe a very influential leader who himself uses advanced intelligence technology to elevate himself above all others.

    image
    Cognitively Improved Reginald Barkley from Star Trek: The Next Generation in the episode "A nth grade"

    The problem is that in the case of UI, a person is endowed with outstanding abilities, and a person, even with improved intelligence, remains imperfect. In this context, it is quite possible that people who deny any moral principles of society, having received an ID, will be mired in hedonism or set other, more specific and ominous goals, such as genocide.

    Artificial superintelligence, in turn, can be created from scratch and simply enter into it the ability to follow a set of your own motivations within those that will be incorporated into it, with an emphasis on good start and stability, developed through positive feedback.

    Researchers of the dilemma often ask themselves the question: “Will the CII reject these motivations?”. No, he will not, as these motivations will form the core of his values, if, of course, they are correctly programmed. There will be no second bottom, a “ghost in the car,” stimulating the AI ​​to ignore or abandon the pre-programmed motives.

    In his work "The Will of Superintelligence" philosopher Nick Bostrom conducts an amazing analysis. The conclusion looks quite optimistic: mercenary goals cannot appear by themselves if the set of disinterested goals are initially and fundamentally embedded in the basic AI motivation system and if the very essence of the existence of a machine is to serve these goals. Evolution has clearly demonstrated that mercenary goals can play a decisive role in survival, especially given the evolutionary constructive limitations of living organisms. But this does not limit our ability to program unselfish agents from scratch.

    Q. What are the difficulties (technological, medical, or ethical) to overcome on the path of development?

    A. The key issue is the development of efficient manufacturing technologies. Now we are not even close to this. Another caveat is finding out what each individual neuron is specifically responsible for, and determining the exact location of these neurons in the brain of a particular person. And this task still seems infinitely far from implementation.

    Finally, we need to create an efficient algorithm capable of quickly assessing the viability of numerous theories about the functioning of the brain. Ed Boyden defines it as “high-throughput screening of neural networks.” The ideal option would be to create a kind of experimental "prototype" of a human being who does not possess consciousness. This would allow to conduct experiments "in pure form" without regard for moral and ethical aspects. But, ironically, I foresee major problems in the implementation of such a concept on the part of most ethics committees.

    If this possibility is not realized, as an alternative, you will have to look for a way to create a precision-precise model of the human brain, which in terms of the implementation of the final goal, will many times exceed the results achieved in today's “brain modeling” projects. Oxford analysis does not look too optimistic: we will be able to get closer to our intended goal by 2080.



    That's all, with you there was a simple service for choosing sophisticated Dronk.Ru equipment  . Do not forget to subscribe to our  blog , there will be many more interesting ...

    image

    Why do online stores give money for purchases? How to find the best deal in China and return some of the money for the purchase? Return your money - choose a cashback service for Aliexpress