Tuesday, July 18, 2017

Response to a Conversation on “Intelligence”

I think much confusion is caused by a lack of clarity about the meaning of the word “intelligence”, and not least a lack of clarity about the nature of the thing(s) we refer to by this word. This is especially true when it comes to discussions of artificial intelligence (AI) and the risks it may pose. A recently published conversation between Tobias Baumann (blue text) and Lukas Gloor (orange text) contains a lot of relevant considerations on this issue, along with some discussion of my views on it, which makes me feel compelled to respond.

The statement that gave rise to the conversation was apparently this:

> Intelligence is the only advantage we have over lions.

My thoughts on which is that this is a simplistic claim. First, I take it that “intelligence” here means cognitive abilities. But cognitive abilities alone — a competent head on a body without legs or arms — will not allow one to escape from lions; it will only enable one to think of and regret all the many useful “non-cognitive” tools one would have liked to have. The sense in which humans have an advantage over other animals, in terms of what has enabled us to take over the world for better or worse, is that we have a unique set of tools — upright walk, vocal cords, hands with fine motor skills, and a brain that can acquire culture. This has enabled us, over time, to build culture, with which we have been able to develop tools that have enabled us to gain an advantage over lions, mostly in the sense of not needing to get near them, as that could easily get fatal, even given our current level of cultural sophistication and “intelligence”.

I could hardly disagree more with the statement that “the reason we humans rule the earth is our big brain”. To the extent we do rule the Earth, there are many reasons, and the brain is just part of the story, and quite a modest one relative to what it gets credit for (which is often all of it). I think Jacob Bronowski’s The Ascent of Man is worth reading for a more nuanced and accurate picture of humanity’s ascent to power than the “it’s all due to the big brain” one.

There is a pretty big threshold effect here between lions (and chimpanzees) and humans, where with a given threshold of intelligence, you're also able to reap all the benefits from culture. (There might be an analogous threshold for self-improvement FOOM benefits.)

The question is what “threshold of intelligence” means in this context. All humans do not reap all the same benefits from culture — some have traits and abilities that enable them to reap far more benefits than others. And many of these traits have nothing to do with “intelligence” in any usual sense. Good looks, for instance. Or a sexy voice.

And the same holds true for cognitive abilities in particular: it is more nuanced than what measurement along a single axis can capture. For instance, some people are mathematical geniuses, yet socially inept. There are many axes along which we can measure abilities, and what allows us to build culture is all these many abilities put together. Again, it is not, I maintain, a single “special intelligence thing”, although we often talk as though it were.

For this reason, I do not believe such a FOOM threshold along a single axis makes much sense. Rather, we see progress along many axes that, when certain thresholds are crossed, allows us to expand our abilities in new ways. For example, at the cultural level we may see progress beyond a certain threshold in the production of good materials, which then leads to progress in our ability to harvest energy, which then leads to better knowledge and materials, etc. A more complicated story with countless little specialized steps and cogs. As far as I can tell, this is the recurrent story of how progress happens, on every level: from biological cells to human civilization.

Magnus Vinding seems to think that because humans do all the cool stuff "only because of tools," innate intelligence differences are not very consequential.

I would like to see a quote that supports this statement. It is accurate to say that I think we do “all the cool stuff only because of tools”, because I think we do everything because of tools. That is, I do not think of that which we call “intelligence” as anything but the product of a lot of tools. I think it’s tools all the way down, if you will. I suppose I could even be considered an “intelligence eliminativist”, in that I think there is just a bunch of hacks; no “special intelligence thing” to be found anywhere. RNA is a tool, which has built another tool, DNA, which, among other things, has built many different brain structures, which are all tools. And so forth. It seems to me that the opposite position with respect to “intelligence” — what may be called “intelligence reification” — is the core basis of many worries about artificial intelligence take-offs.

It is not correct, however, that I think that “innate differences in intelligence [which I assume refers to IQ, not general goal-achieving ability] are not very consequential”. They are clearly consequential in many contexts. Yet IQ is far from being an exhaustive measure of all cognitive abilities (although it sure does say a lot), and cognitive abilities are far from being all that enables us to achieve the wide variety of goals we are able to achieve. It is merely one integral subset among many others.

This seems wrong to me [MV: also to me], and among other things we can observe that e.g. von Neumann’s accomplishments were so much greater than the accomplishments that would be possible with an average human brain.

I wrote a section on Von Neumann in my Reflections on Intelligence, which I will refer readers to. I will just stress, again, that I believe thinking of “accomplishments” and “intelligence” along a single axis is counterproductive. John Von Neumann was no doubt a mathematical genius of the highest rank. Yet with respect to the goal of world domination in particular, which is what we seem especially concerned about in this context, putting Von Neumann in charge hardly seems a recipe for success, but rather the opposite. As he reportedly said:
“If you say why not bomb them tomorrow, I say why not today? If you say today at five o' clock, I say why not one o' clock?”
To me, these do not seem to be the words of a brain optimized for taking over the world. If we want to look at such a brain, we should, by all appearances, rather peer into the skull of Putin or Trump (if it is indeed mainly their brain, rather than their looks, or perhaps a combination of many things, that brought them into power).

One might argue that empirical evidence confirms the existence of a meaningful single measure of intelligence in the human case. I agree with this, but I think it's a collection of modules that happen to correlate in humans for some reason that I don't yet understand.

I think a good analogy is a country’s GDP. It’s a single, highly informative measure, yet a nation’s GDP is a function of countless things. This measure predicts a lot, too. Yet it clearly also leaves out a lot of information. More than that, we do not seem to fear that the GDP of a country (or a city, or the indeed the whole world) will suddenly explode once it reaches a certain level. But why? (For the record, I think global GDP is a far better measure of a randomly selected human’s ability to achieve a wide variety of goals [of the kind we care about] than said person’s IQ is.)

> The "threshold" between chimps and humans just reflects the fact that all the tools, knowledge, etc. was tailored to humans (or maybe tailored to individuals with superior cognitive ability).
So there's a possible world full of lion-tailored tools where the lions are beating our asses all day?

Depending on the meaning of “lion-tailored tool” it seems to me the answer could well be “yes”. In terms of the history of our evolution, for instance, it could well be that a lion tool in the form of, say, powerful armor could have meant that humans were killed by them in high numbers rather than the other way around.

Further down you acknowledge that the difference is "or maybe tailored to individuals with superior cognitive ability" – but what would it mean for a tool to be tailored to inferior cognitive ability? The whole point of cognitive ability is to be good at make the most out of tool-shaped parts of the environment.

I suspect David Pearce might say that that’s a parochially male thing to say. One could also say that the whole point of cognitive abilities is to make others feel good — a drive/task that has no doubt played a large role both for human survival and the increase in our cognitive abilities and goal-achieving abilities in general, arguably just as great as “making the most out of tool-shaped parts of the environment”.

Second, I think the term “inferior cognitive ability” again overlooks that there are many dimensions along which we can measure cognitive abilities. Once again, take the mathematical genius who has bad social skills. How to best make tools — ranging from apps to statements to say to oneself — that improve the capabilities of such an individual seems likely to be different in significant ways from how to best make tools for someone who is, say, socially gifted and mathematically inept.

Magnus takes the human vs. chimp analogy to mean that intelligence is largely “in the (tool-and-culture-rich) environment.

I would delete the word “intelligence” and instead say that the ability to achieve goals is a product of a large set of tools, of which, in our case, the human brain is a necessary but, for virtually all of our purposes, insufficient subset.
Also, chimps display superior cognitive abilities to humans in some respects, so saying that humans are more “intelligent” than chimps, period, is, I think, misleading. The same holds true of our usual employment of the word “intelligence” in general, in my view.

My view implies that quick AI takeover becomes more likely as society advances technologically. Intelligence would not be in the tools, but tools amplify how far you can get by being more intelligent than the competition (this might be mostly semantics, though).

First, it should be noted that “intelligence” here seems to mean “cognitive abilities”, not “the ability to achieve goals”. This distinction must be stressed. Second, as hinted above, I think the dichotomy between “intelligence” (i.e. cognitive ability) on the one hand and “tools” on the other is deeply problematic. I fail to see in what sense cognitive abilities are not tools? (And by “cognitive abilities” I also mean the abilities of computer software.) And I think the causal arrows between the different tools that determine how things unfold are far more mutual than they are according to the story that “intelligence (some subset of cognitive tools) is that which will control all other tools”.
Third, for reasons alluded to above, I think the meaning of “being more intelligent than the competition” stands in need of clarification. It is far from obvious to me what it means. More cognitively able, presumably, but in what ways? What kinds of cognitive abilities are most relevant with respect to the task of taking over the world? And how might they be likely to be created? Relevant questions to clarify, it seems to me.

Some reasons not to think that “quick AI takeover becomes more likely as society advances technologically” include that other agents would be more capable (closer to notional limits of “capabilities”) the more technologically advanced society is, that there would be more technology learned about and mastered by others to learn about and master in order to take over, and finally that society presumably will learn more about the limits and risks of technology, including AI, the more technologically advanced it is, and hence know more about what to expect and how to counter it.

Friday, April 21, 2017

New Book: 'You Are Them'

What follows if we reject belief in any kind of non-physical soul and instead fully embrace what we know about the world? The main implication, this book argues, is a naturalization of personal identity and ethics. A radically different way of thinking about ourself.

“A precondition of rational behaviour is a basic understanding of the nature of oneself and the world. Any fusion of ethical and decision-theoretic rationality into a seamless package runs counter to some our deepest intuitions. But "You Are Them" makes a powerful case. Magnus Vinding's best book to date. Highly recommended.”
— David Pearce, co-founder of The Neuroethics Foundation, co-founder of World Transhumanist Association / Humanity+, and author of The Hedonistic Imperative and The Anti-Speciesist Revolution.

Free download:

Monday, January 30, 2017

Reasons to Focus on Values as Our Main Cause

Given decent clarity about our fundamental values, the all-important question becomes: what causes and interventions optimize those values?

In this post I shall present some of the reasons in favor of focusing directly on fundamental values in these regards. That is, reasons why a good way to optimize our fundamental values is to reflect on, argue for, and work out the implications of, these values themselves. We may call it “the values cause”.

So why is this sensible? In short, because of the unmatched importance of fundamental values. Our fundamental values comprise the most important and fundamental element in our notional ‘tree of ought’. They are what determine the sensibility of any cause and intervention we may take part in, and hence what any reasonable choice of causes and interventions should be based on. An important implication of this is that the (expected) sensibility of any cause or intervention cannot be greater than the (expected) sensibility of our fundamental values. For example, if we have 90 percent confidence in our fundamental values, and then choose a cause or intervention based on these, we cannot have greater confidence in the sensibility of this cause or intervention than 90 percent. Indeed, 90 percent would be the level of credence we should have if we were 100 percent sure that the specific cause or intervention optimizes our fundamental values perfectly; a degree of confidence we will of course never have about any cause or intervention. Thus, we must have greater confidence in the sensibility of our values than in the sensibility of any action taken to optimize those values.

In a world where little seems certain, this relationship is worth taking note of. It means that, of all our beliefs pertaining to ethics, our fundamental values are — or at least should be — what we are the most certain about. This also, I would argue, makes them the thing we should be most confident about arguing for in our pursuit for a better world.

Getting Others to Join Us in the Best Way

Arguing directly for our fundamental values rather than causes and interventions derived from those values also, if done successfully, has the benefit of bringing people down alongside with us in the basement of fundamental values from which the sensibility of causes and interventions must be assessed. In other words, not only do we argue for that which we are most confident about when we argue for our fundamental values, we also invite people to join us in the best possible place: our core base from which we ourselves are trying to find out which causes and interventions that best optimize our values. Having more minds to help optimize our tree of ought from the bottom up seems very positive, and the deeper down they join us, the better.

And even if we do not manage to convince people to fully share our fundamental values, arguing for our values likely does at least make them update somewhat in our direction, which, given the large changes in practical implications that can result from small changes in fundamental values, could well be far more valuable than convincing others to agree with specific causes or interventions we may favor. Not least because it might make them more likely to agree with these interventions, which then leads to another, albeit somewhat speculative, reason to focus on fundamental values in practice.

For one counterargument to the argument I have made above is that people might be more receptive to arguments for specific causes or interventions than they are to the fundamental values that recommend those causes. Yet I think the opposite is generally true. I suspect it is generally easier to convince people of one’s fundamental values, or at least make them update significantly toward them, than it is to convince others of one’s most favored causes or interventions.

For example, it seems much easier to convince people that extreme suffering is of great significance and worth reducing than it does to convince them that they should go vegan. And in order to convince people of the importance of a given cause or intervention, it might well require bottom-up reasoning from first principles — in this case, fundamental values — to see the reasonableness of that given cause or intervention. It can indeed seem naive for us to think, after we ourselves have come to support a given intervention based on an underlying value framework, that we should then be able to convince others to support that intervention without communicating this very framework that led us to consider that intervention a sensible one ourselves.

So not only may people be more receptive to our fundamental values than the causes and interventions we support (an admittedly speculative “may”), it might also be that arguing for our fundamental values is the best way to bring people on board with our preferred causes and interventions in many cases, due to the likely necessity of following a chain of inferential steps. And again, if we invite others to try to step in our own inferential footsteps, we might be lucky to have them spot missteps. In this way, we enable others to help us find even better causes and interventions based on our fundamental values than the ones we presently focus on.

An instructive example of failure here, I think, is found in the strategy of most anti-natalists. The vast majority of anti-natalists seems to share the fundamental goal of reducing net suffering, yet their advocacy tends to focus exclusively on anthropocentric anti-natalism — a highly specific and narrow intervention. They appear to confidently assume that this is the best way to reduce suffering in the world, rather than focusing on the fundamental goal of reducing suffering itself, and encouraging discussion and research about how to best do this. If anti-natalists focused more on the latter, they would likely have more success, both by inspiring more people to take their fundamental values into consideration, and by inviting these others (and themselves not least) to think deeper about which other ideas they might be able to spread that could be more conducive to the goal of reducing suffering than the idea of anthropocentric anti-natalism (which seems rather unlikely to be the best idea to push in order to reduce the most suffering in our future light cone).

Reducing Moral Uncertainty/Updating our Fundamental Values

Another reason to focus on fundamental values is our own moral uncertainty. For given that we may be wrong about what we value, whether in a strong moral realist sense or an idealized personal preferences sense (or anything in-between), we should be keen on updating our fundamental values. And reflecting on and discussing them openly is likely among the best ways to do so. To restate this important point once more: given the immense importance of fundamental values, even small updates here could be among the most significant moves we could make.

And fundamental values do appear quite open to change. Indeed, values are contagious and subject to cultural influence to a great extent, as a map of people’s religious beliefs around the world reveals (such beliefs are undeniably closely tied to beliefs about fundamental values). Arguably, our values are subject to change and cultural influence to a significantly greater extent than technological progress is (cf. What Technology Wants by Kevin Kelly), which may be harder to influence and hence might be less of a leverage point for impacting change than focusing on values is. To put things crudely, technologies tend to be developed regardless, while how they are used generally seems more contingent. And arguing values seems among the best ways to impact how we use our powers.

Values are, to a first approximation, ideas, and ideas tend to be updatable and spreadable. In my own case, I used to not care about ethics at all, then I became a classical utilitarian, and eventually I updated toward negative utilitarianism and suffering-focused ethics as I came upon arguments in their favor. We should expect similar changes to be possible in others, and in ourselves, as we learn more and keep on updating our beliefs.

Better Cooperation?

Not only would we all benefit from having our moral uncertainty reduced/our moral views updated, which is valuable in itself; it seems that we should also expect to benefit from the greater convergence on fundamental values that is likely to follow from mutual discussion and updating on them, even if the magnitude of this updating is small. The reason this is beneficial is that such convergence likely reduces the level of friction in our efforts of cooperation, and on virtually any set of fundamental values, success in achieving the most valuable/least disvaluable future seems to rest on humanity’s ability to cooperate. This makes such cooperation a high priority for all of us. While somewhat speculative, this consideration in favor of convergence on fundamental values, and hence, arguably, in favor of mutual discussion and updating on them, is important to factor in as well.

Fundamental Values and AI Safety

I have tried elsewhere to explain why I think the Bostromesque framing of the issue of “AI safety” is unsound. But even assuming it isn’t, I would argue that fundamental values should likely still be our main focus, the reason being that we have little clarity or consensus about which values to load a notional super-powerful AI with in the first place (and I should note that I find using the term “AI” in this unqualified way highly objectionable — for what does it refer to?).

The main problem claimed to exist within the cause of “AI safety” is the so-called control problem, particularly what is called the value loading problem: how do we load “an AI” with good values? What seems implicit in such a question, however, is that we have a fairly high level of consensus about what constitutes good values. Yet when we look at modern discussions of ethics, especially population ethics, we find that this is not the case — indeed, we see that strong certainty about what constitutes good values is hardly reasonable for any of us. This suggests that we have a lot to clarify before we start programming, namely what values we estimate to be ideal. We must have decent clarity about what constitutes good values before we can implement such values — in anything we do or create. We must solve the values problem before we can solve any notional values loading problem.

For an example of an unresolved question, take the following, in my view critically important one: What are the theoretical upper bounds of the ratio between happiness and suffering in a functional civilization, and can the suffering it contains, if there is any, ever be outweighed by the happiness? At the very least, these questions deserve consideration, yet they are hardly ever asked (not to mention the loud silence on the issue of the utilitronium shockwave that would seem, at least in theory, the main corollary of classical utilitarianism; are classical utilitarians obliged to work toward such a shockwave, contra the present dominant view on “AI ethics”, which seems to be an anthropocentric preference utilitarianism of sorts [see the note on the goals of AI builders below], which appears very bad from a classical utilitarian perspective, at least compared to a utilitronium shockwave?).

Another example would be the aforementioned subject of population ethics, where many ethicists believe that we should bring about the greatest number of happy beings we can, while many others believe that adding an additional happy life to any given population has no intrinsic value. Given such a near-maximal divergence of views on an issue like this, what does it mean to say that we should build a system that does what humans want? What could it mean?

This issue of value implementation underscores the importance of convergence on values, as that would likely make any such project of implementation go smoother (an example of the general point about human cooperation made above). It could well be that trying to make mutual value updating happen among those who try to build the world of tomorrow — both in the realm of software and in other realms — is the better way to implement our values than to bargain at the level of the direct implementation with others who have more divergent values; that is, more divergent values than they would have had if we had put more effort into arguing for our fundamental values directly.

In other words, if humans are going to program values into “an AI”, the best way to impact the outcome of that process could well be to impact the values of these humans and humanity in general. Not least because the goal many of these AI researchers aim to implement in tomorrow’s software simply is “that which humans want” (Paul Christiano: “I want to see a future where AI systems help humans get what they want […]”; OpenAI: “We believe AI should be an extension of individual human wills […]”; the so-called Partnership on AI by Google, Facebook, Microsoft, Amazon, IBM, and Apple seems to have essentially the same goal).

I have argued that the future of “intelligence” on Earth and beyond will be shaped by a collective, distributed process comprised of what many agents do, which also holds true in the case of a software takeover. And the best way to impact such a collective process in a positive direction is, I think, most likely one where we try to impact values directly.

Whether we deem it the main cause or not, it seems clear to me that “the values cause” must be considered a main cause, and an utmost neglected one at that. Our altruistic efforts ought to be informed by careful considerations based on first principles, those principles being our fundamental values. Yet for the most part, this isn’t what we are doing. If it were, we would have better clarity about what exactly our first principles are; at the very least, we would be aware of the fact that we do not have such clarity in the first place. Instead, we go with intuition and vague ideas like “more happiness, less suffering”, believing that to be good enough for all practical purposes. As I have tried to argue, this is far from the case.

Saying that we should focus much more on fundamental values is not, however, to say that we should not focus on other specific causes and interventions that follow from those values, nor that we should not do advocacy for these. I think we should. What I think it does imply, however, is that we should try to communicate our (carefully considered) fundamental values in such advocacy. For instance, when doing concrete anti-speciesist advocacy, we should do so by phrasing it in terms of our fundamental values, e.g. concern for sentience and involuntary suffering. Thereby, we both do advocacy for a (relatively) specific cause recommended by our fundamental values and those values themselves, which invites people to consider and discuss both. It does not have to be a matter of either focusing on values or focusing on “doing”. We can encourage people to reflect on fundamental values with our doing.