Monday, December 13, 2010

Four Classes of Robotic Armaggedon

One estimate puts the number of robots dwelling among us at 11 million (as of mid 2010). If computers serve as guide, we can expect that number to increase rapidly as the years go by. Were the number of robots to increase at a rate of 10% a year, we'd have more robots than the present day human population by 2080. Self-replicating robots could cut that time frame dramatically.

Knowing that 11 million automated beings hide in factories and homes across this globe makes me a little worried. That's not an inconsequential army. Granted, almost all of those bots are highly specialized factory works who couldn't lift a gun if you wrote the software for them, but are we entering the era of the robot, and consequently, the threat of robot uprisings? When should we really start worrying, and what should we do if a mechanical judgment day comes?

The first step is, of course, precaution. The best way to avoid a robotic uprising is not training them to kill. We may have missed the boat on that one. At least the militaries of the world realize the importance of having a human making the shoot/don't shoot decision. Still, it's just a single feature requiring malfunction or tampering. The lack of strategic coordination among the robots also mitigates the threat, thus requiring a programming evil villain in the loop at present. You should be safe in the near term.

In the long term, who's to say? Of course, not all uprisings are created equal. The first question is if the robots have a shot at winning. Without access to heavy weaponry, extensive infrastructure support, and all-terrain movement, the uprising should turn out to be an inconvenience, not an Armageddon. What if we're not so lucky? I'd still divide the outcome into a few categories: bad, terrible, really-really-terrible, and not-all-bad. Let's hope for the last.

Robotic Uprising, the bad kind
This would be a low-intelligence, but highly effective attack by the robots. The most probable cause is haywire nuclear missile control, but a roboticized military turning against us isn't inconceivable. It's bad, because civilization gets destroyed. It's not terrible, because a clock starts ticking as soon as the attack begins. Robots break down. Nuclear arms are used up. As long as some humans living in remote locations make it a year or two, humanity can be rebooted. Even if we don't make it, the life left behind will have a chance to evolve its way to intelligence again. The most vicious nuclear onslaught is likely to leave life behind, if only in the depths of the sea. If surviving life evolves intelligence quickly, they'll even have the ruins of our civilization to learn from.

Robotic Uprising, the terrible kind
Much worse would be self-perpetuating robots with a dislike for biological sentience. Self repairing robots using solar energy could see to it that no new life takes hold as long as the sun keeps shining. Here, a small band of humans couldn't hold out past the end of days to see a new dawn: once the automatons take over, it's over. This scenario requires a much more sophisticated robotic ecosystem. The lesson, though, is that once we have self-perpetuating robots, we're moving into dangerous territory.

Robotic Uprising, Really-Really-Terrible kind
The last scenario had all life, or at least all life more advanced than a fish, being eradicated from the Earth forever. Is there really a more dismal scenario? Yup. I see it playing out much like the terrible version, with one crucial difference: robotic seeder ships. While we're still all buddy-buddy with the robots, we start constructing robotic settlement ships. They're built to fly to distant star systems, maybe construct a city and incubate a few humans when they arrive. Then they mine the local system, and set about building new probes to colonize new worlds. We'll explore and settle the stars!

Until the robots decide they don't like biological intelligence anymore. Then the seeder ships are re-purposed to seek it out and destroy it. In the "terrible scenario" we blotted out life on Earth, but we're just one of a trillion or more planets. In this scenario we set in motion an orchestrated effort to destroy all life anywhere. Whoops. Truly, a disaster of the greatest magnitude. If such a scenario seems at all possible, I hope we'll stockpile a planet destroying supply of fission, or anti-matter, bombs. Better that the Earth be turned into dust then risk eradicating life everywhere. Of course, once the seeder ships are out there, it might already be too late...

Robotic Uprisings: not all bad
I saved the least bad for last, so as to not end on the down note of universal extinction. I believe that the proper way to evaluate the consequences of any disaster are in terms of life, and in particular, intelligent life. The destruction of, say, a nation would be a tragedy, certainly. But in the grand scheme of things its a small blip. The miracle of Earth is first, life, and second, intelligence. The degree to which robots muck that up is the degree to which the uprising is bad, or really really really bad.

And the thing that makes life special is evolution. The biosphere is constantly changing, and usually improving. We went from microbes blindly reacting to environmental cues, through varying degrees of braininess, to the human mind, capable of unlocking cosmological secrets, capable of creative invention. Someday we may build a machine that's at least as capable of evolving as ourselves, where generation after self-replicating generation of machine is more intelligent, more creative, better than the last. Such a system could very well leave humanity behind, achieving intellect unlike anything we currently imagine. If such an improving, super-human creation were to turn against humanity, it'd just be Darwinism, in a way. Competition among species is as old as time, and if we create a truly superior specie, it might out-compete us in the end. I see nothing wrong with fighting such a creature: we need not go gently into the dark night. But, unlikely the other scenarios, mutual destruction would no longer be necessary. As long as life, be it biological or mechanical, is improving itself, things are well enough in the world. If we achieve any degree of artificial intelligence, I suspect we'll eventually reach this mastery of the art. The important thing, then, is seeing to it that we don't let the robots get out of hand, at least until that time.

Wednesday, December 8, 2010

An Ion Blast from Low Orbit

Sometimes I start to write about a topic, and every few sentences find myself resisting getting pulled down some new tangent. These topics can be approached in so many different ways, it's hard not to keep writing until a book comes out. Attempting to expose the backstory for the original point I wanted to make, I'll find myself repeatedly backing up, deciding there's something else I should tell the reader first.

Wikileaks is that topic d'jour. Why has its release of diplomatic cables caused so much more furious a response than anything else its leaked. What's the legal and moral implications of the event? How is everyone reacting, and where's it all leading?

But what I originally wanted to talk about is the fact that Anonymous is attacking the credit card companies.

Anonymous is hard to characterize, exactly. It's a loose affiliation of computer savvy individuals, who've taken on such organizations as Scientology and the RIAA. They fight against censorship, against perceived injustices perpetrated by major corporations against defenseless individual citizens. And lately they have dropped their attention from antipiracy organizations to focus on the financial behemoths behind the economy.

While previous leaks by WikiLeaks have created some controversy, the fallout over the release of classified diplomatic cables is somewhat unprecedented. Many members of many governments throughout the world have been falling over each other to condemn the organization in the strongest words possible. It's a terrorist organization whose members should be assassinated, according to some. Arguably in response to this rhetoric, companies with links to WikiLeaks have been quick to sever them.

Amazon dropped its hosting of the site. Paypal froze its account, refusing any new contributions. Mastercard and Visa have also banned any payments to wikileaks through there systems. The effect of all this is to deny the organization access to capital at the same time its facing major technical attacks and legal battles. It's an odd situation, legally. WikiLeaks has not been formally accused of any crimes, but the attempts to remove it from existence are not really government actions. Is Mastercard in the right deciding that certain organizations don't deserve access to donations? This is one of those tangents I'm going to avoid going down.

Anonymous has sided with Wikileaks, arguing that the companies are in the wrong for trying to cut off a whistle blowing organization, that regardless of the moral questions around the appropriateness of this set of leaks, it's not the government or private industries role to silence undesired speech, even the revelation of secrets. So it's attacking, in its own peculiar form. As a primary internet based group, it fights through the dissemination and stopping of information. Some see it as a illegal mob, others as modern activists: another tangent.

The primary weapon anonymous uses is the DDOS, a technique to bring websites down. A computer server can only handle so many requests for a webpage at a time. By running code to, in essence, refresh a webpage all day, you can slow the page down. By running that code on thousands of computers, you can block entry by anyone. So for much of today mastercard and visa websites (but not the transaction processing servers) were down.

The code in question is called the "Low Orbit Ion Cannon", and can be downloaded by anyone. I find people's participation very interesting. A DDOS attack is a crime. Participating could involve a multiyear jail sentence. Do they not know? Not care? Figure in such a large crowd they won't be singled out? The victim can easily log the requests coming in, and through subpoena's find out who the attackers were. But despite the risks, real or perceived or ignored, people download the code and let their computers bring down credit card websites.

Thinking about it, I've realized they may be protected by an odd ally: the virus. It's easy to view the computer as just an extension of the self, but it is not necessarily only in our control. A traditional DDOS is not a group affair, but the tool of virus writers. a DDOS requires many computers spread out between many networks. Virus ridden machines will often spring silently to life to attack a distant server, without the owner noticing anything except perhaps a slower than usual internet. These botnets are almost certainly also involved in the attack against the credit card companies.

So how do you know who was attacking, and who just had a secretive virus buried in their machine? I suspect computer forensics could tell, but after the first round of trials for this, that would change. Activists would just visit unsafe sites, download trojan-laced programs, knowing that they were helping the attack while retaining plausible deniability.

This technique doesn't stop there, either. There have been horror stories over the years of viruses that pull kiddy porn onto your machine. I suspect these are written by the purveyors of such filth to avoid having to host the content themselves: much safer to let anonymous infected computers handle that risk. But if it's not already used as a screen, I suspect it will be eventually: a virus that downloads inappropriate material to your machine for you. Thousands will be infected unknowingly, a few will seek out the computer infection for the files, and how do you ever tell the two groups apart?

Identity theft was just the start. With computers in between us, it becomes impossible to test for intention. As more of our lives go online, as more crimes are committed in the digital ether, detection of crime may become the easy part of law. The hard part would be figuring out who the computer committed the crime for.

Friday, December 3, 2010

Comcast and L3

I always read the comments on online news. I think it's a wonderful view into the process of society moving. Interactions, the spread of ideas: it's all stored in 1's and 0's for us to examine.

The internet feels nebulous, but it's built on very real hardware spread across the world. No one entity owns the internet, instead millions of companies and billions of people each own a little piece. The science and technology of the connections, the mathematics of the virtual information speeding down copper wires, the economic transactions between billions that makes it all work: it's an interesting topic, but in many ways an esoteric one. Like most other things in moden society, a small group of people specialize deeply in each aspect, and make it work.

So it's interesting when the small details are brought to the broader attention of the public. This happened recently in an ongoing dispute between Comcast and L3. Each presented a different set of facts to the public, and in online news and the related comments sections I've watched as consensuses formed. First, there was anger at Comcast, and accusations this was about stopping Netflix from replacing cable tv. Then Comcast's arguments were considered, and the discussion turned towards absolving Comcast. Maybe L3 was in the wrong here? People taught each other about peering and CDN's, short arguments were developed explaining sides, and last I saw people were solidifying their views, the debate seeming to be won by the L3 side, but with moderate disagreement as is always true in a controversy.

The very short of it is that Comcast, provider of internet to residential areas, and L3, who owns heavy duty connections between far flung ISPs, had a peering agreement. This said that since they were sending about the same amount of data to each other, there was no need to calculate detailed bills. It was a wash, they carried each others data for free. Netflix contracted L3 to stream movies for it, Comcast claimed L3 now owed it money, and L3 said Comcast is just trying to block Netflix.

It's murky, because there are two types of data transfers. If a Comcast customer sends an email to Verizon, and it travels across the L3 network, L3 gets paid. Comcast is paid by its customer, Verizon by its customer, but nobody is paying L3. In contrast, if Comcast and Verizon are directly connected, it's not clear either should pay. Even if Comcast is sending far more emails to Verizon, it's in both companies interest to make sure the email gets through. If not, both have angry customers.

So in that regard, Comcast is sending data across L3, but L3 is supplying Comcast with data it's customers explicitly requested. Comcast argues that the previous Netflix provider did provide them, and that L3 is looking for a unfair advantage. L3, and before it Aakamai, are acting as CDNs, Content Distribution Networks.

But again, the term is ambiguous. In one sense, both L3 and Akamai distributed content, so they're in the same business. But they did it dramatically differently. Aakamia essentially paid Comcast to host the content, acting as a middle man. L3 already has networks across the country, so it just places the computers on its network. In a sense, the computers just move a few miles. But now, they're moving around like the rest of the content on the internet, a middle man was just removed.

It's these subtle distinctions that define the situation. Is a CDN distributed content hosting, or is it a specific way of doing it, purchasing service from your customer's ISP? Is "peering" just traffic that passes across a network elsewhere, or does it include traffic destined for the other network. But these distinctions have the potential to dramatically change who makes billions and who goes under, what sort of internet you get and for what costs, even whether things like high fidelity video conferencing will change business or not.

It's been heartening to watch thousands of people online all unraveling, at least a little, the interconnections and implications of the topic.