I have twinned reading on drones, the invasion of privacy and the commercial, non-military world. First consider Brianna Ehley’s piece in the Fiscal Times, What’s Grounding the Commercial Drone Industry. Ehley writes about the potential markets for civilian drone use in the U.S. While a properly impressive large dollar figure has not yet been associated with the market size, suffice it to say that Ehley correctly recognizes the potential space of drone applications as vast, with thousands of drones performing aerial fertilization, pesticide-spraying, delivery and numerous other tasks. In the future of drones, 2015 figures as an important date we all need to remember, as this is when the FAA must, by law, start to have sufficient regulation in place for this new form of Robot Smog to take off. So, breathe easy for one year? The second reading is by Scott Bomboy: Forget About Drones in Constitution Daily. Bomboy argues that commercial robots- lethal robots for nonmilitary use- are coming, and cites the oft-sited Samsung security guard robot that can be armed with a machine gun and grenade launcher. The connection between the two articles is instructive because it’s a recognition of the legal speed bumps facing both types of deployment: Bomboy notes that the concept of accountability is ill-defined for a system as autonomous as a gun-wielding security robot, making judgements about responsibility quite difficult; Ehley notes that the insurance industry has not constructed the financial models to offer insurance policies that will be required before civilian drone use can take off (sorry for that). I would add that accountability and insurance will be even further muddied by the fact that many of these robots will be mixed-initiative, with partial human operator control and part-time autonomous operation. When bad things happen during transitional periods, prepare for epic battles over accountability and responsibility.
Yesterday’s Washington Post hosts an article by Abbot Koloff: “It’s increasingly hard to hide from the watchful eyes of cameras” which was posted originally in the Leader Telegram last month. The article is a good reminder that infrastructure-based recordings are becoming ubiquitous, from streets and parking lots to dormitory rooms and hallways; and that such image archives can prove very effective in identifying criminal behavior- particularly when you know when and where to look for it in the data record. As infrastructure sensing of human behavior expands, the interesting part of this massive Big Data problem is that making sense of all the data becomes a superhuman endeavour, and a natural direction for innovation will be to apply automatic machine-based interpretation of human behavior to this dataset that far exceeds even a human army’s ability to review. “Show me all pairs of 30-year old men who photographed that bridge” ; “Show me every person who handed another person a package in Central Park that week.” Of course the boundary between identifying criminal behavior and summarizing high-value information is slippery. How about premonition for new clothes trends in New York, storefront gesture analyses, studies of visitorship to competing restaurants. The infrastructure will build will be all-seeing, and we know people will not change their overall behavior profoundly, even if they are intellectually aware of this level of observation.. Look at email: we repeatedly tell everyone that email is essentially public, and plenty of users still manage to embarrass themselves with incredibly poor email decisions. But that very same sensing infrastructure can move from passive recording to meaning-making as AI proceeds, and this means a treasure trove of semantic information will be available to those with either power, money or subpoenas. Every camera on every street corner will start behaving like a very sharp gremlin is in there taking notes- and these gremlins will be happy to strike a bargain.
The Washington Post’s Steven Levingston just published a nice review of Robot Futures; it is an interesting read that does not focus much on the dystopian side of what I evaluate. In fact Steven starts off by saying he hopes to be reincarnated as a robot!
NPR’s Steve Henn broadcast this story on NPR’s Morning Edition about Google I/O and the upcoming roadmap of product intelligence. The unifying theme across the report involves the greater intelligence of Google’s products in providing you with the right information: by taking into account past interaction with you, social conventions, image analysis, your location – you name it- the tools move from search-based information providers to highly customized, responsive agents that give you what you need when you need it. In New Mediocracy I write about the concept of the perfect marketing tool: if you have six billion people, the perfect tool has six billion custom versions. The results of AI, applied so very personally, can be wonderful in bringing us information ever more efficiently. Indeed, the wealthy have personal assistants that do just this. There is one interesting difference however: Bill Gates’ personal assistants work for Bill Gates- they’re paid by him. In this case, there will be a trade; you get far better information, but of course you will also receive far more customized advertisement. And your personal assistant will be part of a global network, able to mine across demographics to draw conclusions that are profitable for business down the road. I am sure that one day we will look back and wonder how we got along with out the customized interfaces we will come to expect. But I am also very curious, in that near future, what expectations will we have about the privacy of our personal behavior?
Kevin Drum writes for Mother Jones’ June 2013 issue: “Welcome, Robot Overlords. Please don’t kill us.” Drum’s article is worth reading, although it is important to note, as with all singularity trajectories laid out in the popular press, the hidden assumption that massively increasing computing speed will lead to robots with ‘true artificial intelligence.’ But, as Drum rightly points out, these future robots needn’t think the way humans do. Whether they are intelligent by our standards or not, machines will become ever more capable at all the prosaic and skilled activities that we partake, from driving to entire classes of working class job positions. Drum brings up the possibility of a further transition from labor to capital-based income generation. Those with money to own the machines win; the rest see their uniqueness being chipped away.
Drum also paints a picture in which these economic dynamics speed up exponentially- as machines double in speed, so the power of capital and the denaturing of labor ramp up. If that model is right, then it is almost guaranteed that when society does respond to obvious change, it’s already too late. Perhaps some outstanding theatrical types can create visions of our future on stage and on camera, so we can jump in a time machine, visit the near future personally, then come back and act in the know.
Your parents told you never to read the bottom half of a web page, but I have disregarded that by checking out all the comments on Tech Review article that I published two days ago. It is not surprising to see a broad diversity of opinion about whether robots are fundamentally a new ingredient in the dynamics of business, or just the same as power nail guns and welding arms. In the arguments made consistently, there are some tropes worth pointing out though, and many opinions center around the same basic flow of reasoning: (1) automation happened before and it was just fine; (2) the advent of new robotics is just more of the same, and so it will be just fine too; (3) no matter what happens, consumer demand and market dynamics is corrective and will ensure that things go well.
But the real world is messy, and these universals are not portraying that reality. Was the advent of automation fine each time it hit? If you pose the long-term question to an economist, sure. If you ask the daughter of a steelworker or auto worker, no; it was a disaster that no amount of retraining succeeded in compensating. Impacts can be generational, and the concept of it all working out can take multiple generations even when trends point the right way. As for the equivalence of robotics and prior forms of automation, the disagreement I have with some Singularity proponents is more on ‘when’ than ‘if’ for concrete, social tasks that robots can eventually undertake. Sure, robots will always lie on a continuum of intelligence, with nail guns near the bottom (I hope). But the sensitivity that underemployment might have to where technology lies on that continuum is not established. We are turning dials that do not have well-defined labels, and that is my principal concern on this front. Finally, there is argument (3)- sometimes stated in terms of demand: without a middle class, there is no demand for goods, and therefore the system of producing via robots will self-correct to ensure enough people consume enough material goods. While corrective forces may influence aggregate demand, it is important to note that such corrections do not necessarily influence social and economic inequity; the gap between classes can increase even with gross corrective forces in place.
Let the debate rage on- it is a healthy step toward building the awareness that has to predate any deliberate action.
MIT Technology Review just published an opinion piece by me entitled We Need to Talk about the Burgeoning Robot Middle Class. The article talks about some of the reasons why I believe robots will authentically displace middle class jobs over the next few decades.
In June last year I wrote about Google Now, which gives you answers before you have formulated your questions. Tom Simonite writes in MIT Tech Review today about several companies providing such predictive services, from directions and check-in to calling a cab for you automatically when you arrive at the airport. And, yes, sending messages informing your friends that you’re late automatically. This article is a good summary of activity in the early space of CEO of Me, Inc.
Amy Chozick writes in the New York Times about the Bloomberg News admission that they snooped Bloomberg terminal usage statistics to see how specific individuals were behaving with respect to the Bloomberg data they were subscribed to receive. Reporters at the news division were given access to behavior data on specific users, and gave away their edge when they asked a Goldman Sachs manager if a particular employee was still at the company since they hadn’t logged into the Bloomberg terminal in a month. This is a good what-if robotics question because the reporters were taking on an ethical misstep in their desire to track specific human behavior so they could write stories that would sell. Now take the New Mediocracy idea in Robot Futures and pursue the what-if exercise: what if a company had access to such human behavior-tracking, but the users were automated market intelligence soft-bots rather than news reporters? News reporters interview folks and eventually trip up, giving away their game. Soft-bots don’t. They mine human data, make decisions about advertising and marketing, and the chance of anyone figuring out just how they are being tracked, whether legally or illegally, is essentially zero. Can such tracking take place? Certainly! Existence proof: Bloomberg. Do robots do it? I don’t know. Neither do you.
The final chapter of Robot Futures suggests that robotic technologies at their very best can help us be more human: they can forge deeper empowerment into our communities, enabling us to make smarter choices for the future. Just this morning TIME released their TIMELapse story, which shows change across the earth, natural and man-made, over the past thirty years at a level of detail that was unthinkable before. TIME has done a fantastic job in researching the stories that these moving, explorable images back, then telling each story in enough detail to give the reader an eye for interactively exploring the data set themselves. The image content comes from NASA and USGS, and the heavy computational lifting is possible because of Google. The CREATE Lab had a role in creating the interface for zooming and exploring massive the massive data in time and space. Some related links that I’m happy to share:
Our GigaPan Time Machine demonstration site