AI regulation is good! Positive benefits solve extinction.
Tzimas ’21 [Themistoklis; 2021; Faculty of Law at the Aristotle University of Thessaloniki; Legal and Ethical Challenges of Artificial Intelligence from an International Law Perspective, “Chapter 2: The Expectations and Risks from AI,” p. 9-32]
Therefore, it is only natural to be at least skeptical towards a future with entities possessing equal or superior intelligence and levels of autonomy; the prospect even of existential risk looms as possible.7
AI that will have reached or surpassed our level of intelligence make us wonder why would highly autonomous and intelligent AI want to give up control back to its original creators?8 Why remain contained in pre-defined goals set for it by us, humans?
Even AI in its current form and narrow intelligence poses risks because of its embedded-ness in an ever-growing number of crucial aspects of our lives. The role of AI in military, financial,9 health, educational, environmental, governance networks-among others—are areas where risk generated by AI—even limited— autonomy can be diffused through non-linear networks, with significant impact— even systemic.10
The answer therefore to the question whether AI brings risk with it is yes; as Eliezer Yudkowski comments the greatest of them all is that people conclude too early that they understand it11 or that they assume that they can achieve it without necessarily having acquired complete and thorough understanding of what intelli- gence means.12
Our projection of our—lack of complete—understanding of the concept of intelligence on AI is owed to our lack of complete comprehension of human intelligence too, which is partially covered by the prevalent and until now self- obvious, anthropomorphism because of which we tend to identify higher intelligence with the human mind.
Yudkowski again however suggests that AI “refers to a vastly greater space of possibilities than does the term “Homo sapiens.” When we talk about “AIs” we are really talking about minds-in-general, or optimization processes in general. Imagine a map of mind design space. In one corner, a tiny little circle contains all humans; within a larger tiny circle containing all biological life; and all the rest of the huge map is the space of minds-in-general. The entire map floats in a still vaster space, the space of optimization processes.”13
Regardless of what our well-established ideas are, there are many, different intelligences and even more significantly, there are potentially, different intelli- gences equally or even more evolved than human.
From such a perspective, the unprecedented—ness of potential AI developments and the mystery surrounding them emerges as not only the outcome of pop culture but of a radical transformation of our—until recently—self—obvious identification of humanity with highly evolved and dominant intelligence.14
The lack of understanding of intelligence and therefore of AI may be frightening but does not lead necessarily to regulation—at least to a proper one. We could even be led into making potentially catastrophic choices, on the basis of false assumptions.
On top of our lack of understanding, we should add a sentiment of anxiety as well as of expectations, which intensifies as an atmosphere of emergency and of expected groundbreaking developments grows. The most graphic description of this feeling is the potential of a moment of singularity, as mentioned above according to the description by Vinge and Kurzweil.
As the mathematician I. J. Good–Alan Turing’s colleague in the team of the latter during World War II—has put it: “Let an ultraintelligent machine be defined as a machine that can far surpass all the intellectual activities of any man however clever. Since the design of machines is one of these intellectual activities, an ultraintelligent machine could design even better machines; there would then unquestionably be an “intelligence explosion,” and the intelligence of man would be left far behind. Thus the first ultraintelligent machine is the last invention that man need ever make, provided that the machine is docile enough to tell us how to keep it under control.”15 This is in a nutshell the moment of singularity.
The estimates currently foresee the emergence of ultra or super intelligence—as it is currently labelled—or in other words of singularity, somewhere between 20 and 50 years from today, further raising the sentiment of emergency.16 We cannot even foretell with precision how singularity would look like but we know that because of its expected groundbreaking impact, both states and private entities compete towards gaining the upper hand in the prospect of the singularity.17
Despite the fact that such predictions have been proven rather optimistic in the past18 and therefore up to some extent inaccurate, there are reasons to assume that their materialization will take place and that the urgency of regulation will be proven realistic.
After all, part of the disappointments from AI should be blamed on the fact that certain activities and standards, which were considered as epitomes of human intelligence have been surpassed by AI, only to indicate that they were not eventu- ally satisfactory thresholds for the surpassing of human intelligence.19 Partially because of AI progress we realize that human intelligence and its thresholds are much more complicated than assumed in the past.
The vastness’s of definitions of intelligence, as well as its etymological roots are enlightening of the difficulties: “to gather, to collect, to assemble or to choose, and to form an impression, thus leading one to finally understand, perceive, or know”.20
As with other relevant concepts, the truth is that until recently our main way to approach intelligence for far too long was “we know it, when we see it”. AI is an additional reason for looking deeper into intelligence and the more we examine it, the most complicated it seems.
The combination of lack of complete understanding of intelligence, the unpredictability of AI, its rapid evolution and the prospect of singularity explain both the fascination and the fear from AI. Once the latter emerges, we have no real knowledge about what will happen next but only speculations, which until recently belonged to the area of science fiction.
We are for example pretty confident that the speed of AI intelligence growth will accelerate, once self—improvement will have been achieved. The expected or possible chain of events will begin from AI capacity to re-write its own algorithms and exponentially self—improve, surpassing human intelligence, which lacks the capacity of such rapid self—improvement and setting its own goals.21
We can somehow guess the speed of AGI and ASI evolution and possibly some of its initial steps but we cannot guess the directions that such AI will choose to follow and the characteristics that it will demonstrate. Practically, we credibly guess the prospects of AI beyond a certain level of development.
Two existential issues could emerge: first, an imbalance of intelligence at our expense—with us, humans becoming the inferior species—in favor of non-biological entities and secondly a lack of even fundamental conceptual communication between the two most intelligent “species”. Both of them heighten the fear of irreversible changes, once we lose the possession of the superior intelligence.22
However, we need to consider the expectations as well. The positive side focuses on the so-called friendly AI, meaning AI which will benefit and not harm humans, thanks to its advanced intelligence.23
AI bears the promise of significantly enhancing human life on various aspects, beginning from the already existing, narrow applications. The enhanced automation24 in the industry and the shift to autonomy,25 the take—over by AI of tasks even at the service sector which can be considered as “tedious”—i.e. in the banking sector—climate and weather forecasting, disaster response,26 the potentially better cooperation among different actors in complicated matters such as in matters of information, geopolitics and international relations, logistics, resources ex.27
The realization of the positive expectations depends up to some extent upon the complementarity or not, of AI with human intelligence. However, what friendly AI will bring in our societies constitutes a matter of debate, given our lack of unanimous approach on what should be considered as beneficial and therefore friendly to humans—as is analyzed in the next chapter.
Friendly AI for example bears the prospect of freeing us from hard labor or even further from unwanted labor; of generating further economic growth; of dealing in unbiased, speedy, effective and cheaper ways with sectors such as policing, justice, health, environmental crisis, natural disasters, education, governance, defense and several more of them which necessitate decision-making, with the involvement of sophisticated intelligence.
The synergies between human intelligence and AI “promise” the enhancement of humans in most of their aspects. Such synergies may remain external—humans using AI as external to themselves, in terms of analysis, forecasts, decision—making and in general as a type of assistant-28 or may evolve into the merging of the two forms of intelligence either temporarily or permanently.
The second profoundly enters humanity, existentially—speaking, into uncharted waters. Elon Musk argues in favor of “having some sort of merger of biological intelligence and machine intelligence” and his company “Neuralink” aims at implanting chips in human brain. Musk argues that through this way humans will keep artificial intelligence under control.29 The proposition is that of “mind design”, with humans playing the role that God had according to theologies.30
While the temptation is strong—exceeding human mind’s capacities, far beyond what nature “created”, by acquiring the capacity for example to connect directly to the cyberspace or to break the barriers of biology31—the risks are significant too: what if a microchip malfunction? Will such a brain be usurped or become captive to malfunctioning AI?
The merging of the two intelligences is most likely to evolve initially by invoking medical reasons, instead of human enhancement. But the merging of the two will most likely continue, as after all the limits between healing and enhancement are most often blurry. This development will give rise, as is analyzed below, to signif- icant questions and issues, the most of crucial of which is the setting of a threshold for the prevalence of the human aspect of intelligence over the artificial one.
Human nature is historically improved, enhanced, healed and now, potentially even re-designed in the future.32 Can a “medical science” endorsing such a goal be ethically acceptable and if yes, under what conditions, when, for whom and by what means? The answers are more difficult than it seems. As the World Health Organi- zation—WHO—provides in its constitution, “Health is a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity”.33
Therefore, why discourage science which aims at human-enhancement, even reaching the levels of post-humanism?34 Or if restrictions are to be imposed on human enhancement, on what ethics and laws will they be justified? How ethically acceptable is it to prohibit or delay technological evolution, which among several other magnificent achievements, promises to treat death as a disease and cure it, by reducing soul to self, self to mind, and mind to brain, which will then be preserved as a “softwarized” program in a hardware other than the human body?35
After all, “According to the strong artificial intelligence program there is no fundamental difference between computers and brains: a computer is different machinery than a person in terms of speed and memory capacity.”36
While such a scientific development and the ones leading potentially to it will be undoubtedly, groundbreaking technologically-speaking, is it actually—ethically- speaking—as ambivalent as it may sound or is it already justified by our well— rooted human-centrism?37
Secular humanism may have very well outdated religious beliefs about afterlife in the area of science but has not diminished the hope for immortality; on the contrary, science, implicitly or explicitly predicts that matter can in various ways surpass death, albeit by means which belong in the realm of scientific proof, instead of that of metaphysical belief.38
If this is the philosophical case, the quest for immortality becomes ethically acceptable; it can be considered as embedded both in the existential anxiety of humans, as well as in the human-centrism of secular philosophical and political victory over the dei-centric approach to the world and to our existence.
From another perspective of course and for the not that distant philosophical reasons, the quest for immortality becomes ethically ambiguous or even unacceptable.39 By seeking endless life we may miss all these that make life worth living in the framework of finiteness. As the gerontologist Paul Hayflick cautioned “Given the possibility that you could replace all your parts, including your brain, then you lose your self-identity, your self-recognition. You lose who you are! You are who you are because of your memory.”40
In other words, once we begin to integrate the two types of intelligence, within ourselves, until when and how we will be sure that it is human intelligence that guides us, instead of the AI? And if we are not guided completely or—even further—at all by human intelligence but on the contrary we are guided by AI which we have embodied and which is trained by our human intelligence, will we be remaining humans or we will have evolved to some type of meta-human or transhumant species, being different persons as well?41
AI promises tor threatens to offer a solution by breaking down our consciousness into small “particles” of information—simplistically speaking—which can then be “software-ized” and therefore “uploaded” into different forms of physical or non-physical existence.
Diane Ackerman states that “The brain is silent, the brain is dark, the brain tastes nothing, the brain hears nothing. All it receives are electrical impulses--not the sumptuous chocolate melting sweetly, not the oboe solo like the flight of a bird, not the pastel pink and lavender sunset over the coral reef--only impulses.”42 Therefore, all that is needed—although it is of course much more complicated than we can imagine—is a way to code and reproduce such impulses.
Even if we consider that without death, we will no more be humans but something else, why should we remain humans once technologies allow us be something “more”, in the sense of an enhanced version of “being”? Why are we to remain bound by biological evolution if we can re-design it and our future form of existence?
Why not try to achieve the major breakthrough, the anticipated or hoped digita- lization of the human mind, which promises immortality of consciousness via the cyberspace or artificial bodies: the uploading of our consciousness so that it can live on forever, turning death into an optional condition.43
Either through an artificial body or emulation-a living, conscious avatar—we hope—or fear—that the domain of immortality will be within reach. It is the prospect of a “substrate-independent minds,” in which human and machine consciousness will merge, transcending biological limits of time, space and mem- ory” that fascinates us.44
As Anders Sandberg explained “The point of brain emulation is to recreate the function of the original brain: if ‘run’ it will be able to think and act as the original,” he says. Progress has been slow but steady. “We are now able to take small brain tissue samples and map them in 3D. These are at exquisite resolution, but the blocks are just a few microns across. We can run simulations of the size of a mouse brain on supercomputers—but we do not have the total connectivity yet. As methods improve, I expect to see automatic conversion of scanned tissue into models that can be run. The different parts exist, but so far there is no pipeline from brains to emulations.”45
The emulation is different from a simulation in the sense that the former mimics not only the outward outcome but also the “internal causal dynamics”, so that the emulated system and in this particular case the human mind behaves as the original.46 Obviously, this is a challenging task: we need to understand the human brain with the help of computational neuroscience and combine simplified parts such as simulated neurons with network structures so that the patterns of the brain are comprehended. We must combine effectively “biological realism (attempting to be faithful to biology), completeness (using all available empirical data about the system), tractability (the possibility of quantitative or qualitative simulation) and understanding (producing a compressed representation of the salient aspects of the system in the mind of the experimenter)”.47
The technological challenges are vast. Technologically speaking, the whole concept is based on some assumptions which must be proven both accurate and feasible.48 We must achieve technology capable of scanning completely the human brain, of creating software on the basis of the acquired information from its scanning and of the interpretation of information and the hardware which will be capable of uploading or downloading such software.49 The steps within these procedures are equally challenging. Their detailed analysis evades the scope of this book.
Some critical questions—they are further analyzed in the next chapters—emerge however: how will we interpret free will in emulation? What will be the impact of the environment and of what environment? How will be missing parts of the human brain re-constructed and emulated? What will be the status of the several emulations which will be created—i.e. failed attempts or emulations of parts of the human brain—in the course of the search for a complete and functioning emulation? Will they be considered as “persons” and therefore as having some right or will they be considered as mere objects in an experimental lab? How are we going to decode the actual subjective sentiments of these emulations? Essentially, are emulations the humans “themselves” who are emulated or a different person? Even further what will human and person mean in the era of emulation?
From a different perspective, the victory over death may be seen as a danger of mass extinction, absorption or de-humanization. In this new, vast universe of emulations will there be place for humans?50
From the above—mentioned discussion, it becomes obvious that at a large extent, the prospect of risk or of expectation is a matter of perspective, for which there is no unanimous agreement in the present. This may be the greatest danger of all, for which Asimov warned us: unleashing technology while we cannot communicate among us, in the face of it.
The existential prospect as well as the risks by AI may self-evidently emerge from technological advances but are determined on the basis of politico—philosophical or in the wider sense, ethical assumptions. This is where the need for legal regulation steps in. Such a need was often underestimated in the past in favor of a solely technologically oriented approach—although exceptions raising issues other than technological can be found too.51 The gradual raising of ethic—political, philosoph- ical and legal issues constitutes a rather recent development, partially because of the realization of the proximity of the risks and of the expectations.
The public debate is often divided between two “contradictory” views: fear of AI or enthusiastic optimism. The opinions of the experts differ respectively.
Kurzweil, who has come with a prediction for a date for the emergence of singularity—until 2045—expects such a development in a positive way: “What’s actually happening is [machines] are powering all of us,” Kurzweil said during the SXSW interview. “They’re making us smarter. They may not yet be inside our bodies, but, by the 2030s, we will connect our neocortex, the part of our brain where we do our thinking, to the cloud.”52
In a well-known article—issued on the occasion of a film—Stephen Hawking, Max Tegmark, Stuart Russell, and Frank Wilczek shared a moderate position: “The potential benefits are huge; everything that civilization has to offer is a product of human intelligence; we cannot predict what we might achieve when this intelligence is magnified by the tools AI may provide, but the eradication of war, disease, and poverty would be high on anyone’s list. Success in creating AI would be the biggest event in human history. . . Unfortunately, it might also be the last, unless we learn how to avoid the risks.”53