“The Singularity” and why the world will change more than (almost?) anyone can imagine in the next century or two

This post is apropos of nothing, aside from being something I wish were more widely understood. It has to do with AI and uploading and “the Singularity” (a confusing term that can actually mean several different things). I’m going to focus on uploading because in some ways it’s the most straightforward thing.

If you really want to understand the uploading issue, I recommend Anders Sandberg and Nick Bostrom’s “Whole Brain Emulation: A Roadmap.” The thing is over a hundred pages, long, though, so here’s a summary. The version of the uploading idea: take a preserved dead brain, slice it into very thin slices, scan the slices, and build a computer simulation of the entire brain.

If this process manages to give you a sufficiently accurate simulation, the implications are huge. For starters, given the right interface (which would have to include a lower-resolution body and environment simulation), you could have a conversation with the simulation that would, from an outside point of view at least, be indistinguishable from a conversation with the simulated person. Perhaps more importantly, such simulations could potentially replace humans in a huge range of fields.

For uploading to be successful in the “can have conversations and replace human workers” sense, surprisingly few assumptions are required. As Sandberg and Bostrom say, “Physicalism (everything supervenes on the physical) is a convenient but not necessary assumption.” Epiphenomenal dualism would actually be just as good an assumption, and uploading could even work on interactionist dualism, if the non-physical part of the mind is also possible to simulate.

Furthermore, the question of whether uploads could replace human workers is independent of philosophical questions like “would an upload really be conscious?” or “would a simulation of a dead person be a form of survival for that dead person?” (Sandberg and Bostrom call these three questions the 6a, 6b, and 6c success criteria, respectively.) And a “yes” to the first question is all it takes for uploads to radically change the world.

Why? Because digital minds would have some immediate advantages over the made of meat minds we have right now. Luke Muehlhauser and Anna Salamon’s excellent paper “Intelligence Exploision: Evidence and Import” explains a number of these advantages in the section on “AI advantages.” Here are two that are especially relevant to uploads:

Communication speed. Axons carry spike signals at 75 meters per second or less (Kandel et al. 2000). That speed is a fixed consequence of our physiology. In contrast, software minds could be ported to faster hardware, and could therefore process information more rapidly. (Of course, this also depends on the efficiency of the algorithms in use; faster hardware compensates for less efficient software.)

Duplicability. Our research colleague Steve Rayhawk likes to describe AI as “instant intelligence; just add hardware!” What Rayhawk means is that, while it will require extensive research to design the first AI, creating additional AIs is just a matter of copying software. The population of digital minds can thus expand to fill the available hardware base, perhaps rapidly surpassing the population of biological minds.

Duplicability also allows the AI population to rapidly become dominated by newly built AIs, with new skills. Since an AI’s skills are stored digitally, its exact current state can be copied, including memories and acquired skills—similar to how a “system state” can be copied by hardware emulation programs or system backup programs. A human who undergoes education increases only his or her own performance, but an AI that becomes 10% better at earning money (per dollar of rentable hardware) than other AIs can be used to replace the others across the hardware base—making each copy 10% more efficient.

I think it’s safe to say that being able to make copies of our most talented and skilled folk (from R&D and marketing departments to scientists and intellectuals), life on Earth will be changed forever. Oh, theoretically we could decide not to use the technology, but I suspect we’ll be unable to resist the temptation of such potentially huge benefits. (For more on this, see the blog and papers of economist Robin Hanson.)

How long until this happens? Well, as Yogi Berra once said, making predictions is hard, especially about the future. But on the one hand, I’d be surprised if getting working human whole brain emulations only took a couple of decades, as some have predicted. Neuroscience is messy, messier than many people think.

On the other hand, the fact that we’re at the point where we can say, in some detail, what further technological advances are necessary to make whole brain emulation feasible (as Sandberg and Bostrom do in their paper) makes me think the technology isn’t too far off. I wouldn’t be surprised if we had it in another century. I would be surprised if it took too long. I can’t say anything for sure, but a couple centuries seems like a safe bet for when we’ll have uploading.

Of course, it’s at least theoretically possible that some kind of catastrophe could greatly damage scientific and technological progress in all areas, including towards uploading. But such a catastrophe would, in itself, but a huge change from “life on Earth as we know it.” So, while I wouldn’t bet too much on specific scenarios (specific scenarios always have a lower probability than more general predictions), it’s a safe bet that some kind of huge changes are in humanity’s future.

How big is huge? I think we can get an idea by comparing likely futures to the visions of science fiction. In most popular science fiction (like Star Trek), humanity has gotten all kinds of cool new toys several centuries from now, but the vast majority of the characters aren’t that much different from folk today. Data and Voyager’s The Doctor are in the minority.

And that kind of Star Trek model of how technological progress works is largely faithful to what’s happened in the past. In reality, though, it seems likely that technological advance will ultimately mean changes in what it is to be a person–or, perhaps I should say, a “person-like entity,” since it will be controversial whether many of the things that are likely to exist in the future will even be people.

Update: I wrote a follow-up post with clarifications to some of the above.

  • Mikey

    Growing up in movies like Short Circuit and Herby the Love Bug, I’m more then ready to great advance AI as equals. Ever hears of a webcomic called Questionable Content? The world is pretty much like ours but theres walking, talking robots and AI and the creator has a very interesting spin on how they achieved equality.

  • Joe

    There is a huge difference between simulating a dead brain, and simulating a functioning living brain. When a brain is starved of oxygen, you have about 4 minutes before irreparable brain damage occurs. Cutting up a dead brain may give you a rudimentary understanding of brain structure, but mostly it just makes a big mess.

    ‘Uploading’ is a scifi geek fantasy we are no where close to achieving. And the ‘singularity’ in this sense has a simple definition: geek rapture.

    • tac

      yeah—but: the things that change after death are the enzymatic processes that run he cells, not the physical connections. Even the types of receptors (proteins) and neurotransmittirs are still present long after death, assuming there has been some form of preservation/fixative/embalming. This is the basis of all the neuro anatomy studies that have been done for the last 100 or more years.

      • Joe

        I was going to respond…. but PZ said it better than I could:

        “If singularitarians were 19th century engineers, they’d be the ones talking about our glorious future of transportation by proposing to hack up horses and replace their muscles with hydraulics. Yes, that’s the future: steam-powered robot horses. And if we shovel more coal into their bellies, they’ll go faster!”

  • anatman

    uploading is an interesting idea but … marketers?

  • Pingback: And everyone gets a robot pony! | Pharyngula

  • InfraredEyes

    I don’t see how taking a snapshot of the state of the brain will give you the ability to “upload” a mind. Seems to me that what we call “mind” is likely to be an emergent property of the functioning brain. In other words, it arises from the dynamics of the brain and won’t be captured by a static image, however precise.

  • Pingback: The singularity gets roasted | dailydouq

  • Jon Hanson

    PZ finally notices you, and this happens.

    Dang.

  • Pingback: And everyone gets a robot pony! [Pharyngula] - BizNax

  • Hunt

    I like to think of Singularity, Intelligence Explosion, Nanotech, etc. all as interesting ideas that are fun to think about and that may actually come to fruition in the future. Nanotech of course, is already bearing fruit. I’ve always found responses like PZ’s disappointing, especially since his critique fails to hit the mark for a number of reasons. All the topics above are potentially science-based, but also some of them come with a certain measure of conjecture, which AFAIC is okay while acknowledge, as it generally is. They don’t deserve to be slammed for being “woo” or new age BS. Reading PZ on this distinctly reminds me of reading an ICR tract presenting “scientific proof” against evolution. I’ve never put him on a pedestal, like some people around here do, and this just chops his legs off a little more in my eyes.

  • Pingback: Alexander Kruel · PZ Myers aptly sums it up

  • ragarth

    Let me start off saying that, yes, I’m a transhumanist. I am not, however, a singularitarian (or at least not of the kurzweilian persuation.)

    I do believe the concept of uploading is feasible, but the majority of its defenders have utterly not idea what such a technology would entail. Simply slicing up a dead brain and scanning it isn’t feasible. With brain death comes the degradation of dendrites and critical components in the synaptic cleft. It would take little time at all for the degredation to be sufficient that any mind duplicated from the dead brain would be unfeasible, even with the fault tolerance of neural networks accounted for (this is also, by the way, why I’m not an advocate of cryogenics–ice crystals). Given this, it is my opinion that any uploading technology would have to function on a living brain, and this opens up a whole new can of technical hurtles to overcome.

    One idea I toyed with was distributing small nanoscale transmitters through the brain to provide information about a large number of neurons, then use math to reverse-model the axon lengths, synaptic weights, and connection of the brain. This is, also, a technology far in advance of anything we can accomplish. The living brain would likely be destroyed (many nano-structures have proven to be toxic to living cells, they get into the lipid layer of the cell membrane and wreak havoc), there’s only so much bandwidth we can broadcast on, power for the nanosystems, and gathering enough data from a collection of a few thousand molecules to do a functional reconstruction are just the first hurdles.

    Its unlikely that anyone living today will see technology such as uploading happen. The above are just a few hurdles, there are many dozen more that would each require a whole host of new technology to overcome, and these are just the bed-monsters we know about. There are doubtless more challenges to the development of these singularitarian technologies that make futurist predictions laughable.

    Uploading is just like fusion power, AI, and flying cars: Always predicted to be 20-100 years away by those who neither understand what they’re talking about, nor understand that there are more monsters under the beds than we know of.

  • Pingback: Clarifications on uploading | The Uncredible Hallq

  • meh1963

    Kind of interesting. I’ve read lots of SF which includes uploading, but it’s always accomplished by handwavium.

    As someone who codes & manages systems for a living, I respectfully suggest we’re a long, long way from even rudimentary uploads. For all practical purposes, it’s a state transfer, and we can barely even do that to generate a VM on a PC over well-established communications interfaces with well-established error-checking, let alone something as complex as a brain.

    Though it’s still a really, really cool idea and a great trope in SF….

  • http://homeschoolingphysicist.blogspot.com PhysicistDave

    Chris,

    I have to agree with ragarth and meh1963. When I was a kid, fifty years ago, controlled fusion was just around the corner: the other day, a friend whose brother is working on it told me they are optimistic. Glad to hear it.

    I was a friend of a friend of Erik Drexler’s when his nanotech Bible, Engines of Creation came out twenty-six years ago: the Drexlerites told me that the creation of the master nanotech assembler was just around the corner. We’re still waiting.

    Similarly for the cure for cancer, the permanent lunar base, etc.

    I think I know why: for a number of years, I did integrated circuit design and worked as a semiconductor device physicist. What limited us was not lack of theoretical understanding but very dirty, mundane, practical problems that lacked clean theoretical solutions: How clean can you get the air in the “clean rooms” used for manufacturing? How uniform could we make the glass lenses used in photolithography? Etc. Even if you had worked out dozens and dozens of steps in the process just right, if one single thing was a sticking point, that limited you.

    Whatever you want to call it: the weakest link principle, the 90/10 rule, etc.

    ragarth and meh1963 have already discussed some of the key problems with uploading; I’ll add another: We are reaching the physical limits of semiconductor technology due to atomic size and electron discreteness. I’m not saying that computers will not be ten times as powerful a decade from now, but they will not be a thousand times as powerful, as Moore’s law used to predict. And, more importantly, four decades from now, computers based on semiconductor technology will not be a trillion times as powerful.

    Alas, that kind of decade-after-decade Moore’s-law growth is what we need for uploading to work, given the complexity of the brain.

    I am far too prudent to predict what the world will be like thousands of years from now.

    But, unless we make some radically different computing technology work (quantum computers? – as a physicist, I think “decoherence” kills you on any large-scale quantum computer), uploading is dead for centuries to come.

    I will stick my neck out on one prediction: Cartesian dualism is going to make a comeback based on work in neural science during this century. But do remember: I am a physicist, not a neuroscientist!

    Dave Miller in Sacramento

  • KG

    The version of the uploading idea: take a preserved dead brain, slice it into very thin slices, scan the slices, and build a computer simulation of the entire brain.

    If this process manages to give you a sufficiently accurate simulation – Chris Hallquist

    It won’t. Which makes the rest of your post completely otiose.

  • http://www.sfgate.com/profile/?User=abb7c0h0&plckPersonaPage=BlogViewPost&plckUserId=abb7c0h0&plckPostId=Blog%3aabb7c0h0Post%3a7b2fdf04-6dc1-4660-80cb-5e8100805a0a&plckController=Perso German Dec

    I’ve not too long ago started an internet site, the info you provide on this site has helped me greatly. Thank you for your whole time & work. “Cultivation to the thoughts is as obligatory as food to the body.” by Marcus Tullius Cicero.

  • Pingback: Comprehensive Copying Not Required for Uploading | Accelerating Future


CLOSE | X

HIDE | X