Superintelligence 5: Forms of Superintelligence

KatjaGrace

22 Superintelligence 5: Forms of Superintelligence

14th Oct 2014

6 min read

22

This is part of a weekly reading group on Nick Bostrom's book, Superintelligence. For more information about the group, and an index of posts so far see the announcement post. For the schedule of future topics, see MIRI's reading guide.

Welcome. This week we discuss the fifth section in the reading guide: Forms of superintelligence. This corresponds to Chapter 3, on different ways in which an intelligence can be super.

This post summarizes the section, and offers a few relevant notes, and ideas for further investigation. Some of my own thoughts and questions for discussion are in the comments.

There is no need to proceed in order through this post, or to look at everything. Feel free to jump straight to the discussion. Where applicable and I remember, page numbers indicate the rough part of the chapter that is most related (not necessarily that the chapter is being cited for the specific claim).

Reading: Chapter 3 (p52-61)

Summary

A speed superintelligence could do what a human does, but faster. This would make the outside world seem very slow to it. It might cope with this partially by being very tiny, or virtual. (p53)
A collective superintelligence is composed of smaller intellects, interacting in some way. It is especially good at tasks that can be broken into parts and completed in parallel. It can be improved by adding more smaller intellects, or by organizing them better. (p54)
A quality superintelligence can carry out intellectual tasks that humans just can't in practice, without necessarily being better or faster at the things humans can do. This can be understood by analogy with the difference between other animals and humans, or the difference between humans with and without certain cognitive capabilities. (p56-7)
These different kinds of superintelligence are especially good at different kinds of tasks. We might say they have different 'direct reach'. Ultimately they could all lead to one another, so can indirectly carry out the same tasks. We might say their 'indirect reach' is the same. (p58-9)
We don't know how smart it is possible for a biological or a synthetic intelligence to be. Nonetheless we can be confident that synthetic entities can be much more intelligent than biological entities.
1. Digital intelligences would have better hardware: they would be made of components ten million times faster than neurons; the components could communicate about two million times faster than neurons can; they could use many more components while our brains are constrained to our skulls; it looks like better memory should be feasible; and they could be built to be more reliable, long-lasting, flexible, and well suited to their environment.
2. Digital intelligences would have better software: they could be cheaply and non-destructively 'edited'; they could be duplicated arbitrarily; they could have well aligned goals as a result of this duplication; they could share memories (at least for some forms of AI); and they could have powerful dedicated software (like our vision system) for domains where we have to rely on slow general reasoning.

Notes

This chapter is about different kinds of superintelligent entities that could exist. I like to think about the closely related question, 'what kinds of better can intelligence be?' You can be a better baker if you can bake a cake faster, or bake more cakes, or bake better cakes. Similarly, a system can become more intelligent if it can do the same intelligent things faster, or if it does things that are qualitatively more intelligent. (Collective intelligence seems somewhat different, in that it appears to be a means to be faster or able to do better things, though it may have benefits in dimensions I'm not thinking of.) I think the chapter is getting at different ways intelligence can be better rather than 'forms' in general, which might vary on many other dimensions (e.g. emulation vs AI, goal directed vs. reflexive, nice vs. nasty).
Some of the hardware and software advantages mentioned would be pretty transformative on their own. If you haven't before, consider taking a moment to think about what the world would be like if people could be cheaply and perfectly replicated, with their skills intact. Or if people could live arbitrarily long by replacing worn components.
The main differences between increasing intelligence of a system via speed and via collectiveness seem to be: (1) the 'collective' route requires that you can break up the task into parallelizable subtasks, (2) it generally has larger costs from communication between those subparts, and (3) it can't produce a single unit as fast as a comparable 'speed-based' system. This suggests that anything a collective intelligence can do, a comparable speed intelligence can do at least as well. One counterexample to this I can think of is that often groups include people with a diversity of knowledge and approaches, and so the group can do a lot more productive thinking than a single person could. It seems wrong to count this as a virtue of collective intelligence in general however, since you could also have a single fast system with varied approaches at different times.
For each task, we can think of curves for how performance increases as we increase intelligence in these different ways. For instance, take the task of finding a fact on the internet quickly. It seems to me that a person who ran at 10x speed would get the figure 10x faster. Ten times as many people working in parallel would do it only a bit faster than one, depending on the variance of their individual performance, and whether they found some clever way to complement each other. It's not obvious how to multiply qualitative intelligence by a particular factor, especially as there are different ways to improve the quality of a system. It also seems non-obvious to me how search speed would scale with a particular measure such as IQ.
How much more intelligent do human systems get as we add more humans? I can't find much of an answer, but people have investigated the effect of things like team size, city size, and scientific collaboration on various measures of productivity.
The things we might think of as collective intelligences - e.g. companies, governments, academic fields - seem notable to me for being slow-moving, relative to their components. If someone were to steal some chewing gum from Target, Target can respond in the sense that an employee can try to stop them. And this is no slower than an individual human acting to stop their chewing gum from being taken. However it also doesn't involve any extra problem-solving from the organization - to the extent that the organization's intelligence goes into the issue, it has to have already done the thinking ahead of time. Target was probably much smarter than an individual human about setting up the procedures and the incentives to have a person there ready to respond quickly and effectively, but that might have happened over months or years.

In-depth investigations

If you are particularly interested in these topics, and want to do further research, these are a few plausible directions, some inspired by Luke Muehlhauser's list, which contains many suggestions related to parts of Superintelligence. These projects could be attempted at various levels of depth.

Produce improved measures of (substrate-independent) general intelligence. Build on the ideas of Legg, Yudkowsky, Goertzel, Hernandez-Orallo & Dowe, etc. Differentiate intelligence quality from speed.
List some feasible but non-realized cognitive talents for humans, and explore what could be achieved if they were given to some humans.
List and examine some types of problems better solved by a speed superintelligence than by a collective superintelligence, and vice versa. Also, what are the returns on “more brains applied to the problem” (collective intelligence) for various problems? If there were merely a huge number of human-level agents added to the economy, how much would it speed up economic growth, technological progress, or other relevant metrics? If there were a large number of researchers added to the field of AI, how would it change progress?
How does intelligence quality improve performance on economically relevant tasks?

If you are interested in anything like this, you might want to mention it in the comments, and see whether other people have useful thoughts.

How to proceed

This has been a collection of notes on the chapter. The most important part of the reading group though is discussion, which is in the comments section. I pose some questions for you there, and I invite you to add your own. Please remember that this group contains a variety of levels of expertise: if a line of discussion seems too basic or too incomprehensible, look around for one that suits you better!

Next week, we will talk about 'intelligence explosion kinetics', a topic at the center of much contemporary debate over the arrival of machine intelligence. To prepare, read Chapter 4, The kinetics of an intelligence explosion (p62-77). The discussion will go live at 6pm Pacific time next Monday 20 October. Sign up to be notified here.

SuperintelligenceReading GroupAI

Personal Blog

22

New Comment

Rendering 0/114 comments, sorted by

top scoring

(show more) Click to highlight new comments since: Today at 5:28 PM

Some comments are truncated due to high volume. (⌘F to expand all)Change truncation settings

Moderation Log

22 Superintelligence 5: Forms of Superintelligence

by KatjaGrace

14th Oct 2014

6 min read

114

22

Welcome. This week we discuss the fifth section in the reading guide: Forms of superintelligence. This corresponds to Chapter 3, on different ways in which an intelligence can be super.

This post summarizes the section, and offers a few relevant notes, and ideas for further investigation. Some of my own thoughts and questions for discussion are in the comments.

Reading: Chapter 3 (p52-61)

Summary

A speed superintelligence could do what a human does, but faster. This would make the outside world seem very slow to it. It might cope with this partially by being very tiny, or virtual. (p53)
A collective superintelligence is composed of smaller intellects, interacting in some way. It is especially good at tasks that can be broken into parts and completed in parallel. It can be improved by adding more smaller intellects, or by organizing them better. (p54)
A quality superintelligence can carry out intellectual tasks that humans just can't in practice, without necessarily being better or faster at the things humans can do. This can be understood by analogy with the difference between other animals and humans, or the difference between humans with and without certain cognitive capabilities. (p56-7)
These different kinds of superintelligence are especially good at different kinds of tasks. We might say they have different 'direct reach'. Ultimately they could all lead to one another, so can indirectly carry out the same tasks. We might say their 'indirect reach' is the same. (p58-9)
We don't know how smart it is possible for a biological or a synthetic intelligence to be. Nonetheless we can be confident that synthetic entities can be much more intelligent than biological entities.
1. Digital intelligences would have better hardware: they would be made of components ten million times faster than neurons; the components could communicate about two million times faster than neurons can; they could use many more components while our brains are constrained to our skulls; it looks like better memory should be feasible; and they could be built to be more reliable, long-lasting, flexible, and well suited to their environment.
2. Digital intelligences would have better software: they could be cheaply and non-destructively 'edited'; they could be duplicated arbitrarily; they could have well aligned goals as a result of this duplication; they could share memories (at least for some forms of AI); and they could have powerful dedicated software (like our vision system) for domains where we have to rely on slow general reasoning.

Notes

This chapter is about different kinds of superintelligent entities that could exist. I like to think about the closely related question, 'what kinds of better can intelligence be?' You can be a better baker if you can bake a cake faster, or bake more cakes, or bake better cakes. Similarly, a system can become more intelligent if it can do the same intelligent things faster, or if it does things that are qualitatively more intelligent. (Collective intelligence seems somewhat different, in that it appears to be a means to be faster or able to do better things, though it may have benefits in dimensions I'm not thinking of.) I think the chapter is getting at different ways intelligence can be better rather than 'forms' in general, which might vary on many other dimensions (e.g. emulation vs AI, goal directed vs. reflexive, nice vs. nasty).
Some of the hardware and software advantages mentioned would be pretty transformative on their own. If you haven't before, consider taking a moment to think about what the world would be like if people could be cheaply and perfectly replicated, with their skills intact. Or if people could live arbitrarily long by replacing worn components.
The main differences between increasing intelligence of a system via speed and via collectiveness seem to be: (1) the 'collective' route requires that you can break up the task into parallelizable subtasks, (2) it generally has larger costs from communication between those subparts, and (3) it can't produce a single unit as fast as a comparable 'speed-based' system. This suggests that anything a collective intelligence can do, a comparable speed intelligence can do at least as well. One counterexample to this I can think of is that often groups include people with a diversity of knowledge and approaches, and so the group can do a lot more productive thinking than a single person could. It seems wrong to count this as a virtue of collective intelligence in general however, since you could also have a single fast system with varied approaches at different times.
For each task, we can think of curves for how performance increases as we increase intelligence in these different ways. For instance, take the task of finding a fact on the internet quickly. It seems to me that a person who ran at 10x speed would get the figure 10x faster. Ten times as many people working in parallel would do it only a bit faster than one, depending on the variance of their individual performance, and whether they found some clever way to complement each other. It's not obvious how to multiply qualitative intelligence by a particular factor, especially as there are different ways to improve the quality of a system. It also seems non-obvious to me how search speed would scale with a particular measure such as IQ.
How much more intelligent do human systems get as we add more humans? I can't find much of an answer, but people have investigated the effect of things like team size, city size, and scientific collaboration on various measures of productivity.
The things we might think of as collective intelligences - e.g. companies, governments, academic fields - seem notable to me for being slow-moving, relative to their components. If someone were to steal some chewing gum from Target, Target can respond in the sense that an employee can try to stop them. And this is no slower than an individual human acting to stop their chewing gum from being taken. However it also doesn't involve any extra problem-solving from the organization - to the extent that the organization's intelligence goes into the issue, it has to have already done the thinking ahead of time. Target was probably much smarter than an individual human about setting up the procedures and the incentives to have a person there ready to respond quickly and effectively, but that might have happened over months or years.

In-depth investigations

Produce improved measures of (substrate-independent) general intelligence. Build on the ideas of Legg, Yudkowsky, Goertzel, Hernandez-Orallo & Dowe, etc. Differentiate intelligence quality from speed.
List some feasible but non-realized cognitive talents for humans, and explore what could be achieved if they were given to some humans.
List and examine some types of problems better solved by a speed superintelligence than by a collective superintelligence, and vice versa. Also, what are the returns on “more brains applied to the problem” (collective intelligence) for various problems? If there were merely a huge number of human-level agents added to the economy, how much would it speed up economic growth, technological progress, or other relevant metrics? If there were a large number of researchers added to the field of AI, how would it change progress?
How does intelligence quality improve performance on economically relevant tasks?

If you are interested in anything like this, you might want to mention it in the comments, and see whether other people have useful thoughts.

How to proceed

SuperintelligenceReading GroupAI

Personal Blog

22

Mentioned in

54We are headed into an extreme compute overhang

36Three Kinds of Competitiveness

31Superintelligence reading group

15"Wide" vs "Tall" superintelligence

New Comment

Rendering 0/114 comments, sorted by

top scoring

(show more) Click to highlight new comments since: Today at 5:28 PM

Some comments are truncated due to high volume. (⌘F to expand all)Change truncation settings

Moderation Log

More from KatjaGrace

Curated and popular this week

114Comments

114

Comment Permalink

satt12y80

When I point out the low-hanging fruit effect to LWers, I do usually get a lot of agreement (and it is appreciated!) but I am starting to wish that someone would dig up some strong contrary evidence.

When the topic of apparent genius deficits and scientific stagnation comes up, people often present multiple explanations, like

intrinsic difficulties in scaling scientific activity
failure to identify/recognize contemporary scientific successes
no more low-hanging fruit
bureaucratization and institutional degradation

but tend to present only anecdotal evidence for each — myself included. And I'm not sure that can be helped; I don't know of readily available evidence which powerfully discriminates between the different explanations.

PhilGoetz has data on scientific & technological progress, but I get the impression that much of it's basically time series of counts of inventions & discoveries, which would establish only the whats and not the whys. Likewise, I think I could substantiate my January comment that cohort explains a substantial part of the variation in scientific eminence. And when I scraped together the data, ran the big regression, and found that birth year accounted for (suppose) 30% of the variance in eminence, that wouldn't refute any of the potential explanations for why cohort correlated with eminence.

A partisan of the scaling hypothesis might say, "Obviously, as science gets bigger over time, it gets less efficient; more recently born scientists just lost the birth year draw".

Someone arguing that scientific stagnation is illusory might say, "Obviously, this is a side effect of overlooking more recent scientific geniuses; scientists are working as effectively as before but we don't recognize that thanks to increasing specialization, or our own complacency, or the difficulty of picking out individual drops from the flood of brilliance, or the fact that we only recognize greatness decades after the fact".

I would say, if I were the kind of person who threw the word "obviously" around willy-nilly, "How many times do you expect general relativity to be invented? Obviously, there are only so many simple but important problems to work on, and when we turn to much harder problems, we make slower and more incremental progress".

Someone most concerned with institutional degradation might say, "Obviously, as science has become more bureaucratic and centralized, that's rendered it more careerist, risk-averse & narrow-minded and less ambitious, so of course later generations of scientists would end up being less eminent, because they're not tackling big scientific questions like they did before".

And we don't get anywhere because each explanation is broadly consistent with the observed facts, and each seems obvious to someone.

PhilGoetz12y00

when I scraped together the data, ran the big regression, and found that birth year accounted for (suppose) 30% of the variance in eminence, that wouldn't refute any of the potential explanations for why cohort correlated with eminence

I'd love to see that data & analysis! Did you post it somewhere? Can you email it to me at gmail?

I think there was a LW post years ago saying that the word "obviously" is only used to cover up the fact that something isn't obvious, and I agree with that more every year.

The evidence against the low-hanging fru... (read more)

4Luke_A_Somers12y

I would put forth three lines of argument that might help. First, what we consider a significant development is put in relation to its context. So, we naturally end up picking out the top-level entities and not the second-layer entities, let alone the third, fourth, fifth... modern science may have the same number of top-level discoveries, but these are underpinned by many more layers of discovery than earlier discoveries. Second, let's stop thinking about the jump from Aristotle to modern science for a minute. Let's think about the jump from Novoselov and Geim's discovery of Graphene to today. In their first paper, they made graphene, put it on a substrate, hooked a few wires up to it, and did low-temperature transport measurements. Worth a nobel prize. Outside of the insight that led to it, pretty simple. Not everyone could do it, but many could. In the following years, a bunch of progressively trickier experiments were performed. As of two years ago, our clearest path to a publishable research paper in this area was to make an enormous pristine sheet of graphene, position a layer of boron nitride on top of it, position another layer of graphene on top of that in such a way that it didn't short to the first piece, place a bunch of wires in very specific locations on this sandwich, then destroy the substrate that was all sitting on, all done so cleanly that it was smooth on every surface. This was insanely hard. This is also only a little trickier than normal for experiments in the field these days. The low-hanging fruit has been taken, here. And it's not simply that other people took it and we're looking at sour grapes. I took some of that low-hanging fruit. There was a simple experiment to do, I did it, published it, and now I cannot do an experiment that simple again in this sub-field. My next experiment was substantially more complicated. The next experiment after that was far more complicated still. Yes, we stand on the shoulders of giants, but we are i

1SteveG12y

Funding agencies these days fund people who get PhDs. To get a PhD, 90% of the time you need to generate a meaningful result of some kind within a limited time horizon. Scientists who go big and create nothing do not graduate and do not get funded later. What some of them do learn to do is to manage several projects at the same time. They diversify, working on some big ideas which may fail, but insuring a steady stream of results by also working through some lesser issues with a higher probability of success. It's true that you can have a career stringing together nothing but small wins, but contrary to popular belief, the funding agencies (who rely on PhD scientific peer review committees) do fund many "high-risk, high-reward" projects. In the private sector, for example, drug discovery projects have a vast failure rate but are funded nonetheless. Science funders do understand the biases toward career safety and are trying (imperfectly) to adjust for them.

See in context