The Power and the Money

Propaganda in the age of artificial intelligence

Have you ever watched television commercials from the 1950s? They seem crude and unpersuasive, even silly. Political propaganda from the 1930s even more so and don’t get me started on print adverts from the 19th century. They actually persuaded people?

Well, yes, they probably did. Our culture has evolved in the presence of ever-more sophisticated advertising. Our defense mechanisms have also risen; ads either have to very sophisticated or do nothing more than present basic information in order to succeed. (See “Len the Plumber” for an example of the latter. All it does it tell you that you can get same-day service at that number; considering the cost of searching for a same-day plumber, it gets business despite presenting no information other than that it exists. And an annoying earwig to make sure you do not forget.)

New technology, however, allows for new ways to get around our cultural immune systems. We have seen the rise of Fox (cable television being the technology)  and fake news (going through social media). But two things are about to make the situation incredibly worse. (Hat tip: Will Baird.)

1. It is getting easier and easier to fake video of people saying things that they never actually said. It is no coincidence that the researchers at the link used footage of President Obama to create their fake clip. 
2. It is getting easier and easier to automate fake online comments and reviews, aka “crowdturfing.” The authors propose some potential fixes which might work for review websites ... but maybe not so much for social media. 

I am fairly sure democracy can survive these innovations. I am also fairly sure that democracies will work less well and lose legitimacy because of them. My worry is that our cultural immune systems will never adapt, or worse yet, adapt by crawling into our own epistemic fortresses.

The 2020 election is shaping up to be awful; 2024 more so; and I do not want to think about 2028 or 2032, when my children will be able to vote. 

Anyone able to provide an optimistic counterargument? Surely some techno-optimists are left, or any form of optimist, really. 

My entry in the children smartphone battles

There is a growing but completely inconclusive literature about young children and screens. I have nothing to add to that debate.

What I can say is that I honestly do not understand when people write things like this:  “For those who, like me, often resort to screens to distract children long enough to get dinner ready or siblings dressed, the news isn’t good.”

You need to resort to screens to make dinner or dress your children? I ... uh ... really? This is a bit like saying that you needed to use your refrigerator in order to start your car. I do not see it. Do I and all my acquaintances just have remarkably well-behaved children or is this writer just spouting off? 

The future is not all bad

My reaction to most new technological developments at the level of personal tech usually falls into one of two categories. (1) “Who gives a fuck?” Fitbit, any new phones since 2007, Google Glass. (2) “This is going to make life worse.” Intelligent personal assistants (Alexa, Siri, et al), autonomous flying drones, ever-smaller cameras, virtual reality, whatever succeeds Google Glass. (Self-driving cars are in an intermediate category if they really bring down accidents.) 

In general, I am not a terribly big fan of the future I find myself in. I like TiVo, cellphones were nifty, search engines are an excellent thing and I wish that more books and datasets were digitized. Beyond that, I am neutral ... at best. Even blogging, hey, it wastes some time, but there is a reason I went off social media as soon as I got on it.

So ... I am happy to discover a list of new inventions that I really can get behind! Now, not all. The Naked 3D fitness tracker is creepy. Ori Systems robotic furniture seems useless. Awair smart air-quality monitor, pretty much ditto. And I cannot really see myself buying an OURA ring.

But the rest? Awesome! A future I can get behind.

Plus, my kids and I watched a spaceship take off from our attic window last night.

Timothy Lee is not making sense

 By now, you have seen the below photo and the assorted reactions on Twitter. (I refuse to write “Twitter reaction” or “internet freakout,” because neither Twitter nor the internet have volition.)

Timothy Lee at Vox then proceeded to write a silly article under a quite reasonable byline: “But the internet's collective freakout misses an important point: No matter how ridiculous it might look to us today, there’s every reason to think virtual reality is going to be a big deal in the next few years.”

The problem with Lee? The reactions on Twitter were not saying that VR is ridiculous and will thus never become common. They were saying that VR will very clearly become common ... and that portends a future that is in equal parts horrible and ridiculous.

And it is coming even if Will Baird is correct in the fascinating debate happening in comments over here ...

Moore’s Law has been repealed

 This blog has evinced some concern about the social effects of rapid progress in information technology, especially as computers using massive statistical techniques appear capable of doing more and more jobs that used to be the domain of humans.

The bad news for the Luddites among us (I probably should count myself as one) is that the hardware ship has already sailed. Even if there are no more advances in computer hardware for the next fifty years, our software still has not caught up. There is a lot of disruption coming down the pike. 

The good news is that Moore’s Law may be dead. In 2012, I wrote that it might have a decade left. But now Intel has announced that not only will chips stop getting faster ... they will start to slow down as engineers try to focus on ways to cut energy use. 

More energy-efficient computers promise a host of changes, some good and some not, but they are unlikely to wake up one morning and decide to stuff us all into tubes.

Ghost Fleet #4: Anonymous is no match for the NSA

 Ghost Fleet’s climactic cyberwar scene reached a depth of stupidity suitable for a SyFy made-for-TV movie, only without the self-awareness. It would have been laugh-out-loud funny had the authors not been telegraphing what was going to happen for about 100 pages.

We have already covered most of what is wrong with the cyberwar climax on page 288 of Ghost Fleet. At the start of the scene, a Chinese superhacker is moving to shut off Akron’s power supply. The set-up, of course, makes no sense. The superhacker is attacking using software “modeled after the Americans’ Project Aurora malware.” You can go to this video to see what a cyberattack really looks like. She would not be flicking rings to “initiate attack protocols.” Maybe she would fool around with the utility’s control panel after the malware had executed, but she wouldn’t be “initiating” anything. 

Second, and more importantly, there is no way on God’s Green Earth that the United States would still be vulnerable to the Aurora vulnerability more than a year (or two; the book is remarkably unclear on chronology) into the war. Consider: “Mitigation is relatively inexpensive and straightforward. It entails placing a hardware device between each substation and its loads. That device monitors for the rapid out-of-phase condition associated with an AUORA event and isolates the substation from its loads before the torque of the grid can be applied to the equipment loads.” Sure, the United States might be vulnerable at the start of the war. But if the enemy doesn’t attack immediately, then it won’t be able to attack at all.

But the real idiocy is at the end of the scene, where our Chinese attacker is thwarted by … Anonymous. Yes, freedom’s secret weapon. Anonymous get a very Hollywood moment where the Chinese hacker’s virtual reality gear and interface to her hacking tools is obstructed by a rendering of a wall of Guy Fawkes masks. Anonymous then proceeds to take over the systems that keep her drugged up (WTF?) and kills her and all the rest of the cyberwarriors.

Anonymous is presented as an ally of convenience than “good guys” or “patriots” for the U.S. At that point, the accuracy stops.

The book paints Anonymous as a coherent and effective force. The truth is rather different. Fact checking their cohesion and ability is not terribly hard. In fact, Anonymous just conducted an operation. One deemed #OpParis.

After the attacks in Paris, Anonymous announced they would go after Daesh, aka “the Islamic State.” They planned to take down twitter accounts associated with Daesh and compromising forums and systems associated with the same. After a week of their #OpParis, Anonymous reported taking down more than 20,000 twitter accounts and claimed to have doxxed several Daesh recruiters. (“Doxxing” means posting private information about names, locations, generally personal details, etc.)

Already one might start to wonder whether the authors of Ghost Fleet were grinding an ideological axe or (more likely) writing a scene they knew was immensely stupid in the hope of getting a movie deal. Taking down Twitter feeds and doxxing Daesh sympathizers is a long way from killing Chinese government operatives by remote control.

Except Anonymous is even less capable than than its claims about OpParis reports would indicate. After all, accounts that were supposedly taken down included such Daesh supporters as the New York Times, U.S. President, BBC and Hillary Clinton.

Wait! What?

Indeed.

Multiple sources reported most of the twitter accounts were active still after the so-called attacks. Twitter stated the lists of supposed Daesh accounts provided from some Anonymous members were wildly inaccurate. So is the claim they took down 20,000 accounts complete hogwash? Is there a way to reconcile Anonymous’ statements with the observations of the outside world?

Yes, actually, there is. At least for some of their claims.

To do so though will require examining how Anonymous attacked its targets. While there are a number of skilled hackers, the majority of Anonymous is little more than what we call script kiddies. These are users of prepackaged software meant to compromise systems. How such software works, we’ve talked about before. Such software is not, however, anywhere as near as sophisticated as Stuxnet or other real cyberweapons.

Anonymous essentially spammed the Twitter accounts so people could not, while under attack, use them. They sent wave after wave of messages including Rick Astley’s 1987 Song “Never Gonna Give You up.”

Yup.  Essentially Anonymous rickrolled the supposed Daesh accounts, spamming all incoming communication and making conversation impossible. The accounts were taken out … but only temporarily. As soon as Anonymous stopped their attacks, the accounts were usable again. This is what is called a denial of service attack, and, frankly, it’s a weak one. In other words, Anonymous didn’t take anything out. All it did was make a few accounts temporarily unusable until Twitter’s management stopped the nonsense.

That brings us to the greatest problem with Anonymous: its a volunteer organization attracting support based on enthusiasm and interest. Its a militia, of a sort. But it not a militia driven by concern for people’s homes or life. Rather, most Anonymous volunteers do what they do for the “lolz;” their own amusement. When they get bored, they stop. #OpParis petered out. The Islamic State steams on.

I’ll mention here that when Anonymous decided to doxx people who collaborated with the Zetas, they learned right quick that it was a bad idea. As for other ways of cyber-attacking the cartels, so far there has been nothing. The cartels are not running scared of Anonymous. 

To be sure, there are probably talented hackers out there affiliated with Anonymous, but the vast majority of them are not. That majority is the cyber equivalent of bunch of high school kids egging a house. Nothing more.

And certainly not the insanely powerful group portrayed in Ghost Fleet. And I see no way for the current group to somehow transmogrify itself into an all-powerful collective in only nine years ... or ever.

In short, the way the authors portrayed cyberwar may succeed as a way to catch a Hollywood agent’s attention. It failed on every other metric. Which is really quite disappointing, given that they have the knowledge to have written a good book.

Ghost Fleet 3½: Really, it makes no sense that the U.S. would not strike back

 Things are slowing down professionally, and so this series will continue. But I just finished re-reading the book ... and man it looks much worse than the first time through. Really shoddy work by the authors in what should have been a classic. I am befuddled by the positive reviews.

Any way, here is Politico (of all places!) reporting on America’s arsenal of offensive cyberweapons. Key quote: “One thing everyone I spoke to agreed on: The U.S. has the most powerful cyber arsenal in the world.” 

Weirdly, the article quotes one of the authors of Ghost Fleet. I know that they know all the stuff Will Baird explained here. So why does the book pretend that they don’t? They really could have done much better ... and realism would have introduced more drama, not less. It is weird.

Next up: the silliness of the way the book treats insurgency and counterinsurgency, followed by fisks of how it presents international politics, conventional military capabilities, and space travel. (Oh, God, the space travel.)

Ghost Fleet #3: Cyberwar is symmetric

 One of the most disappointing things about Ghost Fleet was the dog that didn’t bark: the massive immeidate American cyber-counterattack against China. It’s one thing to write an exaggerated warning against a threat that you believe your country is neglecting. It’s quite another to ignore the loud and massive preparations for cyberwar that the United States has already made — and that those very authors have reported upon extensively. It was weird; weird enough to remove much of our ability to enjoy the book.

One thing that the authors get completely right is that much of any future cyberwar will be fought using malicious programs that are pre-positioned inside hardware. Trojaned electronics are a very real concern. It is entirely conceivable that in a future conflict, the Chinese government could do serious damage to the American civilian infrastructure. Moreover, with the advent of the “internet of things” such an attack — when combined with the more “traditional” operations described in our last post — such an attack would consist of a serious Weapon of Mass Annoyance, as toilets seized up and heating systems went haywire. To their credit, the authors evocatively show just such effects: a home front dealing with unfixable automated refrigerators.

Unfortunately, the authors jumped (a lot of sharks) far away from the great fictional war classics like Red Army or Arc Light or The War in 2020. They moved out of the realm of the plausible. Let’s start with the idea that the Chinese could install Trojan horses in most of our military systems. They couldn’t. There seems to be a fundamental misunderstanding on how some of this works. Adding a modification to a chip is a pretty profound process. It is highly likely to add glitches that would be easily detected. It would likely add significant time to delivery as well … and sudden additional delays for manufacturing a relatively generic chip would catch people’s attention.

Field-programmable gate arrays (FPGAs) are even less susceptible to Trojan horses. (Elsewhere, the authors of Ghost Fleet have pointed out in ominous tones that two-thirds of all the U.S. military’s field-programmable chips are made in China and Taiwan. We’ll ignore the fact that Taiwan is not in fact a part of China but an ally of the United States and pretend that all the key chips come from the People’s Republic.) FPGAs are not where you need to worry about Trojan horses because FPGAs are blank slates: the customers can put whatever programs they want on the chips. There is no way for the manufacturer to know what the customer (whether the U.S. Air Force or Intel) plans to put on the chips.

Remember, in order to subvert a program, a hacker must have an understanding of a program. But there is no way to know what program will be placed on a FPGA. Furthermore, many FPGAs allow for a program to be permanently loaded —  written once, read many –  such that any attempt at overwriting with a new subverted program would be impossible.*

Moreover, the United States has a trusted foundries program administered by the National Security Agency to make chips for the military. (Bizarrely, the authors have the trusted foundries play a key role in the successful U.S. counteroffensive.) Now, the trusted foundries do not make the United States invulnerable. But it does make a breakdown on the scale depicted in Ghost Fleet very hard to believe. The worst case scenario is an attack allowing for the crashing of the software through understanding the hardware —  garnered through espionage.

Recovering from a cyberattack based on Trojaned electronics would “merely” require resetting the software. The scare quotes are because such a reset would be far from simple and would take a substantial amount of time. During a battle, such an attack could be catastrophic. We believed in the U.S. defeat during the initial offensive. (Save perhaps the actual invasion of Hawaii: more on that later.) But over the course of a war, the problem is entirely addressable, and the U.S. should have gotten a handle on the issue much faster than it did in the book.

But to be fair, while the scale of the attack may have been hard to believe, but it was not impossible to believe. Yes, the details were wrong, but the United States could in fact get punched in the nose as hard as it was in the opening battles of a future war.

What was impossible to believe was that the United States would not strike back, hard. The U.S. is already training offensive cyberteams. And this is the latest part of a long-standing program — viz the fact that the only documented cyberattacks in the history of the world have come from the Israel and the United States. Hooah, indeed.

As for Trojan Horses hidden in hardware, the Chinese have far more to fear than we do. They don’t know what they are manufacturing! It’s far more likely the U.S. is slipping in something into their chips than vice versa. Not because the U.S. is miraculous, but because we are doing the design. Chinese firms, for the chips made there, do the manufacture. It’s easier for us to verify that a design is what we think it is than it is for them to verify that a design does what they think it does.

Now, perhaps the United States isn’t installing Trojan horses in Chinese electronics. We’re perfectly willing to accept that as a premise. (Even if it’s improbable.) And we are willing to accept that the Chinese could inflict enough damage on American systems as to allow them to win a naval campaign in the Pacific. What we’re not willing to accept is that the United States would be unable to strike back against Chinese infrastructure.

A realistic counter strike would cripple the Chinese in a manner in parallel to the United States. Those same fridges taken out in the U.S. are likely to be taken down in China. The irony of the Chinese situation is with their modern infrastructure, they are likely to have far more of their critical infrastructure taken down in a cyberwar scenario. The analog systems of yesteryear are completely immune to the hacking of digital systems. Digital upgrades will be very uneven in the U.S. They will be almost ubiquitous in China. The Chinese could pummel the United States. The United States would equally pummel the Chinese in a cyberwar. The authors highlighted Los Angeles and the problems with fridges, a metropolitan area currently of 13 million people. What happens when Beijing’s 21 million can’t store food either?

A realistic scenario for cyberwarfare would be the race of the cripples: compromised systems on both sides will cripple each side’s military and homefront in unexpected ways. Both sides would go into battle with their equipment malfunctioning. Both sides would find some of their premier weapons unusable. No one has a monopoly on cyberwar. As of 2015, the U.S. has a significant lead, despite what the authors imply. It is certainly possible that the U.S. would lose that lead by 2024, but it is simply impossible that America would lose the ability to strike back.

The revelations about Stuxnet and the Equation Group have made it clear that the United States has adapted its Cold War approach to nuclear warfare to the new world of cyberwar. During the Cold War, it was theoretically possible to defend against nuclear weapons on a large scale. (Using nuclear tipped interceptors, for example.) But the costs — and there were many! — were deemed too large.

Instead, American war planners adopted an approach that worked surprisingly well: mutually assured destruction (MAD). If a power launches nuclear weapons against the U.S., then the U.S. will literally destroy them in response.

The U.S. could construct an economy invulnerable to cyberattack, but that would slow progress in development of computers and the internet to a crawl. Moreover, it would require civilian discipline unprecedented outside of the two World Wars: does your grandmother always update her computer? Or your uncle his phone? Your child her tablet? Or your husband his new smart thermostat?

Rather than attempt the impossible, the U.S. has applied MAD to cyberwarfare: China unleashes cybernetic hell on the U.S. and the U.S. will unleash it in turn on China.

Finally, there is no way that a cyberwar on this scale wouldn’t go global. The authors themselves have noted in other venues how Stuxnet spread to systems far outside Iran. The same would happen with Chinese and American cyberweapons unleashed in a Third World War. To some extent, that’s a minor point: why should the reader of a good WW3 yarn care whether municipal sewers and automated metros in Lyon are going haywire? They shouldn’t. But the rest of the world will be affected and they will be angry … and in a scenario where China strikes first, they will be angry at China. Maybe America’s NATO allies would refuse to invoke Article 5. (We’ll get to whether that is believable in a later post.) But there is no way on God’s green Earth that they would pretend as though nothing has happened.

In short: the scale of the effect of the initial cyberattack on the U.S. is believable, even if the depictions of how it was carried out make no sense. But it is improbable. And it is completely impossible that the U.S. wouldn’t strike back. We can believe a Chinese victory in the first phase of the war. But one of on the scale depicted was laughably silly and the lack of an American counterstrike was ludicrous fantasy.

And so, once again, our disbelief was desuspended. If you can’t believe the scenario is possible, then what’s the point? These guys know enough to have done much Much MUCH better.​

Ghost Fleet #2: Cyberwar is boring

 Neuromancer set the template for fictional depictions of hacking: colorful icons swooping through cyberspace in a glorified videogame. Sometimes the depiction is a little more sedate, but only a little: hackers frantically banging at keyboards as they break enemy defenses. Films like Swordfish or Hackers and many, many others make it seem rather glamorous, even visual. 

Ghost Fleet falls right into that trap early on. Page 41: 

“The young hacker focused on her attack, navigating the malware packet through the DIA networks while fighting back the desire to brush a bead of sweat off her nose with her gloved hands. The Pentagon’s autonomous network defenses, sensing the slight anomalies of her network streams, tried to identify and contain her attack. But this was where the integration of woman and machine triumphed above mere ‘big data.’ Hu was already two steps ahead, building systems components and then tearing them down before the data could be integrated enough for the DIA computers to see them as threats. Her left arm coiled and sprung, her fingers outstretched. Then the right did the same, this time a misdirect, steering the defense code to shut down further external access, essentially tricking the programs into focusing on locking the doors of a burning house, but leading a small ember on the outside for them to stamp on, so they’d think the fire was out.”

Needless to say, this is not how hacking works. 

Hacking comes in two major flavors. The first is the all-out assault. The second is the stealthy compromise.

The Distributed Denial of Service (DDOS) attack is a prime example for the all-out assault. A large number of software bots are surreptitiously placed on computers all over the world, usually by getting someone to click on an infected link or attachment. The software then automatically downloads itself and send reports into a controller server (itself likely another hacked system). The number of bots can range up to the hundreds of thousands and be directed by a small handful of servers. When the attack is launched, all those computers fire off requests in such volumes to cause the receiving computers to lock up. This is somewhat like having several hundred 5-year-olds ask questions of one person at the same time.

A variation on this is a massive port scan: using the same botnet, hackers can check all the different places software listens to for communication. The bots will then try to crash the software listening on that port, hopefully getting the system to cough up another entryway in response, like a username or password. The botnet is effectively looking in all the computer’s pockets for a hole from which it can swipe the key. Again, the hacker “merely” pulls the trigger after aiming the botnet. You could, I suppose, have a scene of the hackers feverishly writing the code before they launch the attack. But nobody is going to be sitting in a room swooping through 3-D displays. No typing, no visuals.

Then there is the stealthy attack. The special forces to the DDOS tank army. The classic is stealing access to passwords, preferably from a network administrator. The attacker then moves through the network slowly and carefully. The problem with this method is lots of places log their access and strange places often get flagged. If it happens too often, they lock accounts. That is part of the reason why many sites require periodic password changes.

The problem with this model — the hacker working hands on — is that it’s far more traceable and slow. Better is to compromise a system and let a bot go to work. In fact, far better is to let a bot just plain do all the work. It’s faster and safer.

The most common bot for this is the infamous “malware”: viruses, worms, Trojan horses, etc. The single most famous example of malware that went above-and-beyond was Stuxnet. The authors of Stuxnet didn’t sit and click or wave their hands in real time as they manuevered their way into the Iranian nuclear facilities. They had someone pop in a USB drive and the rest was automated. 

This makes it tremendously unsexy. The Stuxnet infection took a significant amount of time before it compromised the systems at the Iranian nuclear enrichment facility. And once it was in, there was no way to know if it had worked until the U.S. received human intelligence that the centrifuges it targeted had actually broken. The infected Iranian facility wasn’t hooked up to the internet, but even if had been, it would have been very, very dangerous to report back. Communications can be intercepted.

Now, the Chinese hacker is working in nine years in the future. Why couldn’t she use virtual reality to hack systems in 2024? Well, maybe she would write code using a V.R. headset and virtual keyboard, although I’m unclear as to why. But she wouldn’t interact with the cyberattacks once they were underway. That would just result in detection once the communications were intercepted. Moreover, the time it would have taken to send back data about what was happening and send back the response could be up to ten seconds on the round trip. Add another ten seconds for her reaction time. That is a lot of time for the DIA to realize something is wrong and start shutting the doors. 

The authors talk about her and her bots trumping Big Data. It’s a nice buzzword, but it is totally meaningless in this context. Big Data might have been used as a way to train the DIA’s intrusion detection system to recognize which data flow patterns are normal activity and which might indicate an attack. But I have no idea what it would mean for the Chinese hacker to be “building systems components and then tearing them down.” As a description of cyberwarfare, it’s not even wrong; it’s meaningless. 

In summary, the hacking in Ghost Fleet is derived from the silver screen, not the reality of today. It also is highly unlikely to be what comes about in ten years. 

Now, to be fair, these scenes are just window-dressing. There is, in fact, a scene on page 37 in which a Venezuelan refugee accidentally infects U.S. defense systems in a realistic way. (Although I have to admit I have some doubts that you could actually copy malware onto a security badge — the article cited in their footnotes does not describe anything like the scene.) But there are much more serious problems with the way the cyberwar develops in Ghost Fleet, which we will get to in later installments.

Comments and criticisms very welcome!

Your questions about space and the Singularity: Answered!

 Our new guest blogger, Will Baird, has a Reddit thread! I like Reddit. It reminds me of the good old days on Usenet, before all this weird-ass “social media” crap came around to destroy human life.

But this post is not to tell you to stay off my lawn. (Right now, with two young kids, I am very happy for neighborhood children to come over and play on our lawn uninvited.) This post is to tell you all to go read the thread! Even if Will isn’t currently still active on it, it’s fascinating. Go, read. 

Technology and social capital

 There are a lot of questions about how modern technology is transforming society and social interactions. (For modern technology, read smartphones mediated by social media software.) What is remarkable, however, is how little we really know about how older technologies affected society and social interaction. 

For example, television. How much do we really know?

Well, courtesy of Elisabeth Perlman, comes some evidence from Indonesia. From the abstract:

This paper investigates the impact of television and radio on social capital in Indonesia. I use two sources of variation in signal reception — one based on Indonesia’s mountainous terrain, and a second based on the differential introduction of private television throughout Indonesia. I find that increased signal reception, which leads to more time watching television and listening to the radio, is associated with less participation in social organizations and with lower self-reported trust. Improved reception does not affect village governance, at least as measured by discussions in village meetings and by corruption in village road projects.

That does not seem good. Maybe my Luddism does not go a far enough.

Elisabeth, I should add, has a neat paper on the effect of free rural mail service on politics in the early 20th century. From its abstract:

The rollout of Rural Free Delivery (RFD) in the early twentieth century dramatically increased the frequency with which rural voters received information. This paper examines the effect of RFD on voters’ and Representatives’ behavior using a panel dataset and instrumental variables. Communities receiving more routes experienced higher voter turnout and spread their votes to more parties. RFD shifted positions taken by Representatives to ones in line with rural communities, including increasing support for pro-temperance and anti-immigration policies. Our results are stronger in counties with newspapers, supporting the hypothesis that information flows play a crucial role in the political process

It is fascinating stuff.

Cohorts v. cross-sections and you are not your grandparent

 So there is a report out that “Digital savviness peaks sometime between the ages of 14 and 15, and then drops gradually throughout adulthood, before falling rapidly in old age.”

Your alarm bells should immediately start going off. Do people just suddenly start forgetting how to push buttons (or yell at their machines) once they reach a certain age? Is using simple information technology a skill more akin to catching a fastball, in that if you don’t develop it by the time you’re a teenager you’ll never get good at it?

Probably not.

So that has to rank up there as a highly highly misleading sentence. Moreover, the tech-savviness that they’re measuring has to do the use of digital applications deliberately designed to be workable by a total idiot. Like, say, myself. It isn’t coding, or fixing a busted keyboard, or chip-design skillz what they’re measuring; it’s knowledge of Kik and Snapchat and Vine and all that.

My father, who really could be called digitally-savvy in that he not only used the for-morons applications like Skype but learned to program in his sixties, he would not have been amused by this article. 

Thumbs down from Lenny, may he rest in peace.

Two cool links

I wish that Typepad had a “blogroll” function that I could put on the sidebar. Because you all should check into the Dragon’s Tales ever so often.

Two cool links today. Link #1: proof that I am not complete Luddite. Yeah, I worry about the robots. And I worry about the panopticon. But I do not worry about the flying cars and it is serious awesome that Advanced Tactics of El Segundo, California, is trying to build one for the U.S. armed forces.

Link #2: By age 23, 49% of black, 44 % of Hispanic & 38 % of white males have been arrested. That is fucking amazing. I came close to being arrested on two occasions, but it did not happen. My friend Guy can say the same, only more so. It is a good thing to be white in this country ... so I am surprised that the disparity in arrest records is as small as eleven percentage points.

What I am curious about is the result of the arrests. My null hypothesis would be that the consequences are far more serious for black arrestees than for white ones, but I do not know that for certain. Does anyone know of research?

I would also like to know if anyone has looked into how cumulative cohort arrest rates have varied over time. Were men who turned 23 in 1953 or 1973 or 1993 as likely to have been arrested as those who turned 23 in 2013?

I think a good social science history of crime in America is sorely needed.

About that website ...

American presidents are fairly weak as these things go. Really, he can do three things without Congress. First, use the military outside the United States. Second, issue executive orders ... as long as said orders do not include unilaterally sequestering funds or overturning existing regulations. (The Administrative Procedure Act of 1946 greatly limited presidential power in this regard.) Third, administer the executive branch.

And in this respect, the Obama administration has fallen down on the job. The damn website is an own-goal. More so since Obamacare is working fine in states that did the job themselves.

The wrong lesson to take away from this is that state governments do things better than the feds. There is no evidence for that, beyond vague small-is-beautiful prejudices.

The right lesson is that President Obama right messed this up for no particularly good reason.

Sadly, however, it is not true that presidents with administrative experience turn out to be good managers (viz James Earl Carter) or that ones without it turn out to be bad ones (hey hey LBJ). So there seems to not be an easy metric to judge how well somebody will do as an administrator before they actually need to administrate. Worse yet, overseeing the creation new stuff is a very different thing from managing an existing operation.

But ex post, even though Obamacare going to work out in the medium run (my apologies, Republicans, but liberalism is not over and a defective website gets nobody killed) the unnecessary mess seems to me to have shown the President to be deficient in one key skill for an executive: project management. The historical record should reflect that.

Am I being too harsh?

Too good to check

“If you wanna know how not secure you are, just take a look around. Nothing’s secure. Nothing’s safe. I don’t hate technology, I don’t hate hackers, because that’s just what comes with it, without those hackers we wouldn’t solve the problems we need to solve.”

— Fred Durst

The future is here?

Well, sort of. In 1964, Isaac Asimov in speculated about the World’s Fair of 2014. How did he do?

Well, we didn’t have a nuclear war. Check.

We have not landed people on Mars. Check.

“Robots [are] neither be common nor very good in 2014, but they [are] in existence” and driverless cars are beginning to be feasible. Check.

Communications are dirt cheap and “the screen can be used not only to see the people you call but also for studying documents and photographs and reading passages from books. Synchronous satellites, hovering in space will make it possible for you to direct-dial any spot on earth.” Check.

No mention of wireless phones, but that was already a trope of near-future science fiction. Heinlein had a thing for startling his mid-century readers by having characters pick up their phone in the middle of the desert or use “I left my phone in my other coat” as an excuse for not being in contact. It probably just did not occur to Asimov to add ubiquitous mobile. The essay did not include the myriad knock-on effects of cheap computers and the Internet, but it predicted them. So, with a caveat, check.

Wall screens will have replaced cathode-ray tubes for television. Check.

Large-scale solar-power. Check. (He also says fusion plants, but he says they will only be experimental. I guess you could claim that we have one of those already. Check.)

He worries about technological unemployment caused by automation, which is not here yet ... but we may be seeing it cause drops in wages. Give him points. ¾ check.

The rest, though? Ugh. First, he extrapolates his own odd agoraphobia onto the population. Yes, we could be building underground cities in 2014, but with the partial exception of Singapore (which is in sort of a unique position) we ain’t. People like, you know, sunlight and views. Plus it is expensive. Maybe ¼ check, and that only because of the plans in Singapore. ¾ fail.

His kitchen of the future could be built, but mostly has not. People like to cook. And it is easy enough to pop a prepared mean in the microwave; why fancy it up with a conveyor from the fridge? Fail.

He predicted that robots would still be bad, but they are even worse than he predicted. If you think we will have robot maids and gardeners by 2024, then I will give it to him on credit. Otherwise, fail. (Will? Carlos? Anyone else?)

Commonplace radioisotope batteries. Nuclear fission providing half of all electricity. (Asimov saw the future and it turned out to be Belgium!) Moving sidewalks in urban downtowns. No on-street parking in central cities. Compressed air tubes carrying intra-city cargoes. And moon colonies, of course. Fail, fail, fail, fail and fail.

Plus, 3-D holographic cubes and all high-schoolers learning to code. Fail, although itis probably our own failing that we do not teach all high-schoolers to code.

Alright then. He got the big stuff, but missed the implications, and added on a bunch of other big stuff that did not happen. (Most of which show little sign of happening.)

So what’s the right grade?

And will we still have World’s Fairs in 2064?

Robocapalyse maybe

Computers have made some startling advances lately in things like voice and object recognition. Put it together, and we are beginning to see some signs that our glorified typewriters-cun-filing cabinets are turning into the robots and computers of classical science fiction. Which brings us to the obvious question: what will they do to the labor market?

A recent paper tries to answer a more limited question: what percentage of jobs are susceptible to computerization? If that number is small, then there is nothing to worry about. If that number is large, then we have to hope that new human-only jobs will emerge that most humans can do ... or watch income inequality expand to unseen levels.

To make this analysis useful, you need to do three things. First, you need to estimate the limit capability of computers. That is, what will they be able to do once Moore’s Law comes to an end. (For the nerds: Koomey’s Law is just as important in this context. After all, a robot that consumed $500 per hour worth of electricity would not be very valuable.)

Second, you need to identify the current jobs that these hypothetical future computers will be able to replace.

Third, you need a methodology that provides you with an upper bound. Why an upper bound? Because the second thing above pushes your estimates upwards. After all, jobs change ... but any reasonable identification strategy will have to identify computer-susceptible jobs as they are and not as they could be. That already implies significant upwards bias, so in order to get a useful estimate one should run with it. That won’t help much if the percentage of computerizable jobs that you come up with is 98%, but it will help if it is, say, 60% or 40% or 25% ... it will provide a useful ceiling.

OK, so how does the paper stack up?

In terms of the raw theory, not too bad. They take a well-worn Cobb-Douglas production functionand replace labor with three labor inputs: perception and manipulation tasks, creative intelligence tasks, and social intelligence tasks.

In terms of its time horizon, it goes out about two decades, to 2035. “The main challenges to robotic computerisation, perception and manipulation, thus largely remain and are unlikely to be fully resolved in the next decade or two.” (Page 25.) This makes me vaguely unhappy: the big challenges are what happens after that, at the limit. After all, my boy will be my age in 2055 ... and if I live as long as my father did, I will also still be around.

If terms of data ... ugh. They used BLS data to group 702 jobs in terms of the three labor inputs. They then validated that model by using the BLS data to ... predict their subjective evaluations.

The results is that restaurant cooks and models wind up in the same easily-computerizable category as radio operators and cashiers. Down on the hard-to-computerize side, we have logisticians ranked with the guys who work in boiler rooms. (The latter is probably not what you think.) Both are hard to computerize ... but the latter is much harder. (I should add that I have worked in both fields, albeit very briefly and in one courtesy of the Army.)

Is there a non-intuitive way to test the plausibility of the rankings? Carlos Yu and I think we might have one. The less-computerizable jobs should also be the jobs where employees have more leverage over employers. Technically, you could use the the negative log of computerizability as a measure of bargaining power regarding future earnings. In Carlos’ words: “That works fine for data entry and telemarketers, who have -log(.99) ~ zero bargaining power, but given the weight of teaching, counseling, and nursing professions on the low-computerizability end of the scale, it suggests other factors are in play.”

In short, an intriguing paper, with what looks like the right methodology ... but not one that gives a useful roadmap to 2035, let along 2055 or 2100.

The robots are here

I have been beating the robot drum since around 2006, as Will Baird and Carlos Yu will attest. In fact, I have been worrying about it for a long time, ever since I asked myself, riding a number 6 train back in the 1980s, “Why do we need people running this thing at all?” A decade later, in the 1990s, I had an impassioned discussion with my (Texan) then-girlfriend about the terrible effect automated taxis would have on the Mexican middle class. I stopped reading most science fiction (Charlie Stross excepted, and I am wondering if this Blue Remembered Earth is worth it) because too much either bounced between “Singularity! Wild wow whoo-hoo! The humans are dead!” and completely ignoring the effect of automation. There was this science-fiction noir series set in San Francisco that dealt with some of the issues (I cannot remember the name, dammit!) but even it had a scene with a human truck driver that just did not make any sense.

All this is a lead-in to getting you to click this link. It is a conversation between two Navy pilots upon watching an unmanned X-47B perform touch-and-go landings on an aircraft carrier. It is awesome. Read!

The end of progress? Probably not.

Robert Gordon just published a paper that is very pessimistic about long-term economic growth. It essentially argues that the big wave of innovation from IT is over. Infotech may radically reorder our lives in the coming decades, but it won’t allow us to make more stuff. Moreover, the big important inventions have been made. His example: would you rather give up all the electronic and IT stuff invented after 1975, or keep all that but have to abandon indoor plumbing?

As a metaphor for the state of play, that nails it. If it doesn’t for you, that is because you have never had to live without indoor plumbing. At night. In the rain.

There are three ripostes to this thesis. The first is that Moore’s Law ain’t over yet. (More to the point, Koomey’s Law ain’t over yet: it states that the energy efficiency of computers doubles every 18 months.) If Moore and Koomey hold for another decade or two, then there is the possibility that AI will emerge seemingly out of nowhere, sometime in the 2030s. Economic growth will surge, but prepare yourself for the robot riots. (I will be joining the rioters, and training my young son to lead them.)

There are reasons to think this will not happen, possibly ever. (I can hope!)

The second riposte, then, is that even without singularian magic, better computers will improve productivity the old-fashioned way. Consider construction robots. My co-blogger and I went back-and-forth on this with our friends Will Baird and Carlos Yu. We concluded that construction is a super-conservative fragmented industry, so there will be no revolution (or robot riots) ... but there will be a slow trickle out of construction (and vehicle operation) just as there was out of agricultural labor, elevator operation, and telegraph delivery. That could keep growth going for a while yet. (Matt Yglesias points out some other sectors ripe for growth here.)

But there is an even more convincing riposte than the above two. It comes from Carlos Yu, and it is based on work by Alexander Field, an economic historian who has pointed out that the Depression-wracked 1930s were likely the most technologically-productive decade of the past century.

What then is the third riposte?

SHORT VERSION: Stop focusing on the damn computers!

LONGER VERSION: “After reading Alexander Field’s book, I distrust the simple correlation in timing between technological improvements and productivity growth. In any case, Gordon is over-focused on the immediate implications of information technology, and he’s missed the one under his nose: biotechnology.

“The Green Revolution doesn’t show up much in American productivity statistics, since our crops had already been improved through a hundred-plus years of scientific selective breeding, but it does show up in global productivity. (Though biotechnology is probably keeping our crop yields high in the biological arms race against disease and pests, but that’s hard to quantify with typical productivity measurements.)

“Cheap effective medicines is another. For many health issues, we can actually buy a few more years of life — or more than a few — at prices far below the break-even point. If we take his question about toilets versus computers and switch it to toilets versus postwar improvements in the applied biological sciences, I think we’d be carrying our chamberpots to the outhouse.

“And now we’re entering the second wave of improvements in biotechnology, ones based on the information technology revolution. It might be a dud. I strongly suspect not.”

What is the weather service like in Tanzania?

 That is something that I have always wondered. And lo, Doug Muir answers:

Much better than you’d think. Not clear why. But it’s providing good, real-time information for both weather predictions and best-guess seasonal stuff. 

Tanzania has 95% mobile coverage. So almost every community — every dusty little plains village, every wandering band of Masai — now has at least one guy with a cell phone who can get regular text updates on when the rains are coming and how long they’re likely to be.

Satellite data is cheap or free, basic forecasting techniques are known and easily taught. So why don’t other African countries have good weather services too? I’m not sure, but I think this is another positive-legacy-of-socialism thing: lots of little local weather stations, plus a good tradition of a strong government weather service. Apparently the weather guys are something that’s all too rare in Africa: an educated elite who aren’t afraid to spend time out in the villages or get their hands dirty. 

Anyway, good weather service = huge, huge difference in quality of life. If nothing else, it dramatically reduces the uncertainty.

Cool stuff

  Crazy busy these days.  Will try to post more, promise!

In the meantime, two things.  First, I watched the first episode of Flashforward on ABC yesterday.  Not bad!  Despite being a little overly “cinematic,” my wife liked it too.  To her surprise.  The best thing?  They had “Quiet Dog” by Mos Def playing when everyone blacked out.  Awesome.

Second, this is old, but kinda cool.  I think.  Libby Dole and Lebed as the presidents?  Huh.

Department of WTF, once more

 These are not like things.  The data do not say what the people making the connection seem to think that they say.