Home > Digitalizacija > Byron Reese: The Fourth Age; Smart Robots, Conscious Computers, and the Future of Humanity

Byron Reese: The Fourth Age; Smart Robots, Conscious Computers, and the Future of Humanity

Things have only changed three times in human history. Each time it was due to technology. Not just a single technology, but groups of interrelated technologies that changed us in fundamental and permanent, even biological ways.

First change was due to invention of fire and language. Fire allowed us cooking. Cooking allowed us to vastly increase our calories intake. We used new energy to grow our brains to unprecedented complexity. We actually use 20% of all the calories we consume just to support our advanced brains. More complex brains lead to invention of language. Language was a great leap that historian Will Durant says, “made us all human”. Language enabled us to cooperate with each other, which is our singular abilities as a species. Another gift of language is stories. Stories are central to humanity, for they gave form to human imagination, which is the first requisite for progress.

Second change come with agriculture and cities. Agriculture like language, is also a technology, and like language, agriculture brought about a slew of other advances. One of those was the city. Cities promoted commerce and the exchange of ideas, but they also made us stationary. Second one was division of labor. Along with trading and technological advances, the division of labor is one of the only three “free lunches” in conventional economic theory; that is, one of the ways overall wealth can be increased without anyone’s having to work harder. Agriculture and privately-owned land ended the economic equality of the First Age. The natural inequality of ability, birth, and luck led to unequal accumulations of wealth. Imagination that come from brain development is first requisite for progress. Idea of future, that come with agriculture was second.

Third change come with writing and wheels. Money also appeared in this change. With writing, the wheel, and money all coming on the scene concurrently, the basic ingredients needed to make nation-state and empires were in place.

The fourth age is and will be influenced by robots and AI. Some historians say that Leonardo da Vinci was the last man to know everything. Since then knowledge grow too fast for humans to catch it all. Technology advance in incremental improvements on work done by others. With development of scientific method and use of imagination, sense of time and writing, expansion of knowledge was possible. But we also need low-cost way to distribute the knowledge, widespread literacy, the rule of law, non-confiscatory taxation, individual liberty, and a culture that promoted risk taking. The invention of printing press and its widespread use, increased literacy and the free flow of information.

Much of the universe is clearly computational. Hurricanes and DNA are computational, as are snowflakes and sand dunes. The story begins in 1821 in London with Charles Babbage. It was a stroke of genius on Babbage’s part to realize that if stream could make cogs, it could also compute logs. He surmised that steam could power computing machines. Alan Turning in 1936 first described what we now call a Turing machine. The simple imaginary device that can solve an enormous range of computational problems. Almost all of them. Then John Von Neumann developed von Neumann architecture for computers. While Turing machine is purely theoretical, designed to frame the questions of what computer can do, the von Neumann architecture is about how to build actual computer. In 1949 Claude Shannon wrote a paper how to program a computer to play chess. Babbage realized machines can do math. Turing added that they could also run programs. Von Neumann figured out how to build a hardware, and Shannon showed how the software could do things that didn’t at first look like math problems.

If we want to understand what AI is capable for, we need to understand big questions.

  • One of them is what is the composition of universe? First school of thought is that everything in universe is composed of a single substance, atoms. This is known as monism. Physics explains chemistry, chemistry explains biology, biology explains life, life explains consciousness. Only the atoms and the void are real. The other school is dualism. There are atoms and there is something else. If you think that experiencing something is different than knowing something, then you are dualist.
  • Second is what are we exactly? Three possible answers: machines, animals or humans. We may be machines, but we are the most amazing and powerful machines on the planet. Second idea is that we are animals. Life makes us different from machines. Maybe our bodies are machines, but “we” are animals that inhabit those machines. If we are humans, that is based on idea that there is something that separated us from machines and animals. At the end of the day “you” is about brain puts all senses together into a single show and lets only one part speak at a time. Those two things together give illusion of a “you”. The second option is that your “self” is an emergent mind. Emergence at its simplest is when group of things interact with each other, and through that interaction, the collective whole gains characteristics that none of the individual things has. The final option is that your “self” is your soul.

Around 1955 AI began as real science with John McCarthy, using term artificial intelligence, he later wanted to change it into computational intelligence. Narrow AI is the ability for a computer to solve a specific kind of problem or perform a specific task. The other kind of AI is referred to by three different names: general AI, strong AI or artificial general intelligence (AGI).

There are three different approaches to how to build AI:

  • Classic AI – thinking through all the factors and building a model that makes those factors, weights them accordingly and makes suggestion.
  • Expert system – you write down every rule. You then arrange those rules in a way that someone can enter any relevant variables and then system will make a suggestion based on those rules.
  • Machine learning – you take all the data and you get computer to build rules.

Large data sets, colloquially called “big data”, combined with powerful computers and cunning algorithms have largely been responsible for the current renewed excitement about AI and its recent successes. Making AGI that can solve problems that it has not been programmed to solve may be profoundly different from making an AI that can solve a narrowly defined problem. A natural housing for an AI is within a machine that gives it an ability to interact with the physical world – a robot.

Human fears about robots are wide. The primary fear is that a robot will compete with us in the job market and that we will lose to them. A massive displacement of this sort would represent a dramatic shift of economic power away from labor and toward those who own the robots. Another concern is that we become permanently sedentary and our brains atrophy as well. Some fear that emotional attachment to robots will lead to human detangelement. And the ultimate fear is about robot uprising.

Some technical limitations to widespread of AI are based on things AI can do. AI can’t contextualize. Context is derived from the differences in a series of still images, and there are innumerable possible permutations of that. Training a computer to make those intuitive leaps is quite a difficult. AI has problems with transfer learning. Transfer learning comes natural to humans, we don’t understand why it happens, so we will have hard time teaching machines how to do it. On other hand, if we have lots of data, that is the area where AI will excel. And AI is very good at doing one thing at the time. You don’t ask it to generalize contextualize, be creative, or anything else. Just teach it that one thing. Ironically, the more you specialize, the easier it is to be replaced by machine.

Robots outside factories are still novelties, facing a laundry list of challenges, including locomotion, sensing and the manipulation of environment. The first challenge robots have is figuring out where they are. This is both sensing challenge and an AI challenge. The way robots work is through thinking process, they need to dissect every action to the most minuscule detail. Machines are, on the whole, more reliable than people in what they do. However, generally speaking, machine failures have a larger potential to cascade.

When we talk about job loss, whether the machine is pushing around atoms or bits doesn’t matter too much when we are talking about employment. We can talk about some possibilities.

  • Possibility one – machines take all the jobs. In order for this to happen we can look at some assumptions and check if their validity.
    • Humans are machines
    • Since humans are machines, we can build a mechanical one.
    • Mechanical humans would have the full range of our mental abilities, including creativity.
    • The conscious machine would want to do our dirty work.
    • Even if they don’t want to, we will compel it to do it, creating de facto mechanical slaves.
    • It would be economically practical to build such mechanical humans.
    • Machine would become so inexpensive or efficient that they would be cheaper to deploy than human labor.
    • The programming cost to teach the machine a new skill plus the cost of running the machine will always be less than the labor cost of paying human to do it.
    • Humans lack the ability to find other tasks that machines cannot do.

If all nine of these assumptions play out, then virtually all paid employment to humans will vanish and most institutions in society will have to be rethought.

  • Possibility two – the machines take some of the jobs.
    • Machines and technology cause a net loss of jobs. When you can introduce efficiencies in an industry, you lower costs or increase quality. Lower costs and increased quality invite higher production, which creates jobs. So that could challenge this assumption. Second challenge to this assumption is what kind of jobs will machine take over. Since they will take repetitive jobs, this can actually improve status of jobs and our long-term goal as a species should be to build machines that will do boring jobs, so that people can do jobs that only people can do. Boredom was only invented once we had factories. It was first used in Dickens novel Bleak House.
    • Too many jobs will be destroyed too quickly.
    • Not enough job will be created quickly enough.
    • Low-skill workers will be the first to go.
    • There won’t be enough jobs for these workers in the future. This is not always the case. If we remember Moravec paradox, that was names after Hans Moravec, who noted that it is easier to do hard, brainy things with computers that “easy” things. And even when we look at creations of new jobs at the top of chain, when this happens, everybody gets a promotion. We all want new jobs to be created, so that everyone gets a chance to move up a rung on the ladder of success.
  • Possibility three – the machines take none of the jobs.
    • Reasons why machines could not take jobs:
      • There exists a range of jobs that machines will not be able to do.
      • There are unlimited number of jobs, because any time people figure out how to sell a service they offer or a product they make, they just created a job. In reality there will be unlimited jobs indefinitely, for they are created by humans mind, not by outside force.
      • Even if robots take all the jobs and the earnings are distributed evenly among population, so that no one needed to work, most people would still choose to have job of some kind. Keynes says that while human want may be insatiable, humans needs are essentially fixed. And given that we will have so much more money, than we need to maintain life, he believes we will “prefer to devote our further energies to non-economic purposes”. We can actually stay in “work pattern” just out of habit or because we will not find happiness or meaning in leisure. As long as you want more income, you will likely find a way to use your skills to add value somewhere, and that action is what creates a job. Maybe we are not so much Homo sapiens, the reasoning man, as we are Homo dissatisfactus.

When we talk about robots taking over, we need to estimate if there are such jobs as robot proof jobs. Some examples of stable jobs regardless of robot capabilities.

  • Jobs robots can do, but probably never will.
  • Jobs we won’t want robots to do.
  • Unpredictable jobs – like CEO, whose description of job could be, come in every morning and fix everything that is broke and seize every opportunity that you can.
  • Jobs done on-site.
  • Jobs that require creativity or abstract thinking.
  • Jobs no one has thought of yet.

Income inequality is one of the main questions in thinking about AI influenced future. Technology can, and frequently does, raise corporate profits without raising wages. This is why the stock market can go up while incomes remain flat. Productivity gains of technology benefit higher-income people directly while helping low-income earners only indirectly. Technology requires investment and investment requires wealth.

With rise of non-labor created values, the idea of UBI (universal basic income) is getting popular again. Argument for UBI are old, argument of basic human rights it that no one should earn a living, second one is that there is a set of shared social resources, like knowledge, institutions, language, money, law, that can legally be seen as owned by everybody, so basically all wealth was made using this resources, so everyone has equal right to claim all that money. But even UBI will not solve problem of people losing jobs and losing their purpose too.

Robots with AI can be used in war. There are three reasons weapons are compelling to them. First, they will be more effective at their missions than human soldiers. Second, there is fear that potential adversaries are developing these technologies. And third, they will reduce the humans casualties of the militaries that deploy them.

When we estimate development of AGI – general artificial intelligence, we try to estimate its capacity towards our brains. But surprisingly we know very little about inner workings of our brain. That is because we had little opportunity to study them anatomically while working. We don’t know how we encode information in the brain or how we retrieve it. To try to capture the nature of human intelligence or human choice is a colossal problem way beyond the limits of contemporary science. The brain, however has some idiosyncrasies. To begin with, it has hundreds of cognitive biases. Plus, it makes some irrational decisions. Can this be our advantage against computers. Among all the objects in the known universe, the human brains are in a category all by itself.

The brain is an organ made up of three pounds of goo that behaves in a mechanistic way. The mind is all the metal stuff that you can do that seems way more difficult than this goo could possibly pull off. The mind is the source of emotion, imagination, judgement, intelligence, volition, and will. There are three possible explanation for the self, and thus by extension the mind.

  • Your self is a trick of brain – this view believes that brain is all that is and all the abilities that we don’t understand like creativity or emotions are all just brain functions.
  • Your self is and emergent property of the brain. It is when whole of something takes attributes and abilities that no individual part has.
  • Your soul is an aspect of you that exist outside the laws of physics.

To earn the title of AGI, the aspiring technology would have to exhibit the entire range of the various types of intelligence that humans have, such as social and emotional intelligence, the ability to ponder the past and the future, as well as creativity and true originality. AGI could actually help us understand what we really are. When debates about AGI took question of “understanding” something, we can use example of Chinese room argument. It is about message in Chinese coming to library and person inside the room doesn’t know what is says (he doesn’t speak Chinese), so he looks for first sign and find it in a book, and clues for next sign. He retrieves all the answers, put them into message and send it back through doors, actually replying correctly, but no knowing what the reply also is. Question here is does a person understand Chinese. So, computers don’t understand things, they just follow set of instructions. Different views on AGI is more about believing different things, not too much about technology.

Evolution of technology will be about moving from electrification century ago to cognization now with AI. IN regard to AGI it is like with heart surgery. One good outcome and a hundred things that could go wrong. Some fears include fear of machines having different goals that us. Second is that they will grow their power too quickly. Some are actually advocating that we should leave computer evolution to lead to computer supremacy (Dr. Hugo de Garis).

When we think about timeline, there is disagreement about how complex intelligence is. Secondly there is not even consensus about how to build AGI.

Another big challenge in AI is its attitude towards ethics. In 1942 Isaac Asimov coined three laws of robots behavior. Future of Life Institute describe challenge in this way: “Some AI systems do generate decisions based on consequences, but consequences are not all there is to morality. Moral judgments are also affected by rights (such as privacy), roles (such as in families), past actions (such as promises), motives and intentions, and other morally relevant features. These diverse factors have not yet been built into AI systems. Internet itself brings some threats, because you only need to change laws to move it from information infrastructure to surveillance one.

Simulation and potential that we are already living in one is also one of questions connected with building AGI.

One aspect that we need to address in order to build AGI is can we build conscious computer. We need to explore two concepts:

  • Sentience – sentient means the ability to feel or “sense” things. Can machine sense things?
  • Free will – in your daily life, you have experiences, which are the basis of consciousness and you make choices, which is free will. Generally speaking dualist believe in free will whereas monist don’t. In 1990 Dan Wegner and Thalia Wheatley proposed revolutionary idea. Instead of the traditional order of the sequence – a person decide to do something and then it happens – they maintained that things in the brain actually run backward from that. First, the theory goes, you do something, then you tell yourself later that you decided to do it. Samuel Johnson captured the conflict of free will quite well by saying that theory holds that we do not have it and experience holds that we do.

All this leads us to concept of consciousness. It is experience of being you. It is single most important aspect of our existence. It is different from sentience. It is also different from intelligence, since intelligence is about reasoning, consciousness is a feeling of experiencing. Daniel C. Dennet suggest consciousness might have something to do with a tension between competition and cooperation. While consciousness is subjective experience, that subjective experience either objectively happens or not. When looking at the levels of consciousness, we can see that some things are a little conscious and other a lot.

One of the implications of believing in consciousness computers are their rights. Nietzsche believed that you only have rights that you can take. In these light computers could only have rights they could size. Second theory is based on rights being created by consensus. And another theory of rights holds at least some of them as inalienable.

Out of this we could ask our self can computer become conscious? Is the computer of the future a thing or a being? Some sees future as merger between humans and machines and that action will secure peace as with marriages between royal families in previous times. There are two possible roads to machine consciousness. One that they achieve it them self and second, that they are empty vessels into which consciousness is uploaded.

There are eight theories about how consciousness came about:

  • Weak Emergence – Ray Kurzweil believes that consciousness is emergent property of a complex physical system. Weak emergence is an outcome that’s unexpected but explainable (at least in theory). Can machine achieve it, probably yes.
  • Strong Emergence – strong emergence says that the emergent property is completely inexplicable as simply the interaction of the parts. If that is true, we are still not sure if machine can achieve it.
  • Physical Property of Matter – consciousness is a physical property of matter. Once we understand matter better, we will understand consciousness. Can machine achieve it, probably yes.
  • Quantum Phenomenon – another variant of “physical property of matter” theory is that consciousness is a quantum phenomenon. Roger Penrose believes that consciousness is created by quantum effects in neurons. If that is true, we are still not sure if machine can achieve it.
  • Consciousness is Fundamental – this is third variant of physical property theory, that consciousness is a fundamental force of the universe. It is simply a reordering of the hierarchy of science, with life explained by biology, which is explained by chemistry, which is explained by physics, which sits on top of space, time, consciousness and other fundamentals. If that is true, we are still not sure if machine can achieve it.
  • Consciousness is Universal – it means it is everywhere and everything is to one degree or another conscious. Can machine achieve it, probably yes.
  • Consciousness is a Trick of the Brain – the functions, it does its thing, and that sense that you have of an inner voice is just part of how brain works. Can machine achieve it, probably yes.
  • Something spiritual – something like soul. Can machine achieve it, probably yes.

But ultimately conscious computers may just be something that may be truly impossible, like traveling back in time.

Three breakthroughs would be needed to accomplish a meaningful merger of people and machines, and they may not be possible. First, a computer must be able to read a human thought. Second, a computer must be able to write a thought back to the brain. And third, a computer must do both those things at speeds substantially faster than what we are presently accustomed to.

Humans have never really defined themselves by biology, but by ability. In the wake of redefining our roles, in a world, where machines can step into human zone, with human like abilities, how will we react.

What will Fourth Age ultimately bring about? The catalyst are AI and robotics, which will increase productivity, expand wealth, accelerate the acquisition of knowledge, prolong life, and a lot more. Will progress continue. Author believes that there are few requirements for that. Imagination, a sense of time, a system for the accumulation and expansion of knowledge, and so on. Progress happens because of the symbiotic relationship between civilization and technology. Civilization facilitates the free flow of information. It is a social order promoting cultural creation. Elements of civilization are: economic provision, political organization, moral traditions and the pursuit of knowledge and the arts. Progress will continue as long as technology advances, because that’s what increases productivity.

What can development of technology do, for some fundamental questions of our time like hunger, poverty and disease. AI can improve our processes of food growing and so lower prices and increase yield. Poverty is also technical problem. It is possible that AI, robotics and other technologies will create prosperity so pervasive that no part of the world will be left behind. It is the same with disease, but in this area, we would need to solve also economic surrounding the industry.

Another area of progress will be field of clean energy. That is probably the quickest ticket for prosperity of all. Energy is the most abundant thing in universe, we need to solve technical challenge of how to manage it.

Wars could also become things of past. As the entire world becomes wealthier, war becomes more financially unthinkable. World is becoming trading partners and if we end poverty, we reduce war. The worldwide culture is shifting. We live in a world where economic accomplishments have largely replaced military ones of men.

With development of Internet we created infrastructure that can enable us to express ourselves, engage with each other, help each other, enable us to be creative, that we have impact and we can search for truth.

You may also like
Marco Iansiti, Karim R. Lakhani: Competing in The Age of AI; Strategy and Leadership When Algorithms and Networks Run the World
IoT, AI and blockchain by Oracle
Asilomar Principles – 2017
Max Tegmark: Life 3.0., Being human in the age of Artifical Intelligence