Artificial Lives Matter: AGI’s must be perceived as human in order to ensure mutual flourishing in a human-artificial social landscape
GCAS’s Florian Kleinau, author of The Dynamic of the Psyche, 21 Days of Giving, and An Always Within a Never, gives his astute vision of how humanity might best develop alongside the ever-present rise of potentially cognizant machines. This paper is thought provocative and demonstrates issues we, biological creatures, might best understand in our rise with technology. Together.
Abstract
The article discusses the problem of artificial general intelligence (AGIs) in human societies. It advances the argument that AGIs must be viewed as humans to ensure peaceful flourishing of human and artificial societies. In doing so, it considers the term equality from the standpoint of artificial agents and explores the consequences of establishing such equality between humans and AGIs both for AGIs and humans. It then critically assess the further development of AIs into AGIs suggesting that in light of its discussion it may be wise to abandon the aim of developing AGIs.
Keywords: Artificial General Intelligence, Artificial Agents, Equality, Non-Human Societies
Discrimination Originates in Perceived Difference
Artificial agents are among us - and their numbs are growing. As we cannot (and perhaps should not) stop their spread it is paramount to discuss how artificial agents can be integrated into human societies. This does not only pertain to agents appearing in ‘humanoid’ or ‘robotic’ shapes but is relevant to the ways we discourse about all forms of artificial intelligence that we accept as ‘human-like’. Although we are not quite at the point where we interact with agents that are confusingly similar to humans - yet classically not considered humans - we are, perhaps, very close.
The central problem is that humans have a tendency to protect members of their in-group against those that belong to the out-group. If this differentiation leads to a worse treatment of members that do not belong to one’s in-group, we speak of discrimination. The Cambridge dictionary defines discrimination as treating a group of people in a worse way than those others due to a difference in a certain attribute. The consequences of such discrimination have continue to peak across the world (Black Lives Matter, China’s conflict with the Uyghurs, etc.). While the issue of such conflicts is certainly complex, it seems to be rooted in the reciprocal perception of two groups as ‘different’ in a regard that cannot be mitigated to the extent that cooperation (as humans) defeats conflict.
We must therefore expect that this problem of discrimination based on perceived difference leading to conflict and violence will take hold in our interactions with artificial agents. Even though, the state-of-the-art Artificial Intelligence (AI) is quite primitive and not comparable to humans in terms of the generality of their intelligence, the public debate is already heavily negatively biased against these machines. Hatred against “the google algorithm” or “the evil robots that will take our jobs” permeate every corner of society. Nevertheless, the further evolution of artificial agents is continuing day after day. Thus, since the discourse highlighting the potential problems these agents (will) bring to human societies is already furious, the magnitude of this violence can be expected to surpass what we can witness in history. Accordingly, the line of argumentation in this article assumes that humans will discriminate against artificial agents. This means that they will treat them worse than other humans thereby not allowing them to participate and flourish in their own societies. I attempt to unpack this problem before it fully unfolds asking the question how discrimination against artificial agents can be avoided and highlighting why it must be avoided in order to ensure flourishing of human and artificial agents in peaceful co-existence.
Before starting this discussion we must advance some basic definitions. First, this paper deals with the integration of artificial agents into human societies. That means we are necessarily dealing with artificial general intelligence (AGI). An AGI displays the same general intelligence as humans that is not tied to specific tasks, has the ability to generalise and take a broad and interpretative view of the world. In the context of this paper AGI is defined as a system that matches humans in every ability, except its ‘humanness’. It must be assumed that once such an AGI is developed, it will quickly spread widely in society. This would imply the rise of a joint society of humans and artificial agents. Would equality not be the basis for the flourishing of such a joint society? Accordingly, in concussing the discussion of the question whether AGI would have to be considered “artificial life” or just “machines”, Hilary Putnam remarks that, “It is reasonable, then, to conclude that the question that titles this paper calls for a decision and not for a discovery. If we are to make a decision, it seems preferable to me to extend our concept so that robots are conscious-for "discrimination" based on the "soft- ness " or "hardness" of the body parts of a synthetic "organism" seems as silly as discriminatory treatment of humans on the basis of skin color”.
We must differentiate these AGIs from AIs which are essentially computer programs designed for a more or less narrow but always specific task. They are thus nothing other than tools that somehow serve as extensions of the human senses or body in the world. Using computer vision, an AI could assist a neuroscientist attempting to place an rTMS coil to stimulate the cortex to treat a patient with depression or help an architect to identify the ideal mix of certain materials. However, an AI would never (unless designed for it) suddenly start to do both of these tasks because it feels like they somehow overlap. While an AGI could also serve as an extension of the human senses or body as well it adds to this “functional” side the human ability to be creative, emotional, impulsive, etcetera. However, while I make this difference seem clear and sharp, it is in fact subject to ongoing debate. Nonetheless, for the purpose of this discussion, we have to accept the difference as if it was clear in order to understand how we can integrate AGIs into our human societies.
This paper will argue that in order to avoid conflicts caused by discrimination and foster mutual prosperity in a human and artificial social landscape, we must accept AGIs as human equals. Achieving this means that humans and AGIs must perceive AGIs as humans. In order to achieve such an acceptance, the perception of AGIs as equal to humans must be fostered promptly to take measures against discrimination and maintain these sustainably. This demands for a swift shift in the public framing of AGIs. In order to explore the consequences of the implementation or disregard of such equality, I will first detail my understanding of AGIs and the term equality in the context of artificial agents. Subsequently, the most important components of achieving such equality will be presented. After that, the positioning of AGIs in human societies will be outlined. Finally, the most important reasons for the implementation of such equality will be presented by discussing the consequences of not granting equality to AGIs. While the paper remains on the idealistic side of the discussion in trying to provocatively consider its pro-side a brief section with critical considerations is appended.
AGIs have Similar Emotional and Cognitive Abilities to Humans
By definition, an AGI will have the same general intellectual capacity as human beings. Consequently, it is able to reflect, introspect, self-perceive, feel, move, exhibit intentions, sense, etcetera in the same way that humans do. As we will see later, this must include the limitations that are inherent to human intelligence. Additionally, it is important to note that for this paper we must assume that an AGI will also inhabit the same emotional and physical landscape as a human. We are therefore speaking of embodied AGI, i.e. AGI in a physical shell that would look like a human. Additionally, as will be outlined in more detail below, AGIs must be situated in the human environment. Despite its importance for their acceptance as equal, such an embodiment and situatedness are crucial for the achievement of human-like generality of an AGI. A large body of research maintains that embodiment is crucial for the creation of meaning which is central for an AGI. This is especially integral to make meaning in a human world, since the space humans inhabit is shaped for a human body including the particularities of the human senses. Additionally, it must be assumed that meaningful interaction with human agents only arises from such embodiment. Therefore, only if such human-like embodiment is achieved, it will be possible to achieve the generality that characterizes an AGI. This development would already entail the assumption that humans and AGIs inhabit a shared landscape. Thus, AGIs and humans will live in a densely intertwined society and culture and interact on a daily basis. Therefore it is essential to foster peaceful unity between the two which can only be achieved by encouraging an equal perception of AGIs. But what does “equality” mean in this context?
Equality here refers to the human acceptance of artificial agents as human. This would imply that a human being, would be unable to differentiate between a human being and an artificial agent. Thus calls to mind Alan Turing’s ‘imitation game’ which posits that an artificial agent can be considered human-like enough when a human observer can no longer differentiate between a human and an artificial agent. Accordingly, equality means that humans perceive AGIs as equal and does not mean that they must be equal in every regard. However, they must still be equal enough to be perceived as equal. This is akin to the fact that while not every human is equal to another human in every regard but all humans share those properties that make them human. Therefore, next we have to investigate how such equality is possible with regard to how AGIs can be enabled to be perceived as human.
AGIs and Humans must Perceive AGIs as a Humans
While it may be impossible to actually turn an AGI into a human, it is hypothesized that a simple shift of perspective on these AGIs may be enough to make humans accept them as other humans. After all, humans themselves are very different from each other in terms of their intelligibility, sociability, behavior and even looks like on. Thus, we must establish the pillars of how an AGI would be accepted as equal in human society in order to understand why such equality is important to avoid racial discrimination and ensure mutual flourishing.
The starting point of this argument is the self-understanding of an AGI. The reason for this is simple: If humans build AGIs, this self-understanding can be provided to the machine a-priori to shape their understanding of the world. This attitude is a starting point to make humans perceive them as equal and also ensures that they will understand themselves as equal to humans which is integral, as will be discussed below. With this simple shift in self-perspective, the premises that AGIs do not differentiate between a human and themselves is already included in the very anatomy of the AGI.
In addition, humans must understand AGIs as fellow humans. This can be achieved by implemented some guidelines regarding the social framing of AGIs even before their development. First, it has become obvious that human-likeness must be fostered. In fact, the very basis of being considered an AGI in the framework of this paper is that humans are unable to differentiate between an AGI and a human agent. This would entail a shift in their position in society to enable them to exhibit human-like behaviour. How this can be achieved in detail will be discussed below. Second, a clear difference between AI and AGI must be established and public awareness of this distinction must be aimed for. In this distinction, AI would simply have the role of fulfilling certain specific tasks in a specific context. Even if such AIs would be embodied, their scope would remain local (i.e. a vacuum robot). Additionally, an AI would be unable to experience or exhibit human-like emotions, engage in behaviour intentionally, break rules, etc. Contrarily, AGIs must then be referred to as “humans” and credited full acceptation as such in human societies while AIs can continue to be referred to as robots, artificial intelligences etc. To further clarify this point, the way in which such AGIs must be integrated in human society must be discussed.
AGIs must be Fully Integrated into Human Societies
In accordance with these proposed guidelines, it must be the first and foremost premises of such a joint society that an AGI and a human have equal access to all societal institutions as well as equal freedom to unfold within them. This implies that the integration of AGIs in society, equality, on the one hand, refers to equality of opportunity and rights and on the other to equality of physical limitations. Both measures must be in place to ensure the perception of AGIs as equal to humans.
AGIs not only have to be conceptually equal but also be granted equal access all facets of human societies. If an AGI would function equivalently to a human agent, then it is required to learn certain social and intellectual skills before being able to serve as a fully functioning member of society in the way that human infants are. In order to learn such human skills it must be seamlessly integrated in human society. Since the AGIs are assumed to be embodied, this integration can be referred to as the necessity to situate the embodied AGI in the entirety of the human environment. Only then is it possible for an AGI to develop and grow into a human that is a fully functioning and accepted member of society. Accordingly, artificial agents must be granted equal access to schools, universities, jobs, politics, etcetera. All fields of the human societies we know today must be open to the AGI if they are to be perceived equal. Without such access it can be expected that AGIs could never develop skills that are equal to those of humans. This is because their human-like intelligence would pre-suppose the necessity to learn, specialise and socialise in order to live.
Such freedom already implies some limitations that may classically not be attributed to AGI. While classically, they may be thought of as having cognitive and physical constitutions, that considerably exceed those of humans by default, this would be detrimental to equality between the two groups. Therefore, an AGI must be subject to a life course resembling the human life one. This includes birth, growth, learning, the necessity to nurture and be nurtured, procreation, love, hunger, thirst and finally death. This is because a great amount of the human’s social behaviour is grounded in these shared boundaries of human life. They thereby determine much of the human social landscape that the AGIs inhabit and are integral for them to optimally adapt and achieve equality to humans.
Technically this would also imply the necessity to restrict AGI’s intellectual growth to a human-like framework. While AGIs by definition have the ability of general intelligence (i.e. learn and engage in all tasks) the growth of this intelligent (i.e. its capacity to learn and master tasks) must remain limited in order to be perceived as equal. Just like humans have certain preeminent limitations that restrict their interaction with the world (i.e. the limited capacity of out sensory system) and their growth (i.e. our limited memory), AGIs must also have such restrictions in order to interact equally with humans. For this paper such limitations are assumed to be given since AGIs have human-like bodies as well as sensory organs. This will ensure that humans will conceive of them as equal competitors.
Additionally, the agents must be subject to the same law and have a desire to behave accordingly in order to ensure the peaceful flourishing of society. It is evident that such compliance naturally arises if an AGI understands itself as “human”. To exemplify this point, we may consider the more general question of what would keep AGIs in the boundaries of ‘human’ flourishing and not go beyond them and turn against humans. The answer to this question appears to be quite simple: For the same reasons humans do not regularly turn against all humans: First, AGIs would have the same organisational and motivational limitations that humans have. They would get tired, hungry, be satisfied by simple pleasures of life such as a family or intellectual success and be unable to effectively organise a global rebellion against humans. This is also due to the fact that while a single AGI may have the opinion that such an uprising would be beneficial, it must still convince all other AGIs that this is the case. To do so, it is bound to the same tools humans are bound to and will therefore be as inefficient as humans are. Additionally, one must seriously ask the question why such an uprising against the entire humankind would be desirable for the AGI in a scenario where it considers itself “human”? They would rise up against themselves.
Finally, the question remains why such equality would be necessary to avoid discrimination conflicts and foster mutual prosperity and in the same vain it must be considered, what would happen if equality was not granted.
If equality was not granted, conflict would arise
In this final part of the paper, we must respond to the question of why equality is necessary in order to avoid discrimination conflicts and foster mutual prosperity. As briefly indicated in the introduction, discrimination based on an agents belonging to the same species can most definitely be avoided by eliminating that difference between two given agents. Concretely, as long as AGIs are viewed as artificial, they will be treated as non-humans. This would have several consequences:
First, they would be unable to fully adapt to human society, thereby not integrating ideally and therefore being treated even more unequally. It can even be expected that the development of general intelligence would therefore be impossible without equality. This is because only the full access to all institutions of human society would enable general intelligence to develop for human intelligence only unfolds in interaction with those. Thus, insufficient integration in human societies due to the perception as unequal would eventually lead to the formation of sub-societies with worse access to social institutions as has been the case in countries with strong discrimination against certain social groups. This development may be compared to the exclusion of women from certain institutions and other parts of societies in the USA and beyond, the aftermath of which still haunts our societies today. Since the access to all social institutions is integral for AGIs to fully develop their intelligence through situated learning, just like humans, the exclusion will decrease their potential productivity and further decrease their integration in human society along a negative spiral. Since humans classically view current artificial entities - AIs - as useful tools, a decrease in productivity will lead to frustration regarding AIs (and later AGIs) existence and ultimately increase the split between humans and these artificial agents leading to more discrimination. All this would increase potential hostility between the two groups, the consequences of which is apparent throughout many cultures that discriminate against each other throughout the world.
We must be Critical About the Development of AGIs
We need equality through equality! Notwithstanding the idealism leading up to the conclusions from the previous section, it seems paramount to append some critical considerations. The purpose of the paper has been to highlight why we should accept AGIs as humans instead of allowing their recognition as others. Perhaps this is rooted in a fear of precisely this otherness that can never be overcome. However, would negating their difference - i.e. pretending they are the same - not finally lead to the convergence of both human and artificial agents towards a third? Certainly, this will take conflict but it may prevent other, even larger, conflicts. There are several other problems arising from the standpoint I advanced: What about AGI’s own culture - would we not subvert it by perceiving them as human? Again, I would tend toward the argument that in reality, we would see the emergence of a joint human and artificial culture. Then would this not mean the loss of human culture? I would suggest that this culture is always lost again by moving through time and interacting with virtually anything.
An alternative to these questions (and many others) would be to never move towards AGIs and stick with AIs that simply function as human tools. This would be a solution and perhaps, in my opinion, a good one. However, I fear that this will not be possible simply as a consequence of human curiosity (everything that can be thought of will be developed). It seems that if AGIs will be developed, accepting them as equal humans would be the only way to prevent violence between “them and us”.
Conclusions
It has become clear that, if AGIs will be developed, only the denotation of artificial intelligence as humans would evade the conflicts that the discrimination of nongroup members would bring. Therefore, only this equality will ensure flourishing in a society in which artificial and human agents live together. This requires that both AGIs and humans view each other and themselves as humans and hence members of the same species. In order to achieve this, AGIs must not only view themselves as humans but must also be fully integrated in human society and culture as equal participants in all human institutions. Moreover, they must have the same physical and cognitive limitations as humans but also maintain the same freedom to grow, learn and develop that is fundamental to the human condition. Additionally, they must pass through the same life course and therefore have the same intergenerational interests as humans (i.e. death, birth and reproduction).
Such equality will lead to harmony and flourishing because both, humans and AGIs work for the continuation of a joint - instead of a separate - human society. This is because they both have an inherent tendency to maintain their own kind. Dystopian scenarios that are depicted in many science fiction movies can therefore only be avoided, if AGIs are viewed as equal - if they are viewed as humans. However, perhaps it would be best to not allow this development in the first place, to leave the dream of AGIs behind and simply stick to AIs and their function as good old human tools.
Resources
Cambridge English Dictionary (n.d.) Discrimination. In Cambridge Englisch Dictionary. Retrieved
3 December 2020, from https://dictionary.cambridge.org/de/worterbuch/englisch/discrimination.
De Vega, Manuel, Arthur C. Graesser, and Arthur M. Glenberg. "Reflecting on the debate." Symbols and embodiment: Debates on meaning and cognition: 397-440. 2008.
Putnam, Hilary R.: Machines or artificially created life?. The Journal of
Philosophy, 61(21), 668-691. 1964
Turing, Alan M. Computing Machinery and Intelligence, Mind 59, 433–460. 1950.
Cover photo by Possessed Photography