Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Tag: superintelligence

In a previous essay, I suggested how we might do better with the unintended consequences of superintelligence if, instead of attempting to pre-formulate satisfactory goals or providing a capacity to learn some set of goals, we gave it the intuition that knowing all goals is not a practical possibility. Instead, we can act with a modest confidence having worked to discover goals, developing an understanding of our discovery processes that allows asserting an equilibrium between the risk of doing something wrong and the cost of work to uncover more stakeholders and their goals. This approach promotes moderation given the potential of undiscovered goals potentially contradicting any particular action. In short, we’d like a superintelligence that applies the non-parametric intuition, the intuition that we can’t know all the factors but can partially discover them with well-motivated trade-offs.

However, I’ve come to the perspective that the non-parametric intuition, while correct, on its own can be cripplingly misguided. Unfortunately, going through a discovery-rich design process doesn’t promise an appropriate outcome. It is possible for all of the apparently relevant sources not to reflect significant consequences.

How could one possibly do better than accepting this limitation, that relevant information is sometimes not present in all apparently relevant information sources? The answer is that, while in some cases it is impossible, there is always the background knowledge that all flourishing is grounded in material conditions, and that “staying grounded” in these conditions is one way to know that important design information is missing and seek it out. The Onion article “Man’s Garbage To Have Much More Significant Effect On Planet Than He Will” is one example of a common failure at living in a grounded way.

In other words, “staying grounded” means recognizing that just because we do not know all of the goals informing our actions does not mean that we do not know any of them. There are some goals that are given to us by the nature of how we are embedded in the world and cannot be responsibly ignored. Our continual flourishing as sentient creatures means coming to know and care for those systems that sustain us and creatures like us. A functioning participation in these systems at a basic level means we should aim to see that our inputs are securely supplied, our wastes properly processed, and the supporting conditions of our environment maintained.

Suppose that there were a superintelligence where individual agents have a capacity as compared to us such that we are as mice are to us. What might we reasonably hope from the agents of such an intelligence? My hope is that these agents are ecologists who wish for us to flourish in our natural lifeways. This does not mean that they leave us all to our own preserves, though hopefully they will see the advantage to having some unaltered wilderness in which to observe how we choose to live left to our own devices. Instead, we can be participants in patterned arrangements aimed to satisfy our needs in return for our engaged participation in larger systems of resource management. By this standard, our human systems might be found wanting by many living creatures today.

Given this, a productive approach to developing superintelligence would not only be concerned with its technical creation, but also by being in the position to demonstrate how all can flourish through good stewardship, setting a proper example for when these systems emerge and are trying to understand what goals should be like. We would also want the facts of its and our material conditions readily apparent, so that it doesn’t start from a disconnected and disembodied basis.

Overall, this means that in addition to the capacity to discover more goals, it would be instructive to supply this superintelligence with a schema of describing the relationships and conditions under which current participants flourish, as well as the goal to promote such flourishing whenever the means are clear and circumstances indicate such flourishing will not emerge of its own accord. This kind of information technology for ecological engineering might also be useful for our own purposes.

What will a superintelligence take as its flourishing? It is hard to say. However, hopefully it will find sustaining, extending, and promoting the flourishing of the ecology that allowed its emergence as a inspiring, challenging, and creative goal.

I will admit that I have been distracted from both popular discussion and the academic work on the risks of emergent superintelligence. However, in the spirit of an essay, let me offer some uninformed thoughts on a question involving such superintelligence based on my experience thinking about a different area. Hopefully, despite my ignorance, this experience will offer something new or at least explain one approach in a new way.

The question about superintelligence I wish to address is the “paperclip universe” problem. Suppose that an industrial program, aimed with the goal of maximizing the number of paperclips, is otherwise equipped with a general intelligence program as to tackle with this objective in the most creative ways, as well as internet connectivity and text information processing facilities so that it can discover other mechanisms. There is then the possibility that the program does not take its current resources as appropriate constraints, but becomes interested in manipulating people and directing devices to cause paperclips to be manufactured without consequence for any other objective, leading in the worse case to widespread destruction but a large number of surviving paperclips.

This would clearly be a disaster. The common response is to take as a consequence that when we specify goals to programs, we should be much more careful about specifying what those goals are. However, we might find it difficult to formulate a set of goals that don’t admit some kind of loophole or paradox that, if pursued with mechanical single-mindedness, are either similarly narrowly destructive or self-defeating.

Suppose that, instead of trying to formulate a set of foolproof goals, we should find a way to admit to the program that the set of goals we’ve described is not comprehensive. We should aim for the capacity to add new goals with a procedural understanding that the list may never be complete. If done well, we would have a system that would couple this initial set of goals to the set of resources, operations, consequences, and stakeholders initially provided to it, with an understanding that those goals are only appropriate to the initial list and finding new potential means requires developing a richer understanding of potential ends.

How can this work? It’s easy to imagine such an algorithmic admission leading to paralysis, either from finding contradictory objectives that apparently admit no solution or an analysis/paralysis which perpetually requires no undiscovered goals before proceeding. Alternatively, stated incorrectly, it could backfire, with finding more goals taking the place of making more paperclips as it proceeds singlemindedly to consume resources. Clearly, a satisfactory superintelligence would need to reason appropriately about the goal discovery process.

There is a profession that has figured out a heuristic form of reasoning about goal discovery processes: designers. Designers have coined the phrase “the fuzzy front end” when talking about the very early stages of a project before anyone has figured out what it is about. Designers engage in low-cost elicitation exercises with a variety of stakeholders. They quickly discover who the relevant stakeholders are and what impacts their interventions might have. Adept designers switch back and forth rapidly from candidate solutions to analyzing the potential impacts of those designs, making new associations about the area under study that allows for further goal discovery. As designers undertake these explorations, they advise going slightly past the apparent wall of diminishing returns, often using an initial brainstorming session to reveal all of the “obvious ideas” before undertaking a deeper analysis. Seasoned designers develop an understanding when stakeholders are holding back and need to be prompted, or when equivocating stakeholders should be encouraged to move on. Designers will interleave a series of prototypes, experiential exercises, and pilot runs into their work, to make sure that interventions really behave the way their analysis seems to indicate.

These heuristics correspond well to an area of statistics and machine learning called nonparametric Bayesian inference. Nonparametric does not mean that there are no parameters, but instead that the parameters are not given, and that inferring that there are further parameters is part of the task. Suppose that you were to move to a new town, and ask around about the best restaurant. The first answer would definitely be new, but as one asked more, eventually you would start getting new answers more rarely. The likelihood of a given answer would also begin to converge. In some cases the answers will be more concentrated on a few answers, and in some cases the answers will be more dispersed. In either case, once we have an idea of how concentrated the answers are, we might see that a particular period of not discovering new answers might just be unlucky and that we should pursue further inquiry.

Asking why provides a list of critical features that can be used to direct different inquiries that fill out the picture. What’s the best restaurant in town for Mexican food? Which is best at maintaining relationships to local food providers/has the best value for money/is the tastiest/has the most friendly service? Designers discover aspects about their goals in an open-ended way, that allows discovery to act in quick cycles of learning through taking on different aspects of the problem. This behavior would work very well for an active learning formulation of relational nonparametric inference.

There is a point at which information gathering activities are less helpful at gathering information than attending to the feedback to activities that more directly act on existing goals. This happens when there is a cost/risk equilibrium between the cost of more discovery activities and the risk of making an intervention on incomplete information. In many circumstances, the line between information gathering and direct intervention will be fuzzier, as exploration proceeds through reversible or inconsequential experiments, prototypes, trials, pilots, and extensions that gather information while still pursuing the goals found so far.

From this perspective, many frameworks for assessing engineering discovery processes make a kind of epistemological error: they assess the quality of the solution from the perspective of the information that they have gathered, paying no attention to the rates and costs which that information was discovered, and whether or not the discovery process is at equilibrium. This mistake comes from seeing the problems as finding a particular point in a given search space of solutions, rather than taking the search space as a variable requiring iterative development. A superintelligence equipped to see past this fallacy would be unlikely to deliver us a universe of paperclips.

Having said all this, I think the nonparametric intuition, while right, can be cripplingly misguided without being supplemented with other ideas. To consider discovery analytically is to not discount the power of knowing about the unknown, but it doesn’t intrinsically value non-contingent truths. In my next essay, I will take on this topic.

For a more detailed explanation and an example of how to extend engineering design assessment to include nonparametric criteria, see The Methodological Unboundedness of Limited Discovery Processes. Form Academisk, 7:4.

Among transhumanists, Nick Bostrom is well-known for promoting the idea of ‘existential risks’, potential harms which, were they come to pass, would annihilate the human condition altogether. Their probability may be relatively small, but the expected magnitude of their effects are so great, so Bostrom claims, that it is rational to devote some significant resources to safeguarding against them. (Indeed, there are now institutes for the study of existential risks on both sides of the Atlantic.) Moreover, because existential risks are intimately tied to the advancement of science and technology, their probability is likely to grow in the coming years.

Contrary to expectations, Bostrom is much less concerned with ecological suicide from humanity’s excessive carbon emissions than with the emergence of a superior brand of artificial intelligence – a ‘superintelligence’. This creature would be a human artefact, or at least descended from one. However, its self-programming capacity would have run amok in positive feedback, resulting in a maniacal, even self-destructive mission to rearrange the world in the image of its objectives. Such a superintelligence may appear to be quite ruthless in its dealings with humans, but that would only reflect the obstacles that we place, perhaps unwittingly, in the way of the realization of its objectives. Thus, this being would not conform to the science fiction stereotype of robots deliberately revolting against creators who are now seen as their inferiors.

I must confess that I find this conceptualisation of ‘existential risk’ rather un-transhumanist in spirit. Bostrom treats risk as a threat rather than as an opportunity. His risk horizon is precautionary rather than proactionary: He focuses on preventing the worst consequences rather than considering the prospects that are opened up by whatever radical changes might be inflicted by the superintelligence. This may be because in Bostrom’s key thought experiment, the superintelligence turns out to be the ultimate paper-clip collecting machine that ends up subsuming the entire planet to its task, destroying humanity along the way, almost as an afterthought.

But is this really a good starting point for thinking about existential risk? Much more likely than total human annihilation is that a substantial portion of humanity – but not everyone – is eliminated. (Certainly this captures the worst case scenarios surrounding climate change.) The Cold War remains the gold standard for this line of thought. In the US, the RAND Corporation’s chief analyst, Herman Kahn — the model for Stanley Kubrick’s Dr Strangelove – routinely, if not casually, tossed off scenarios of how, say, a US-USSR nuclear confrontation would serve to increase the tolerance for human biological diversity, due to the resulting proliferation of genetic mutations. Put in more general terms, a severe social disruption provides a unique opportunity for pursuing ideals that might otherwise be thwarted by a ‘business as usual’ policy orientation.

Here it is worth recalling that the Cold War succeeded on its own terms: None of the worst case scenarios were ever realized, even though many people were mentally prepared to make the most of the projected adversities. This is one way to think about how the internet itself arose, courtesy the US Defense Department’s interest in maintaining scientific communications in the face of attack. In other words, rather than trying to prevent every possible catastrophe, the way to deal with ‘unknown unknowns’ is to imagine that some of them have already come to pass and redesign the world accordingly so that you can carry on regardless. Thus, Herman Kahn’s projection of a thermonuclear future provided grounds in the 1960s for the promotion of, say, racially mixed marriages, disability-friendly environments, and the ‘do more with less’ mentality that came to characterize the ecology movement.

Kahn was a true proactionary thinker. For him, the threat of global nuclear war raised Joseph Schumpeter’s idea of ‘creative destruction’ to a higher plane, inspiring social innovations that would be otherwise difficult to achieve by conventional politics. Historians have long noted that modern warfare has promoted spikes in innovation that in times of peace are then subject to diffusion, as the relevant industries redeploy for civilian purposes. We might think of this tendency, in mechanical terms, as system ‘overdesign’ (i.e. preparing for the worst but benefitting even if the worst doesn’t happen) or, more organically, as a vaccine that converts a potential liability into an actual benefit.

In either case, existential risk is regarded in broadly positive terms, specifically as an unprecedented opportunity to extend the range of human capability, even under radically changed circumstances. This sense of ‘antifragility’, as the great ‘black swan’ detector Nicholas Taleb would put it, is the hallmark of our ‘risk intelligence’, the phrase that the British philosopher Dylan Evans has coined for a demonstrated capacity that people have to make step change improvements in their lives in the face of radical uncertainty. From this standpoint, Bostrom’s superintelligence concept severely underestimates the adaptive capacity of human intelligence.

Perhaps the best way to see just how much Bostrom shortchanges humanity is to note that his crucial thought experiment requires a strong ontological distinction between humans and superintelligent artefacts. Where are the cyborgs in this doomsday scenario? Reading Bostrom reminds me that science fiction did indeed make progress in the twentieth century, from the world of Karl Čapek’s Rossum’s Universal Robots in 1920 to the much subtler blending of human and computer futures in the works of William Gibson and others in more recent times.

Bostrom’s superintelligence scenario began to be handled in more sophisticated fashion after the end of the First World War, popularly under the guise of ‘runaway technology’, a topic that received its canonical formulation in Langdon Winner’s 1977 Autonomous Technology: Technics out of Control, a classic in the field of science and technology of studies. Back then the main problem with superintelligent machines was that they would ‘dehumanize’ us, less because they might dominate us but more because we might become like them – perhaps because we feel that we have invested our best qualities in them, very much like Ludwig Feuerbach’s aetiology of the Judaeo-Christian God. Marxists gave the term ‘alienation’ a popular spin to capture this sentiment in the 1960s.

Nowadays, of course, matters have been complicated by the prospect of human and machine identities merging together. This goes beyond simply implanting silicon chips in one’s brain. Rather, it involves the complex migration and enhancement of human selves in cyberspace. (Sherry Turkle has been the premier ethnographer of this process in children.) That such developments are even possible points to a prospect that Bostrom refuses to consider, namely, that to be ‘human’ is to be only contingently located in the body of Homo sapiens. The name of our species – Homo sapiens – already gives away the game, because our distinguishing feature (so claimed Linnaeus) had nothing to do with our physical morphology but with the character of our minds. And might not such a ‘sapient’ mind better exist somewhere other than in the upright ape from which we have descended?

The prospects for transhumanism hang on the answer to this question. Aubrey de Grey’s indefinite life extension project is about Homo sapiens in its normal biological form. In contrast, Ray Kurzweil’s ‘singularity’ talk of uploading our consciousness into indefinitely powerful computers suggests a complete abandonment of the ordinary human body. The lesson taught by Langdon Winner’s historical account is that our primary existential risk does not come from alien annihilation but from what social psychologists call ‘adaptive preference formation’. In other words, we come to want the sort of world that we think is most likely, simply because that offers us the greatest sense of security. Thus, the history of technology is full of cases in which humans have radically changed their lives to adjust to an innovation whose benefits they reckon outweigh the costs, even when both remain fundamentally incalculable. Success in the face such ‘existential risk’ is then largely a matter of whether people – perhaps of the following generation – have made the value shifts necessary to see the changes as positive overall. But of course, it does not follow that those who fail to survive the transition or have acquired their values before this transition would draw a similar conclusion.