Sustensis VIDEO on ‘Who could save Humanity from Superintelligence’
This video was first shown at the University College London for the London Futurists on 29.04.2017. It is accompanied by a presentation shown below. You can view the presentation in Full Screen mode (Click on the bottom right hand icon). Alternatively, it will open in a new Window in an animated mode. You can download the presentation by clicking the download link underneath.
Sustensis PRESENTATION on ‘Who could save Humanity from Superintelligence’
Here is this book review by dr. Stephen Cooney and my response below:
Tony Czarnecki’s response to the review of his book ‘Who Could Save Humanity from Superintelligence?’ by Dr Stephen Cooney
London, 26/10/2020
Introductory note
Stephen Cooney has Ph,D. from the London School of Economics. He is now retired, having worked most recently at the Congressional Research Service of the Library of Congress in Washington, DC. He continues to follow economic and political issues at Chatham House in London.
I am grateful for Stephen’s review of my book and very precise, well-argued comments. I should mention that it was my first book on the subject of the impact of Superintelligence on the future of Humanity, first published in 2018. Since then I have published three more books as a Posthumans series: ‘Federate to Survive!’, ‘Democracy for a Human Federation’ and ‘Becoming a Butterfly’ – in June 2020. My comments are below Stephen’s review.
This is a very serious book about an extremely serious subject. Nothing that this author will write intends to denigrate from the ideas of Tony Czarnecki, the book’s author. Quite the opposite: my theme, in criticising this work is, to use Mies’ famous cliche, ‘Less is more.’
Let’s start at the beginning. Tony lays out in the introduction his proposition that ‘superintelligence,’ derived from the technology of artificial intelligence (AI), is already on its way of becoming a different animal species altogether. He quotes another futurist, Nick Bostrom, as describing how AI may evolve into ‘superintelligence’ (a form of artificial general intelligence). It will have the ability to perform every task in every field, including ‘scientific creativity, general wisdom and social skills … better than the best human brains.’ Tony lays out his fear that, if it cannot be managed or controlled (i.e., tamed) superintelligence can take over the planet from the extant human species within a period of time as short as 50 years.
As illustrated in the development of robotics, over time more and more simple mechanical functions have been transferred from human beings to robots. But that is ancient history. Advanced AI technology packs in progressively more transistors and can use supercomputers as a tool. These AI machines now have been able to defeat the best expert players at in games such as chess and the Asian game of go. Of course, machines still can’t predict the weather with 100% accuracy – but they’re getting better all the time. And new technology features are constantly being added, which give AI-enhanced robots vision, hearing, and all the other senses, plus increased functional utility and manipulability.
But this is not yet superintelligence.
Even the most enhanced AI equipment has to be programmed, even when computers design other computers, with a human manager and controller. To a achieve higher skill levels, AI will probably have to advance to quantum computing, a science still in its infancy. Put at its simplest, as this author understands, quantum computers will be able to solve simultaneous equations with multiple outcomes. ‘Brownian movement,’ as one example, is a case in which the observer’s interaction with a particle has an ongoing effect on the particle, thereby changing its state from the initial prediction. Election polls, in another specific case, as they are published may affect voting outcomes, thereby rendering original predictions inaccurate. To borrow a phrase used by US Defense Secretary Donald Rumsfeld, quantum computers may not only be able to calculate all ‘known unknowns,’ but also to predict ‘unknown unknowns’ – or even unknowable unknowns?
So far this is pretty intriguing, and I would have liked to learn more about how the mechanics of quantum computing and superintelligence will work. But Tony then veers off into a lengthy assessment of anthropogenic and non-anthropogenic risks, and how they will affect our planet’s future and possible demise. This includes a high probability of superintelligence taking over the world within the next century, either by accident or intentional design. Both of these possible outcomes he considers, with the consequences either benevolent or malevolent for the future of the human race. Mostly, he says, we will ultimately have to live with superintelligence as an autonomous force and integrate autonomous AI with the human species – possibly even physically with implanted chips in human brains (chapter 3).
But in outlining other ‘existential risks’ in chapter 2, and in more detail in chapters 4-7, this book gets badly off track in focus. He rates the risks from everything from nuclear war (high), through future pandemics (more unknown unknowns), climate change (slow but probable if we don’t do anything), and not excluding being hit by an asteroid. These are ipso facto not directly related to superintelligence, except insofar as human control over AI/superintelligence can contribute to an amelioration or mitigation of these catastrophic outcomes.
Then in chapters 7-8, and the subsequent parts 2-3, he gets to what seems to be the real subject of the book. If we need superintelligence to avert, control or mitigate existential catastrophes, including runaway superintelligence itself, how are we to do this? We need a world government, and we have at most 10, or at the outside 20 years, before a superintelligence, or to use its full name, artificial general intelligence, takes over.
In the second part of the book, Tony evaluates alternative means of global reorganization, concluding, surprise, surprise, that it must be a democracy. Moreover, after exhaustively surveying the options, he concludes, somewhat more surprisingly, that it should be based on a modified version of the European Union. Only a democracy can deliver on the ‘Universal Values of Humanity,’ which must be the basis of the future global order. The United Nations, which initially published its charter of human rights, cannot accomplish this in a reasonable time frame, because of its structural requirement for consensus among specified leading military powers. Therefore, the new world order must grow out of an existing subset of territorial organizations, and it better do it quick.
There are lots of problems with this. First Tony ignores a rich history dating back at least to the 1600s on plans for world government and why they all failed. For one thing, surprising for an economist like Tony, he builds on the UN’s human rights declaration, but he ignores the UN Charter on the Economic Rights and Duties of States. This equally significant and later companion document sets global systems development in a necessary and underlying economic context. Without attention to the economic context, any plan for global confederation or consensus will fail. This despite the fact that in some of its activities and specialized agencies, the UN has achieved many notable successes.
But, secondly, Tony ignores a whole body of modern political science and international relations theory which has moved beyond idealized legalistic structures to analysis of existing of power relationships. In the early Twentieth Century, at the time of the League of Nations and the Briand-Kellogg Pact that outlawed war, it was believed that designing the perfect global constitution was the key to a global regime of peace and harmony. But our understanding has evolved, through functionalism, neo-functionalism, power politics, systems theory and the evolution of capitalist liberalism. Designing an ‘ideal’ system from the outset, which ignores the realities about which we have learned, will not work.
In part 2, Tony comes to the conclusion that only the European Union, appropriately modified, has the potential to become the basis for the necessary world government. The United States provides a useful model, but its idiosyncratic rules would most probably not be subject to requisite flexibility. Elsewhere, he suggests that the only potential substitute would be China, but this would be at the expense of accepting an authoritarian dictatorship and the consequent loss of personal freedom.
In part 3, he describes in detail the options that should be considered in designing the modifications for the new ‘European Federation,’ and how its transition may be accomplished. Internally, the EU is not prepared to take the necessary steps Tony projects will be needed to accomplish a broader federation.
Brexit could be viewed in support of this idea as the subtraction that will aid addition. The EU is now a more legally coherent organization without the UK. But still, not every country is in the euro. The economic recovery bargain of grants and loans occasioned by the joint response to the coronavirus pandemic may be an important step in the direction of internal solidarity, if not a ‘Hamiltonian moment.’ But Stein Rokkan, a Swedish political scientist, brilliantly analysed how the then-EEC, a fundamentally confederal structure, did not have the internal means to reorganize into a centralized state. One of Tony’s first projects in this direction is the establishment of a European Army. Without NATO, an extra-EU entity, as its basis, this is a modest idea. The sames is the same for many other changes within Europe, that Tony posits must happen to create an idealized European Federation, which would ultimately be led by a Consensual Presidential Democracy. Some of his specific ideas are positively outlandish. The idea of ‘sortition,’ based on Athenian democracy and Anglo-American jury selection, to choose ‘qualified’ decision makers in some type of future Senate is a fantastic sketch of an idea, but divorced from practical reality.
But the external dimensions of the proposed European Federation are even more problematic. Tony consistently proposes that as it evolves toward a world government, it must be based on ‘Judeo-Christian values.’ No chance to sell that to Muslims, Hindus, Buddhists and officially designated atheists (such as Chinese) who make up the bulk of the world’s population. Not to mention the proposal that English must be the globe’s official lingua franca. Try selling that in Latin America. This is more than insensitivity; it becomes a self-denying prophecy.
Tony deals with this diversity by allowing a range of 5 Zones of access to the European Federation, starting outward from a core that accepts the full details and principles in full. This could hopefully include a UK that realizes the wisdom of the revised plan. Other nations will accept in varying degrees a progressively increasing structure of core values, democratic principles and degrees of economic relationships.
Tony assumes the ‘Planet Earth’ will be so nonplussed by the looming threat of superintelligence, that it will be willing to forego millennia of cultural inheritance. The great German General von Moltke is quoted as saying that all plans will work, until the first shot is fired. Tony’s elaborate hypothesized structure of a European Federation as the nexus of future global government cannot be implemented until the need is clearly felt by everyone. Then it will be too late to implement such elaborate plans laid out in advance, even if they might work in the first place. It would be far better to work toward an incremental understanding of superintelligence as it evolves, and how we can control it, at both the national and international collective levels.
TONY CZARNECKI’S RESPONSE:
On developing AI: (SC: Even the most enhanced AI equipment has to be programmed, even when computers design other computers, with a human manager and controller.)
This is no longer the case. AI has already past the Machine Learning phase and human trained learning (assisted learning), moving to unassisted learning, where an AI agent is not even given a goal – it has to arrive at it itself, as well as how to solve a problem, or which decision is most optimal.
On Superintelligence: (SC: New technology features are constantly being added, which give AI-enhanced robots vision, hearing, and all the other senses, plus increased functional utility and manipulability. But this is not yet superintelligence.)
I agree that the current AI is not Superintelligence. This is a product in the making, which as a mature entity may be delivered in a few decades from now.
On Quantum computing: (SC: AI will probably have to advance to quantum computing, a science still in its infancy.)
Even in 2018 some mathematicians, physicists and computer specialists were proclaiming that making a working quantum computer will not be possible, mainly because of decoherence – see here: https://www.quantamagazine.org/gil-kalais-argument-against-quantum-computers-20180207/ . Today, any scientist may log on to IBM website https://quantum-computing.ibm.com/login to do his research using a quantum computer. There are at least a dozen more such cloud quantum computing facilities: https://en.wikipedia.org/wiki/Cloud-based_quantum_computing#:~:text=Quantum%20computers%20achieve%20their%20massive,quantum%20computing%20within%20the%20cloud
On existential risks: (SC: He rates the risks from everything from nuclear war (high), through future pandemics (more unknown unknowns), climate change (slow but probable if we don’t do anything), and not excluding being hit by an asteroid. These are ipso facto not directly related to superintelligence…)
I have explored both anthropogenic (man-made) and non-anthropogenic (natural) existential risks to make a relative assessment of the overall risk of delivering a malicious Superintelligence. Only in such a context it becomes obvious that Superintelligence is the earliest and most profound threat for a human species.
On treating other cultures with due respect: (SC: Tony consistently proposes that as it (the Federation- TC) evolves toward a world government, it must be based on ‘Judeo-Christian values.’ No chance to sell that to Muslims, Hindus, Buddhists and officially designated atheists (such as Chinese) who make up the bulk of the world’s population.)
I am aware that some suggestions, like that one about founding the future European Federation’s values on Judaeo-Christian principles, may feel offensive to some faiths. If some people feel that way, then I am really sorry. I would like to assert everyone that my ideal view would be for all cultures, faiths, and civilizations, such as Chinese, to contribute to a new set of Universal Values of Humanity.
Where does it then leave the process of melting cultures into one planetary cohesive culture? I would love that to happen but being a realist, I do not think it will. It is too late for what would have been required to do it properly and what is possible in the shortest time (just within a decade). We need to focus on the ultimate goal – minimizing a possibility of a human species’ extinction. Life is the top value. Everything else, even freedom to some extent, is trumped up by that value. Either we will be extinct by the end of this century for various reasons, or we have a mature Superintelligence in the second half of the century, which will be better ‘qualified’ to deal with such a problem like a common culture.
Nevertheless, we would need somehow a cohesive view on what kind of a heritage and values we would like Superintelligence to get. When we have a benevolent, mature Superintelligence, it may view our values and some ideals stemming from our anthropic thinking. It will certainly be capable to show us all inconsistences in our values and absurdities that we apply in various spheres of life. We will be in its (hers or his) ‘hands’.
And yet, despite the awareness that realism should prevail, since there is so little time left, I take an almost idealistic view, on how this set of Universal Values of Humanity could be agreed, making it the cornerstone of the future Constitution. That becomes then the basis for suggesting various electoral systems. This includes an auxiliary method Sortition for arriving at a common consensus in deciding very important issues for a country or a Federation. Incidentally, the term Sortition, is now almost universally substituted by a Citizens’ Assembly and is to form part of the Future of Europe Conference, which is to decide in the next two years the depth of the European integration (read ‘federalization). On the subject of federalization, I make a very strong point saying, it should be very shallow, adopting the principle of the so-called Minimal State.
On the World Government. The rest of Stephen’s review deals with this subject and I only quote this fragment: (SC:…Tony ignores a whole body of modern political science and international relations theory which has moved beyond idealized legalistic structures to analysis of existing of power relationships)
I have analysed in detail 10 potential candidates, including the UN and NATO, which would best meet 10 criteria to act initially as a pseudo World Government. I have concluded that the European Union has the best chance to convert itself into the first version of a de-facto World Government by 2030. This is, in my view, a key criteria, since as I argue in the follow-on books, we may lose the control over the development of AI by 2030, which incidentally coincides with a tipping point for stopping the worsening of the climate change. I am also acutely aware of the near exponential pace of change in many domains, as Covid-19 has proven, which requires an immediate international action.
By suggesting the European Federation as the best candidate for a pseudo World Government and a prototype of a future Human Federation, I am not saying that it is doable. There is a good chance that it will not happen at all, because as the current US elections prove, people in general are emotional beings, and quite often think irrationally, believing in something that they want to believe against all the odds. As things stand today, anyone like myself, calling for the EU to be converted into a European Federation by about 2025 and then extending its scope to become a pseudo World Government by 2030, may be considered not living on this planet. I agree. I am only suggesting what should be done, not if it is definitely going to be achieved within the suggested timescales. Incidentally, I mention in the book that should the current EU be transformed into a federation and take the mantle of a pseudo World Government, it may create a serious, though perhaps not an existential risks, by doing that. As Stephen rightly points out – what about Russia, China or even the USA? However, I believe humanity simply does not have another option and wait till everyone agrees to everything – it will never happen.
I do not try to denigrate the role of the UN at all. I am only lamenting that such a theoretically great organization with so many achievements, including the Universal Declaration of Human Rights (in an urgent need of a deep revision) is incapable of governing the world. I also focus a lot on how to make the enlargement process of the future European Federation (or whatever we call it) more realistic by helping to distribute wealth globally through Global Wealth Redistribution Fund. This is a reference to the role of economics in the process of creating a Human Federation.
I do not believe that a process of building a Human Federation, can happen in one stage. It has to be gradual but also extremely fast, even somewhat chaotic. That is why I propose a 4-Zone system for non-member countries, attached to the European Federation, where countries move from a lower Zone to a full membership of the Federation over a few decades.
I tend to agree with Stephen that a European Federation, nor a Human Federation, can be elaborately built over time. It will rather happen in a chaotic fashion, under pressure, as people almost touch a danger and understand it by experiencing its first symptoms. In this sense Covid-19 illustrates it best. That is why I have presented three scenarios for creating a European Federation, including the third one, that it does not happen at all, with probably fatal consequences not just for Europe but for the world. Today, just nearly 3 years later after writing the book, I see the creation of the European Federation by 2025 as rather a late date, if we consider that by 2030, we may already lose control over the maturing AI.