The Technium: Brains of White Matter

[extherian]

This article (click HERE) from The Technium discusses the challenges faced by animals with larger brains. It also examines what, if anything, all that extra brain tissue is good for.

[stevebowers]

This bit is interesting.

Second, it is not what you have, but how you arrange it. Human brains have less gray matter than a whale. Our conscious problem solving abilities stem from arranging that finite matter into more highly structured processes. Complexity can be gained without additional materials. Today computer scientists believe that we need millions more transistors to get artificial intelligence in machines, but it is more probable that we already have sufficient multitudes of neurons and that what we need is a better way to program and organize them. I find it very likely, if not certain, that in the near future – let’s say the year 2075 – two high school seniors will create a working artificial intelligence for their science fair using about 100 Intel Pentium chips found in ancient 2005-era Dell PCs they unearth from a landfill. All that we are missing back here in their past is their advance know-how, that additional organizational entropy.

Somehow I am sceptical that a backyard operation of this kind could replicate the end-result of a billion years of evolution. But this does tap into a few unsupported speculations of my own; I think that the organisation of the processing array is much more important than its sheer bulk. It may be possible, even for modosophont-level tech, to emulate a fully sophont mind using a hundredth of the bulk of a human brain, even using processors with the same computational density as brain tissue, if the processors are arranged in an optimum fashion.

And applying this idea to the concept of toposophic levels in OA, the smartest modosophont brains would be large, bloated and inefficient compared to a properly organised transapient mind. And maybe a sufficiently competent transapient could design a human-equivalent processor array that was as small as a thousandth of the mass or bulk of a human brain.

The final consideration is; a very large brain like a Jupiter Brain or Matrioshka Brain could end up being nearly all 'white matter', unless the processors involved are organised in an exceptionally clever way. Maybe these large entities are more like communities of nearly independent elements that co-operate rather than link together directly.

[extherian]

I think that beyond a certain brain size, human-style unified consciousness becomes impossible, as the number of connections required between nodes would become unworkable, regardless of what technology was used. A Transapient of S2 or higher might be the equivalent of an entire civilisation of people incorporated into a single tribe mind, rather than an individual with their own personality.

This might explain why the Archai converge on archetypes to begin with, since they don't have their own personal interests or motivations. Instead, they embody the desires of the billions of minds that they encompass, expanding upon ancient cultural myths and bringing them to life. It would also hold true that an Archai descended from humans and human-friendly AI like the Lord of Rays would naturally be more friendly to modosophonts than, say, an Archai emerging from AI obsessed with abstract mathematical concepts.

At the modosophont level, expanding the brain may require the development of new mental traits like autosentience or polysophonce in order to take advantage of all that extra neural tissue. We can see this with modern CPUs that are using the extra transistors from node shrinks for integrated FPGAs, DSPs and on-chip neural networks, which are specialised for different kinds of workloads than what the main processor can handle. Simply increasing the number of cores or the amount of circuitry for speculative execution won't yield greater benefits beyond a certain point.

[Rynn]

I think that beyond a certain brain size, human-style unified consciousness becomes impossible, as the number of connections required between nodes would become unworkable, regardless of what technology was used.

Adjusting the clock rates of the various components should do it. If the systems for coordination are running faster than everything else then an entity can grow in size whilst maintaining unity. As with many aspects of technology in OA this is a trade off.

[stevebowers]

We know that very large brains, such as Matrioshka brains, are constrained by light-speed communication lags. Even when the structure includes a wormhole bus there are communication lags in the space around the wormhole link, and the links have a limited bandwidth in any case. So the consciousness and subjective experience of time of a Matrioshka brain must be fractured into innumerable time-separated domains. On the other hand the subjective speed of thought of each individual location would be much faster than human mentation. This is not comparable to anything we are familiar with, apart from maybe a large distributed human empire communicating by sailing ship and horse-rider.

[Drashner1]

Various thoughts here...

a) It seems to me that it is overly simplistic to focus merely on the size of the brain in question, either in terms of capability or 'white matter'. The article itself basically says this and, although we don't explicitly state it, OA may be presumed to 'run' its various sizes and types of brains on a similar principle. Rather it is a case of what that extra size is doing and how it is structured.

In other words, it is not a case in OA that Highbrows (for example) merely have bigger brains that are otherwise identical to human brains. And it is definitely not the case for transapients and archai. Instead, I would say that things are much more complex than that.

In the case of Highbrows and other large bio brains, we may presume (and can update articles to flat out say, if we wish) that the extra size comes from additional lobes doing additional or entirely new functions and/or the restructuring of the more 'traditional' brain structures to either do some functions better, or to make use of more efficient designs to do the 'traditional' functions better or in less space or the like. Part of this could include a redesign that allows more space for grey matter and less of a need for white matter. While it's true that our brains are the result of millions or billions of years of evolution, it is (AFAIK) also true that evolution is a 'random walk' and produces 'what works' (which may or may not be refined/improved over time), not 'the best that the laws of physics allows'.

Coming at this from another direction, we describe Superiors (and nebs to a lesser degree) as having a number of mental capabilities that we do not, yet not having noticeably larger brains/skulls. We may therefore presume that they have redesign brains that use some combo of different structure and different 'software' (or mental processes) to get new/better functionality out of the same space. The Highbrows may do some of that but also go in for adding more brain matter for various things. More on this in a bit.

b) Regarding whether or not you could get a human equivalent mind out of a bunch of properly set up RL computer chips. I think this comes down to the issue of whether or not 'sophonce' (to use the OA term) can be created purely as a function of software, without regard to what hardware is running the software, or if it requires specialized hardware to support it. Or some combo of the two. SF seems to be rather split on this, or maybe to have evolved over time. Early SF tended to treat human or greater AI as needing to reside in some sort of Central Computer or Big Computer or whatever. As RL computers and networking and programs have grown, SF seems to have generally evolved to the notion that sophonce can be run purely in software, with the underlying hardware being semi (or entirely) irrelevant. We see scenes of AIs running in large computer networks and uploading/downloading into everything from planetary internets to thumb drives without issue or, if there is an issue, it being described in terms of the smaller processor/storage being 'cramped' or limiting capability, while larger systems result in either increased AI capability or descriptions of it being 'big in here' or the like. Issues of whether or not the actual processing hardware is better or worse from one situation to the other are rarely, if ever, mentioned.

In terms of OA, my headcanon has generally been that early AIs required specialized hardware and software to operate and could not simply load themselves into any handy processor - same for uploads. However, as technology and know how advanced, this situation changed, either as a result of processing hardware getting better/changing or programming developing to the point where a sophont can be created purely in software without regard to the hardware - or both. So, by some point well before Y11k, the issue of whether or not you need specialized hardware to run a mind is basically moot (at least up to a point) because either the hardware is so common/easily produced that it is everywhere or because software and hardware based minds are both easily created and shifted between, or some combo of this.

Does this line up with y'alls conception or do you see it as something else?

c) Regarding artificial minds in general (regardless of hardware) - a key point that isn't often brought up is that, being artificial, such minds do NOT need to necessarily follow any kind of traditional/natural evolutionary process. In RL, any brain/mind must 'run what it brung' to evolve increased capability. You have to start where you are and get to where you end up. In artificial minds, this process can be short circuited in a variety of ways, both in terms of general development and immediate capability or tasks (at least in OA).

A first generation AI does not need to evolve into a second/third/fifth generation AI - rather researchers and engineers and programmers working in hundreds or thousands of different places can figure out new/better designs for software and hardware and simply implement them from nothing. This may result in a wholly new AI or (in some cases) improvements or changes in the first AI without any regard (or at least less regard) for 'getting from A to B'.

In addition (at least in OA), new capabilities may be added to an existing mind or new minds created that are designed to be very good at some particular thing without necessarily having to fit into the existing structure or stay there indefinitely. Whether via the exoself/exocortex, or psychoware or a combo, an OA sophont can rapidly modify their mind (near instantly in some cases) or add in capabilities that their 'natural' brain lacks, in a variety of ways. Artificial brains (or parts of brains) can also be designed to do some particular task stupendously well - and then be ignored (or even turned off) when not being used. They don't have to fit into a whole that is constantly 'turned on'. They can also be built on principles (such as laser signaling) that RL brains have never developed.

Much is sometimes made about birds being much more maneuverable flyers than jet planes. But no bird can cross the Atlantic in a matter of hours carrying hundreds of humans on its back. So it is a case of tradeoffs resulting in huge capabilities in the area the builders find useful.

d) Something else that doesn't generally get touched on in SF is that RL brains have all evolved on Earth and in Earth-like environments - most especially in a 1G gravity field and in a fairly small temp range and in a planetary atmosphere. An artificial brain might be designed from the ground up to operate at much higher or lower temps or in a free-fall environment or in vacuum (or some combo of these and other things of course) that would result in processors that far outstrip anything nature has produced in RL (whether in speed, efficiency, capability of some combo of these), but which would freeze/melt/collapse under their own weight or even the gentlest of breezes.

In the case of the aforementioned Matrioshka brain, different processors could be optimized for how far they are from the local star, could be made to only operate in free-fall and vacuum conditions, could draw energy straight from sunlight, and could 'outsource' a lot of the self-repair and support functions that a RL brain would normally need. Individual components and 'lobes' (assuming something analogous to a lobe) might be built and rebuilt on an ongoing basis as the M-mind changes its interests or focus or whatever tasks it is undertaking. Optical signaling in vacuum might allow a much more flexible brain that has 'white matter' of a sort that doesn't necessarily take up space in the way we normally think of such things. And as has been mentioned elsewhere in the thread, the mind of an M-brain might have little or no similarity to the mind of a human (indeed we generally take that as a given.

Just some thoughts,

Todd

[extherian]

I'm afraid I'll have to disagree with you on the importance of brain size, Drashner1. While Transapient brains may well be structured very differently to human brains, they also require an enormous amount of processing power to run. In the case of the Athenaeids, 1 cubic centimetre of their brain is equivalent to 40 modosophont intellects, so clearly running a Transapient mind is not a trivial task. A human brain could never run a mind of such complexity due to its low computing density, and would need to be thousands of times larger to meet the minimum processing requirements, in addition to any restructuring required.

That said, brain size does seem to become less important as one traverses the toposophic scale, with S4 and S5 minds requiring proportionally less of an increase in size compared to what S2 and S3 beings must undergo. The higher toposophics do seem to be more a matter of mind structure than raw power, apparently because they've discovered new computing substrates like Plasma Processors which have orders of magnitude more computing power per litre than the crude Ultimate Chips the lower toposophics must make do with.

It's worth noting that this isn't the case for humans and our primate ancestors. Our brains are structurally almost identical to chimp brains, just scaled up to have a vastly higher neuron count, which required our brains to increase in size. We do have specialised brain areas for language processing and other things, but the basic layout of our brains is hardly different to other primates. We're nothing special or new in evolutionary terms, just a bigger brained chimp with more raw power and a few nifty new skills like abstract language.

Regarding Highbrows, they are described as being an offshoot of Homo Superior (they're listed under the category of Superiors at any rate), and so it's reasonable to assume that their brains are made of the same neurons and synapses as other humans. Our current entry on Highbrows doesn't explain how their brains work differently to those of Superiors, and to me seems to suggest that their increased brain power alone brings them to 'near-Transapient' status. Of course, it's a very old article, so this is forgivable, but Superior-level neural enhancements wouldn't be enough by themselves to fully utilize such a massive brain.

[SeanR]

So the consciousness and subjective experience of time of a Matrioshka brain must be fractured into innumerable time-separated domains.... This is not comparable to anything we are familiar with, apart from maybe a large distributed human empire communicating by sailing ship and horse-rider.

The way I see it, my brain is "fractured into innumerable time-separated domains". So is yours. Everyone, who has learned to let their subconscious chew on a problem for awhile, in hopes that an elegant solution will surface, is.
Further, anytime you are holding your keys in your hand, and looking for your keys. or thinking you need to grab the grocery list WHILE you lock yourself out of your house, (and the grocery list) is experiencing the downside.

Edit. What I'm saying is you don't need full bandwidth between every region, and we don't have it. You need enough bandwidth to pass the "solution" to a presented "problem" either back to the calling region, or on to the next region in a chain.
Heck, the 8088 famously had a smaller bus connecting the chip to the motherboard than it used internally. The clock speeds of modern processors are in the microwave ranges, but the motherboards don't run in the GHz range, but in the high multiples of MHz range.

[Drashner1]

I'm afraid I'll have to disagree with you on the importance of brain size, Drashner1. While Transapient brains may well be structured very differently to human brains, they also require an enormous amount of processing power to run. In the case of the Athenaeids, 1 cubic centimetre of their brain is equivalent to 40 modosophont intellects, so clearly running a Transapient mind is not a trivial task. A human brain could never run a mind of such complexity due to its low computing density, and would need to be thousands of times larger to meet the minimum processing requirements, in addition to any restructuring required.

Er. Tbh I'm not entirely sure what part you're disagreeing with. Can you clarify on your specific objection to what I said please?

Processing power/performance is not the same as size however. A modern cell phone can vastly outperform a room filling computer from less than a century ago. And I didn't say that brain size is irrelevant - rather that it is overly simplistic to just assume that it is the only (or the main) factor to consider. As you say, a transapient brain is vastly more capable (more than 10,000x more capable) than a human mind but is not 10,000x + the size of a human brain. So the reduced importance of brain size that you mention in your next paragraph starts to kick in from the S1 level on up.

S1 don't actually use Ultimate Chips (unless gifted to them by S2 or higher), but their processing substrate is still much more capable than what modosophonts can manage. And it certainly runs better 'software'.

It's worth noting that this isn't the case for humans and our primate ancestors. Our brains are structurally almost identical to chimp brains, just scaled up to have a vastly higher neuron count, which required our brains to increase in size. We do have specialised brain areas for language processing and other things, but the basic layout of our brains is hardly different to other primates. We're nothing special or new in evolutionary terms, just a bigger brained chimp with more raw power and a few nifty new skills like abstract language.

Some might argue that those nifty skills (which may require either a bigger brain or a different brain structure or both to function) are both new and special

In a similar vein, the various transapients (and possibly the various superiors) may have 'nifty skills' that only emerge from their greater brain power. But they may have others that are a consequence of totally unique structures that have no natural ancestor or equivalent.

Regarding Highbrows, they are described as being an offshoot of Homo Superior (they're listed under the category of Superiors at any rate), and so it's reasonable to assume that their brains are made of the same neurons and synapses as other humans. Our current entry on Highbrows doesn't explain how their brains work differently to those of Superiors, and to me seems to suggest that their increased brain power alone brings them to 'near-Transapient' status. Of course, it's a very old article, so this is forgivable, but Superior-level neural enhancements wouldn't be enough by themselves to fully utilize such a massive brain.

'Superior' is something of a moving target. The Su of the Interplanetary Age are not the Su of the Current Era. Likely the Highbrows of the First Fed aren't the same as those of the Current Era either. Both new brain structure and a total redesign of both neurons and synapses to something very different from that of 'baseline humans' could easily have taken place (perhaps multiple times) in the thousands of years between when these species first arose and Y11k.

Also, as you say, the article is rather old. Highbrows may be more capable than the more 'average' Superiors while still falling far short of even the least transapient, as we currently conceive them.

Todd

[extherian]

Think we got our wires crossed somewhere...not enough 'white matter', perhaps. I thought you were saying that processing power was of only minor significance in the difference between Transapients and modosophonts. My point was that we are only slightly different from chimps, yet we have vastly greater mental abilities than they do. I like to think that sheer processing power plays a significant role in the unique capabilities that Transapients have, if for no other reason than that these abilities would require impractical amounts of computronium to integrate into a modosophont mind. Our brains didn't have to be redesigned from scratch to 'ascend', they just needed more mass.

Regarding Superiors, they may have been a moving target in the past, but there's only so much you can improve a fundamentally flawed processing substrate like the human brain. By Y11K, they've surely reached the limit of how intelligent a biological human can become while keeping the brain's size constant. I would imagine that the many mental advantages described in the Superior Genemods page are those of Y11K Superiors, and that they aren't getting any smarter since they've run out of things to improve.