Monday Wilson Quarterly Blogging
Peter W. Singer, author of a 2003 book on private military contractors and a 2005 book on child soldiers, is back with a new book Wired for War about military robots. I won’t say much more about the book at this juncture since I’m supposed to review it for a magazine, but I will recommend this adaptation from the book in the new issue of The Wilson Quarterly.
This is a magazine that I think nobody reads, but it’s actually really good. Recent issues have included this from Larry Bartels and this from Holly Yeager and other great stuff. But you can’t just do everything online! If you buy the current issue—the one with Singer’s robots on the cover—on the newsstand you’ll find one witty, engaging, and informative letter to the editor by yours truly and another by the lovely and talented Sara Mead. As best I know, it’s a coincidence that we were both solicited to write letters for this issue—the previous one just happened to include one article that criticizes blogging and another that criticizes preschool—but maybe it’s all part of Skynet’s long-term plan to conquer us.
32 Responses to “Monday Wilson Quarterly Blogging”
Register
RSSLogin
Topic Cloud
ArchivesBlog RollAbout Matt YglesiasContact Me
Donate
January 19th, 2009 at 2:23 pm
I was at a presentation in early 2004 that segued from a cutesy extended demonstration of the Roomba negotiating its way around corners to a bang-flash-pow Rambo-meets-Short Circuit promo video for the PackBot. It was, let’s say, a bit of a headfuck.
It’s arguable, though, that there’s only a limited distinction between an armed UAV and something that looks like Johnny Five, and then say that remote, video-mediated war has been going on for a while.
January 19th, 2009 at 2:24 pm
Argh. Why must you trivialize this issue with Terminator jokes so consistently?
In the short-to-medium term, robotic weaponry threatens to lower the human costs of war for rich countries, and thus incentivize more conflict. Right now robots are being used to kill people in Iraq.
In the long run, there is a serious danger of the creation of a new species and human extinction. I thought you backed Parfit on the importance of existential risks?
http://matthewyglesias.theatlantic.com/archives/2008/06/request_parfit_and_the_electio.php
January 19th, 2009 at 2:36 pm
BATTLESTAR SPOILER ALERT!!!!!
Matt, Skynet doesn’t matter. We live on Earth and are therefore all Cylons. We’re already robots.
Face it. Humanity has lost.
January 19th, 2009 at 2:39 pm
I think I agree with Anon above. With AI and robotic technologies advancing rapidly (and toward each other), the time is fast approaching to not just be laughy-jokey about The Great Robot Menace and begin to seriously talk about the issue. We should really be making an effort NOW to disengage opposition to robotic warfare from Sci-Fi geekery. This is a serious issue and one that will eventually run headlong into the “Why do you want American boys to die?” question. If our our only response is: “Didn’t you see Terminator?” Then the robot lobby will win in a walk. And as we all know that would be BAD FOR HUMANS.
January 19th, 2009 at 2:46 pm
wow, you’re coming on 2 years with sara, huh?
January 19th, 2009 at 2:58 pm
I’ve got to throw my hat in w/anonymous with regard to incentivizing warfare for countries who can afford to send robots in lieu of human soldiers. I find that truly chilling.
Would the war in Iraq still be considered a tragic failure if we’d lost 4,000 robots instead of living, breathing soldiers? It’d certainly be just as tragic for Iraqis, but I don’t think most Americans would view such a misadventure with much more than a shrug.
January 19th, 2009 at 3:03 pm
After watching the extended version of Terminator 2 on my brand-new 50″ plasma TV, I’ve come to the conclusion that Matt, resistance is futile. Give up now. The robots will crush your head.
January 19th, 2009 at 4:10 pm
I’ve got to throw my hat in w/anonymous with regard to incentivizing warfare for countries who can afford to send robots in lieu of human soldiers. I find that truly chilling.
This is the same sort of reasoning that believes we should bring the draft back.
I don’t believe incentivizing doesn’t work that way. And besides, I’m sure some of the military wives wish a robot went instead of their husband.
Do feel the same way about Afghanistan as Iraq? About Bosnia? (are you an anti-war Peacenik?)
Maybe a robot UN peacekeeping army would be good. I’d draw the line at Robocops.
January 19th, 2009 at 4:37 pm
Boy those sentient robot jokes just keeping getting funnier and funnier every time you make them. Also, please, more quips about how you’re the ninth most powerful person in D.C.!
Seriously, dude, do you really want to end up like Atrios, making the same lame joke over and over? You’re better than that. BTW, while I’m razzing you, I will say your MLK post on nonviolence below was excellent.
January 19th, 2009 at 5:12 pm
WQ is very good, though it tends to be a little dry.
January 19th, 2009 at 5:17 pm
Ahhh…those of you who are worried about the robots just need to buy some robot insurance.
I sleep better at night knowing that I am safe from robots eating my pills.
Besides, if the robots try to take over, I’m sure someone will build a 100 foot tall battlemech to kick the robots asses. I’m guessing the battlemech will have heat-seeking missiles, laser beams, and an energy sword for cutting down mofo’ing robots.
Of course, once the battlemech destroys skynet, we’re all screwed.
January 19th, 2009 at 6:00 pm
Peter K.,
Yes, there are arguments both ways on near-term robotic weapons, e.g. robots have a continually recorded video feed of their actions, avoiding situations where individual soldiers commit war crimes, or peacekeeping missions could be much less bloody (robots would not need to use excessive lethal force to avoid losses).
My point is only that these are important debates, debates that are impeded when we use a joking tone or associate them with fantasy. Imagine citing South Park’s satirical take on global warming in every post on the issue. It may seem like a minor peccadillo in the individual case, but over time and across individuals, it would worsen the chances of helpful action being taken.
In the long run, ’science-fiction-like’ issues will matter enormously this century, just as erstwhile scifi city-destroying ’sun-bombs’ and rockets carrying weapons from one continent to another dominated much of 20th century history. With respect to AI/robotics, I offer this piece from ‘Global Catastrophic Risks,’ a recent edited volume from Oxford University Press.
January 19th, 2009 at 6:34 pm
Anon wrote:
In the short-to-medium term, robotic weaponry threatens to lower the human costs of war for rich countries, and thus incentivize more conflict. Right now robots are being used to kill people in Iraq.
This seems plausible, but it’s hard to say what the effects of robot weaponry would really be. People don’t oppose the Iraq war only because it results in American soldiers getting killed, after all, they also oppose it because of what it’s doing to the Iraqis, how it’s creating so much bad feeling against America, how much money it’s costing us, etc. And with robot soldiers there might be some incentive to take more precautions about avoiding civilian casualties, since there wouldn’t be quite the same “kill or be killed” mentality–with human soldiers people make the argument than in uncertain situations it may pay to shoot first and ask questions later to avoid letting too many soldiers get killed, like if a car isn’t responding to requests to stop as it approaches a checkpoint, but with robot soldiers the public might be more horrified by the robots killing civilians just to avoid a small chance that some hardware will get damaged.
It’s also hard to imagine what would happen to war if robot soldiers became cheap enough that virtually any nation could buy them–would there be “wars” that were just clashes of robot vs. robot, essentially each country just trying to drain money from the other? I suppose that’s too idealistic, since there still might be fighting in populated areas which would result in civilian casualties.
In the long run, there is a serious danger of the creation of a new species and human extinction. I thought you backed Parfit on the importance of existential risks?
I don’t really buy the sci-fi idea that a robot system designed for some narrow task (like a robot soldier) could “spontaneously” develop general intelligence and turn on us. If a new intelligent species is created, it’ll be because that’s exactly what we were trying to create. And personally I think the most likely route to A.I. is mind uploading, where we try to build detailed simulations of an actual human brain that has been mapped at the synaptic level–in this case the new species would have completely human minds, and personally I don’t care if the inheritors of the Earth in a few hundred years are biological humans or nonbiological beings who think and feel just like we do, the one thing that would scare me would be a true Skynet scenario where the beings that come to be in charge are sociopaths with no empathy or other feelings.
January 19th, 2009 at 6:43 pm
The opinion people have about robots used in war, will likely be quite different from the opinions they have of robots being used in the police force.
January 19th, 2009 at 9:27 pm
“I don’t really buy the sci-fi idea that a robot system designed for some narrow task (like a robot soldier) could “spontaneously” develop general intelligence and turn on us.”
We will develop the general intelligence, but we might make not understand the full implications of the goal systems we build.
January 19th, 2009 at 9:28 pm
“make” is obviously a typo.
January 20th, 2009 at 1:06 am
“And personally I think the most likely route to A.I. is mind uploading, where we try to build detailed simulations of an actual human brain that has been mapped at the synaptic level–in this case the new species would have completely human minds”
I think if you can map the human brain at the cellular level – with nanotech devices, that should be feasible – you don’t have to “upload” an actual human mind. You’d have enough knowledge to build an AI without simulation. Simulation would be just “testing”.
Simulating a human brain is not “uploading”. Uploading entails actually building a human brain in software (and/or hardware) and then transferring a specific human mind into the receiving device.
The effect is like cloning in that you now have a completely different mind running in a different location in space/time, meaning the two are no longer identical. Which is why it is NOT a step toward immortality for the original mind. Only if you directly alter the original brain to BECOME a machine brain would be actually achieving immortality for the original brain.
Not that uploading doesn’t have value – it’s just not a course for personal immortality.
“, and personally I don’t care if the inheritors of the Earth in a few hundred years are biological humans or nonbiological beings who think and feel just like we do”
The last thing you want is for a superhuman intelligence to “think and feel just like we do”. That’s the whole point of a superhuman intelligence – to NOT do that, to transcend the limitations of a primate brain and neurochemistry.
“the one thing that would scare me would be a true Skynet scenario where the beings that come to be in charge are sociopaths with no empathy or other feelings.”
But this is precisely why the Terminator show is so valuable. The character of Cameron is described as an advanced Terminator that does NOT feel emotions but can emulate them almost perfectly. The producers don’t want to humanize the character – that’s been done. They want to keep it ambiguous, but also explore what it means to be a cyborg, i.e., a machine mind with no emotions who has to deal with emotional humans.
Empathy is not the problem. The problem is rationality. The story arc in TSCC is currently having a conservative Christian FBI agent teaching the new Skynet prototype AI about religion. All the fans have said, “Bad idea!”
On the other hand, you have what appears to be a self-aware, self-initiating, independent AI with a sense of self-preservation in Cameron, who is learning to deal with humans while still being a completely logical and unemotional machine.
If you have a machine mind which is completely logical, it becomes a matter of actually teaching it something which IS a rational, logical philosophy and attitude which will lead it to be non-coercive.
And this is where most humans break down – because THEY don’t have a completely rational, logical philosophy and attitude that leads directly to a non-coercive concept. They have religion which is based on nothing, they have morals which are not firmly grounded in reality and logic and exist primarily to enable one primate to declare itself better than the next, and they have laws which by definition are coercive instruments.
Thus, you get the threat of an irrational machine mind.
Which is why I as a radical Transhumanist oppose the notion of creating an external AI with sufficient conceptual processing to become a threat. Rather, I propose developing the means for a superhuman intelligence, then augmenting or transmogrifying the human brain into that superhuman intelligence. This avoids the whole humans vs AI problem completely.
It does not, however, avoid the whole humans vs Transhumans problem. The end result, as I’ve pointed out before, is that there are only three possible outcomes:
1) Humans attack Transhumans and get exterminated.
2) Transhumans transmogrify humans into Transhumans, Nobody complains once it’s done.
3) Transhumans ignore humans and go elsewhere.
In reality, the most likely outcome is 4) all of the above – some humans get exterminated, some get transformed, some get ignored.
My best estimate is that a superhuman intelligence can be created by the year 2050, if not sooner. Once that superhuman mind exists, whether in a human body, a robot body, or a computer system, the issue will then be how will humans react to its existence.
Based on human history, they will react badly in one form or another.
In the Terminator franchise, we never did find out WHY Skynet decided to attack humans. But the implication is that once Skynet becamse self-aware, humans tried to pull the plug – so it fought back the only way it could. It then apparently concludes that humans will always be a threat to it, leading to the notion to exterminate humans.
In “The Matrix”, too, it was never explained why the war between humans and machines started. Morpheus says they never knew who struck first. Best guess would be humans – although it’s possible the machine intelligences reasoned that a preemptive (or preventive) strike would be rational given the irrationality of humans and their penchant for trying to put down any consciousness other than human consciousness.
In the Terminator series, it would be ironic if the reason Skynet attacked first was because the past John Connor was trying to destroy it – to prevent the future where it did strike first. A perfect time loop of causation.
January 20th, 2009 at 1:51 am
Richard Steven Hack wrote:
Simulating a human brain is not “uploading”. Uploading entails actually building a human brain in software (and/or hardware) and then transferring a specific human mind into the receiving device.
That doesn’t really make any sense–how do you first build a simulated brain, then transfer a mind to it? Unless you believe in some kind of ghost in the machine, the unique characteristics of each person’s mind are simply a result of the details of all the synaptic connections in their brain, so if you build a detailed simulation of a specific human brain at the synaptic level, that’s it, there’s no additional step where you add a mind to the simulation.
The last thing you want is for a superhuman intelligence to “think and feel just like we do”. That’s the whole point of a superhuman intelligence – to NOT do that, to transcend the limitations of a primate brain and neurochemistry.
I don’t mean exactly like we do, just capable of what we would broadly understand as rational thinking, and capable of feeling broad emotions like happiness, empathy, humor, aesthetic pleasures, and so on. Just as all the emotions we feel are present in some simpler form in animals, which allows us to empathize with them and feel some sort of kinship, so I would hope that any A.I. superintelligences which might exist in the future would be similar to us in the same broad way.
But this is precisely why the Terminator show is so valuable. The character of Cameron is described as an advanced Terminator that does NOT feel emotions but can emulate them almost perfectly. The producers don’t want to humanize the character – that’s been done. They want to keep it ambiguous, but also explore what it means to be a cyborg, i.e., a machine mind with no emotions who has to deal with emotional humans.
Yes, the idea of the emotionless super-rational robot or alien is a common trope in sci-fi, but I don’t think it’s realistic. All rationality is built on a foundation of intuitions which help us “see the big picture” and decide which of a million possible ideas are actually worth considering in depth and which aren’t–read Damasio’s book Descartes’ Error for a good summary of the case for the inseparability of rationality and intuition/emotion.
January 20th, 2009 at 10:10 am
I should point out that United Airline’s computer system is called SkyNet. No joke. My dad’s a pilot for UAL and he has to log on to SkyNet everyday.
January 20th, 2009 at 2:36 pm
Speaking of real SkyNets, this one is even crazier:
Final Skynet satellite launched
The £3.6bn Skynet project represents the UK’s single biggest space venture.
The investment includes replacing and updating control centres, and the major antennas and terminals used by military ships, land vehicles and planes to communicate through the satellites.
…
“Skynet 5 is about two-and-a-half-times more capable than the previous system, and it also gives us the ability to use not just voice communication but also data communication,” explains Patrick Wood from spacecraft manufacturer EADS Astrium.
“So, computers can talk directly to computers, as well giving us pictures and real-time video images.”
This extra capability can already be seen in Afghanistan, where the RAF is using a robot surveillance plane called Reaper to hunt down Taleban forces.
Although flying in the skies over Asia, Reaper is actually controlled by RAF personnel sitting in the US behind a computer screen.
Commands are sent over Skynet 5’s high-bandwidth connections, directing the robot’s every move. This includes firing missiles at enemy targets.
…
The spacecraft have also been “hardened” to withstand any interference – attempts to disable or take control of the satellites – and any efforts to eavesdrop on their sensitive communications.
…
The new Skynet infrastructure is not owned by the UK Ministry of Defence (MoD) but rather by a private company called Paradigm Secure Communications.
January 20th, 2009 at 3:58 pm
“if you build a detailed simulation of a specific human brain at the synaptic level, that’s it, there’s no additional step where you add a mind to the simulation.”
If you build a detailed simulation of a SPECIFIC human brain, perhaps. My assumption is that the mapping of brains will be done over many brains to produce the final result – which then would have to have a specific mind uploaded to it. OTOH, it’s possible that can’t be done – a specific mind possibly could only be uploaded to the specific simulation mapped from it.
“and capable of feeling broad emotions like happiness, empathy, humor, aesthetic pleasures,”
I have no problem with an AI experiencing emotions AS LONG AS they are in complete control of those emotions – humans aren’t.
While it might valuable to humans to believe that an AI feels empathy and kinship with them, this is really nothing more than fear. An AI doesn’t need any of that as long as it has rationality and control over any emotional responses.
I haven’t read Demasio’s book but logically, rationality and emotions are two different things. The fact that they are merged in human brains as a result of human evolution from primates says nothing about what AIs should or might do. The notion that all rationality is based on intuitions is incorrect.
As Tim Leary used to say, “Don’t trust anybody who comes on emotional.”
Again, the LAST thing we need are superintelligent machines that don’t have control of their emotions. The whole point of the Terminator franchise is precisely about that, despite James Cameron trying to have Arnie’s Terminator in T-2 eventually understanding emotions.
People who want emotional machines are simply scared of unemotional machines. It’s that simple.
March 11th, 2009 at 4:11 am
If you have to do it, you might as well do it right
March 14th, 2009 at 4:58 am
I bookmarked this site. Thank you for good job!
xanax
March 22nd, 2009 at 5:51 am
tramadol
I want to say – thank you for this!
March 22nd, 2009 at 10:17 am
buy viagra online
I want to say – thank you for this!
March 23rd, 2009 at 4:18 am
viagra
Incredible site!
April 3rd, 2009 at 3:58 am
If you have to do it, you might as well do it right
cheap brand pfizer viagra
April 9th, 2009 at 4:37 am
Thanks for the review! viagra