« Do Not Taunt Massive Quake Ball | Main | Monday Topsight, June 30, 2008 »

Singularities Enough, and Time

brain-sil.pngA few people have asked me what I thought of Karl Schroeder's recent article at Worldchanging, "No Time for the Singularity." Karl argues that we can't count on super-intelligent AIs to save us from environmental disaster, since by the time they're possible (assuming that they're possible), things will have gotten so bad that they won't matter (and/or won't have any resources available to act, or even persist). It's a pretty straightforward argument, and echoes pieces I've written on parallel themes. In short, my initial reaction, was "yeah, of course."

But giving it a bit more thought, I see that Karl's argument has a couple of subtle, but important, flaws.

The first is that he makes the assumption that nearly every casual discussion of the Singularity concept makes, in that he defines it as "...within about 25 years, computers will exceed human intelligence and rapidly bootstrap themselves to godlike status." But if you go back to Vinge's original piece, you'll see that he actually suggests four different pathways to a Singularity, only two of which arguably include super-intelligent AI. His four pathways are:

• There may be developed computers that are "awake" and superhumanly intelligent. (To date, there has been much controversy as to whether we can create human equivalence in a machine. But if the answer is "yes, we can", then there is little doubt that beings more intelligent can be constructed shortly thereafter.)
• Large computer networks (and their associated users) may "wake up" as a superhumanly intelligent entity.
• Computer/human interfaces may become so intimate that users may reasonably be considered superhumanly intelligent.
• Biological science may provide means to improve natural human intellect.

The first two depend upon computers gaining self-awareness and boostrapping themselves into super-intelligence through some handwaved process. People don't talk much about the Internet "waking up" these days, but talk of artificially intelligent systems remains quite popular. And while the details of how we might get from here to a seemingly intelligent machine grow more sophisticated, there's still quite a bit of handwaving about how that bootstrapping to super-intelligence would actually take place.

The second two -- computer/human interfaces and biological enhancement -- fall into the category of "intelligence augmentation," or IA. Here, the notion is that the human brain remains the smartest thing around, but has either cybernetic or biotechnological turbo chargers. It's important to note that the cyber version of this concept does not require that the embedded/connected computer be anything other than a fancy dumb system -- you wouldn't necessarily have to put up with an AI in your head.

So when Karl says that the Singularity, if it's even possible, wouldn't arrive in nearly enough time to deal with global environmental disasters, he's really only talking about one kind of Singularity. It's this narrowing of terms that leads to the second flaw in his argument.

Karl seems to suggest that only super-intelligent AIs would be able to figure out what to do about an eco-pocalypse. But there's still quite a bit of advancement to be had between the present level of intelligence-related technologies, and Singularity-scale technologies -- and that pathway of advancement will almost certainly be of tremendous value to figuring out how to avoid disaster.

This pathway is especially clear when it comes to the two non-AI versions of the Singularity concept. With bio-enhancement, it's easy to find stories about how Ritalin or Adderall or Provigil have become productivity tools in school and in the workplace. To the degree that our sense of "intelligence" depends on a capacity to learn and process new information, these drugs are simple intelligence boosters (ones with potential risks, as the linked articles suggest). While they're simple, they're also indicative of where things are going: our increasing understanding of how the brain functions will very likely lead to more powerful cognitive modifications.

The intelligence-boosting through human-computer connections is even easier to see -- just look in front of you. We're already offloading certain cognitive functions to our computing systems, functions such as memory, math, and increasingly, information analysis. Powerful simulations and petabyte-scale datasets allow us to do things with our brains that would once have been literally unimaginable. That the interface between our brains and our computers requires typing and/or pointing, rather than just thinking, is arguably a benefit rather than a drawback: upgrading is much simpler when there's no surgery involved.

You don't have to believe in godlike super-AIs to see that this kind of intelligence enhancement can lead to some pretty significant results as the systems get more complex, datasets get bigger, connections get faster, and interfaces become ever more useable.

So we have intelligence augmentation through both biochemistry and human-computer interface well underway and increasingly powerful, with artificial intelligence on some possible horizon. Let's cast aside the loaded term "Singularity" and just talk about getting smarter. This is happening now, and will under nearly any plausible scenario keep happening for at least the next decade and a half. Enhanced intelligence alone won't solve global warming and other environmental threats, but it will almost certainly make the solutions we come up with more effective. We could deal with these crises without getting any smarter, to be sure, and we shouldn't depend on getting smarter later as a way of avoiding hard work today. But we should certainly take advantage of whatever new capacities or advantages may emerge.

I still say that the Singularity is not a sustainability strategy, and agree with Karl that it's ludicrous to consider future advances in technology as our only hope. But we should at the same time be ready to embrace such advances if they do, in fact, emerge. The situation we face, particularly with regards to climate disruption, is so potentially devastating that we have to be willing to accept new strategies based on new conditions and opportunities. In the end, the best tool we have for dealing with potential catastrophe is our ability to innovate.

Comments

"In the end, the best tool we have for dealing with potential catastrophe is our ability to innovate."

Dealing with the coming catastrophe(s) has less to do with innovation than with our ability to organize ourselves for change. It has taken 30 years for CFLs to garner a 6% market share of the light bulb business in the US.

I was in NYC a few days ago to see John Todd receive the Buckminster Fuller Challenge award. He showed slides of one of his first Living Machines, cleaning sewage in Harwich MA. That was in 1986. How many municipalities are using that proven technology today? 1/100th of 1%? If you look back, you could argue that New Alchemy Institute had documented, with numbers, most of the solutions to our food, heating, and home power problems but how many people today even remember what they did?

We have had technologies,techniques, and innovations to ameliorate the possible coming catastrophe(s) for years if not decades now. We haven't had the organization and political will to put them into practice.

New Orleans anyone?

Jamais, between your post and gmoke's response, there's the start of a very interesting discussion. I'm old enough to have been around when The Limits To Growth was published, and to have known its authors, John Todd, Amory Lovins, Kenneth Boulding, and a whole bunch of other folks who knew the path toward sustainability 35 years ago. Most of them would be thrilled by the new tools and enhancements you cite, but would caution us against thinking that they automatically lead to or cause anything. Sure, new tools give us new possibilities, often unforeseen. But for any tool, what we do with it depends on how we wield it. To wield our tools well requires training, foresight, patience, intelligence, and ethics.

Despite dazzling technical advances, the future remains something we choose to create, rather than an inevitable trend. We have choices. Those choices are constrained by ultimate laws of physics, widened by technical innovations. But the future is still a creative act, made with the tools at hand. We could have started creating a sustainable world 30 years ago, and in many ways, it would have been easier than starting now. Since we're starting far later than prudent, I'm glad we have powerful tools - we'll need them. But - Singularity? We don't need no stinkin' Singularity.

1. If there was godlike AI then geoengineering a restored earth would be easy

2. We already know several technological solutions to climate change and to peak oil.

Mass produce nuclear power.
Uranium hydride reactors, refinements of existing reactors, molten salt reactors or all of the above.

Use MIT annular fuel approach for uprating existing reactors by 50%.

Start switching cars and vehicles over to all electric. Use ultracapacitor/battery combos and lighten cars with inflatable cars (XP vehicles)

The technology that we will have in hand over the next 8-12 years will be plenty.

Kitegen, coolearth, sunrgi for the wind and solar paths.

http://peswiki.com/index.php/Directory:Concentrated_Solar_Power

I think your analysis is quite on the mark, Jamais. I'm known for being a critic of the singularity idea, but I'll admit there's a bit of a straw man argument to this particular piece. In fact, I'm thinking of writing a near-future novel (a kind of answer to John Brunner's Stand on Zanzibar) in which various mutations of today's social networking and web 2.0 software enable massively distributed problem solving by people. Basically, collective intelligence as the enabling factor for speeding up actual implementation of ideas--because, as David points out, the ideas are already there. I don't personally identify collective intelligence apps as "singularity tech" but they are in line with the options you're talking about here.

My intent with the piece was, of course, just to snap the technophiles in the audience out of any complacency they might be feeling. It seems to have done that.

Superintelligence would allow us not to just solve the climate change problem, but dozens of other dire problems, like the development of WMDs, genocide, malaria, AIDS, etc. I choose to live sustainably and encourage work towards superintelligence. I don't think Karl's language ("mythology", etc.) gives the proposal the seriousness it deserves.

And as Brian says, I think the solutions to solving the climate change problem are available now, but environmentalists refuse to embrace them. This mainly means nuclear power.

As one of the co-founders of WorldChanging and an environmentalist blogger, you can help by promoting nuclear power.

Michael, I'll address the "mythology" aspect soon (I'm in a workshop today).

As for nuclear... I honestly don't understand the obsession with nuclear power among many Singularity/transhumanism enthusiasts, especially those focusing on safe and responsible change. It's precisely the kind of centralized, expensive, slow-to-change, hard-to-reverse, risk-encouraging technology that should be anathema.

Nuclear power isn't the only solution out there, and it's hardly the best.

I certainly support continued research into safer/simpler/cheaper forms of nuclear. I've already written about thorium nuclear, especially molten-salt thorium (no proliferation risk, thorium more abundant than uranium, much shorter half-life). I have no doubt that nuclear will be part of the mix, albeit not anytime soon -- the construction time for current, available forms of nuclear power is measured in decades.

In terms of economies of scope, where innovations in one arena support benefits in others, distributed, cheaper, fast-changing, easily-reversible, safer technologies are far more attractive. As is efficiency -- bear in mind that transportation is less than a third of the overall source of greenhouse gases. Moving to electric vehicles (itself not a fast process, because it's not just cars, it's the support infrastructure) isn't enough in and of itself -- updating building design (homes and business) actually has a bigger impact.

It takes between 10 and 14 years for the fleet of cars now on the road to be replaced by newer vehicles. At least, according to what I've read.

The lag time built into our energy infrastructure makes responding to climate change extremely difficult, even without the deniers and foot-draggers and competing interests. If we had a clear path that we all agreed on, we'd still have to go far, far beyond business as usual practices in order to make the changes happen in time to avoid the worst effects of climate change.

In addition, any prospective changes we can make now will not stop the climate change that is already in the pipeline (pun intended).

Such views make me the cheerful and positive person that I am today.

However, I do like the approach that John Todd won the first Buckminster Fuller Challenge Award with:

http://www.dailykos.com/story/2008/7/1/165138/5528

China is in discussions with Westinghouse to increase orders of the AP1000 up to 100 built or under construction by 2020. AP1000's being built in China are of the 1250MW type. Next batch are the 1400MW design and following that will be 1700MW.

China starts building first high temp reactor 200MW version in 2009. Successful develop will see factory mass production of this reactor. Two year construction times. Smaller total budget needed. A project that built eight of these reactors would build each reactor and turn it on to generate utility money. Never getting in debt for more than one 200MW reactor. Would be compatible with providing high temp industrial heat for some applications.

I have never understood the obsession that some futurist environmental types have with not looking in depth at nuclear power, what projects are actually being built, the database of actual construction times, the nuclear fuel cycle. Obviously Jamais has done better than most, but still needs to consider how much is possible with nuclear fission. Karl Schroeder has not tried at all.

I think Karl is being too complacent is his analysis of the singularity and nuclear technological solutions and other technological solutions. Karl talks about complacency but does not back up his own position with in depth analysis of projects in development, patents and research papers of the technology he criticizes.

On my site:
laser refinement of nuclear fuel will provide three or more times lower operating costs for nuclear fuel. (GE, 2012 target commercialization)

MIT 50% uprating of nuclear reactors (Westinghouse working on commercializing)

Higher efficiency reactor research [not just molten salt, high temp reactors (china), accelerator driven reactors -EU, national lab research]

Progress on restoring and expanding the nuclear supply chain. Workarounds on the temporary limit on large forgings. Make two halves of the 600 ton forgings and weld them together, French redesign of plants to only need more common 300 ton forging limit, Candu reactor design, S Korea/China/Russia and others bringing large forging capacity online.

for many of the other environmental and energy solutions, we cannot point to a country and say see they built up a lot of solar or wind and are running their country on it. Nuclear can say see France did it in the eighties to get to 80% nuclear power for electricity.

Nuclear matters to reduce the air pollution from coal and fossil fuel power. Air pollution (indoor and outdoor) cause 5 million deaths per year worldwide.

The US is politically dicking around and slowly moving to nuclear power,but even the US passed the 2005 energy bill and allows southern states to recharge customers for new power supply. So thirty new reactors will be built even without new legislative change. A 2009 climate change bill should penalize coal and further accelerate the next economic and large scale option for the big utilities to choose which is nuclear.

Any reactors that get built over the next few decades that leave unburned fuel can have that fuel burned with follow on reactors with higher efficiency. Like China's high temp reactors or uranium hydride reactors or molten salt reactors.

Inflatable electric cars are targeting less than $10,000 prices with one model at $2500 price to compete with Tata nano. 300 mile range with regular batteries and 3000 mile range with a special swappable fuel cell power pack. $2500 is the price of a upscale refridgerator or a big screen TV. People can keep their old cars and buy an inflatable one as an extra regular commute car.

Don't be lazy and stick to outdated positions. Look at and support the best solutions that are or are becoming available. Look at the actual shifts to better solutions anywhere in the world. Global problems do not have to wait for the USA to implement a solution.

If Karl or Jamais or someone else reads and addresses all of my key articles on nuclear power and points out where I am seriously off course in supporting it then I may reconsider.

Likewise Karl can look at and address my mundane Singularity and manufacturing revolution articles to dispute my belief that the benefits attributed to a Singularity are being developed and implemented.

Archives

Creative Commons License
This weblog is licensed under a Creative Commons License.
Powered By MovableType 4.37