Link to the University of Waterloo home page

Link to the Centre for Society, Technology and Values home page

jump to navigation

Is AI bad for us? February 8, 2016

Posted by Cameron Shelley in : STV302 , add a comment

Several recent news items take up the ongoing controversy as to whether or not the benefits of Artificial Intelligence (AI) software are worth the risks.

For example, an NPR article discusses how the “Google generation” has become dependent on search engines to locate and process knowledge. Kentucky English teacher Terry Heick relates his chagrin when his students, faced with the question, “How does a novel represent humanity?” responded not by considering it but by Googling it:

Heick had intended for his students to take a moment to think, figure out what type of information they needed, how to evaluate the data and how to reconcile conflicting viewpoints. He did not intend for them to immediately Google the question, word by word — eliminating the process of critical thinking.

Search engine AI tries harder and harder to provide information that is relevant to users. There is the risk that users will respond by not trying very hard to think for themselves. This result may be due to pure laziness or to the fact that other activities are competing for their time and attention, like whatever their friends are doing on Facebook.

Heick recommends that teachers make assignment questions “Google-proof”, that is, questions that cannot be answered through a few simple web searches.

A FastCompany article takes up a related theme, encapsulated by the term Fear Of Deciding Alone (FODA). The claim is that AI decision-support software, e.g., recommendation engines, make people anxious about making decisions in the absence of such “expert” advice. The danger appears to be that people who experience FODA will relieve it by uncritical acceptance of whatever these software systems put forward. Software may be objective in the sense that it has no feelings about its recommendations one way or the other. However, it is a mistake to conclude that such software is unbiased:

Behind every computer algorithm is a programmer. And behind that programmer is a strategy set by people with business and political motives. It would be easy enough for the people who design AI systems, motivated by greed, self-interest, or politics, to train computers to manipulate our lives in subtle and insidious ways, essentially lying to us through the algorithms that guide our thinking.

In other words, software may be unfeeling, but it still embodies choices about what factors to consider, how they are measured, and how they should be weighed. In fact, this point is true of any software that is used for decision support, whether it involves AI or not.

Lastly, a new website called Watson2016 has appeared, advocating for IBM’s AI system known as Watson to become the next President of the United States. Yes, that is the same Watson that kicked human ass in a recent Jeopardy game show.

The website, apparently set up by the Electronic Frontier Foundation, advocates Watson because of its unique, data-crunching abilities:

It is our belief that Watson’s unique capabilities to assess information and make informed and transparent decisions define it as an ideal candidate for the job responsibilities required by the president.

According to the website, Watson has arrived at a number of left-of-center answers to policy questions, such as single-payer national healthcare, and free tuition for public universities.

I, for one, welcome America’s new robot chief executive. Certainly, I hope to hear it take on Donald Trump in a debate in the upcoming U.S. elections.

Watson gives a press conference in the White House


DNN: 5 Feb 2016 February 5, 2016

Posted by Cameron Shelley in : STV202 , add a comment

The capabilities and limitations of drones continue to be explored.

For example, New Scientist notes that the British National Health Service is preparing to use drones to help first responders. Fifteen Hazardous Area Response Teams are to receive drones to use for reconnaissance when called upon to deal with hazardous situations and materials. While on their way, advance drones can give operators an idea of what lies in store for the teams.

Indian doctors with the Multi-Organ Harvesting Aid Network (MOHAN) are planning to use drones to transfer organs for transplantation. Congestion in Indian cities can make it difficult to move sensitive organs in time. A heart, for example, must be transplanted within four to six hours of its removal for a decent chance of success. To date, police have been used to try to move organs through cities speedily. Four Spanish design students thought that specially-equipped drones would work better and founded DronLife. Testing in Delhi will begin later this year.

Slate relates a project headed by Eelke Folmer at the University of Nevada, Reno, who adapted a drone to help lead blind runners around a track. The approach is simple enough: Runners simply follow the buzzing sound made the drones’ motors. Prof. Folmer says that the trickiest part is getting the legal permission to proceed.

Speaking of legal limitations, the FAA has announced a strict no-drone zone around Levi’s Stadium during this weekend’s Super Bowl event. If they find anyone’s drone within 32 miles (52km) of the stadium, they will get medieval on that person’s ass. Violators risk fines of $1,100 for persons and $27,500 for businesses.

Interestingly, the FAA’s rules state that “deadly force” may be used in the event that aircraft over forbidden zones are determined to be a threat. How NORAD might undertake to use deadly force against a consumer drone is unclear. However, Dutch police may have an answer. Gizmag reports that they have trained birds of prey to take out consumer drones on command. Raptors have been seen attacking drones in the wild before, so perhaps this observation was inspiration for the project.

I could even see how, in future games, a bald eagle taking out a drone could be part of the half-time show.

Uber’s new icon is sphincterrific! February 3, 2016

Posted by Cameron Shelley in : STV202 , add a comment

Uber has unveiled a new pair of icons, along with other measures aimed at public rebranding. Wired magazine gives a detailed report of the effort, which was undertaken by a dedicated design team at the company, which found that external consultants were not satisfactory.

Here is the previous logo:


(Kobolen/Wikimedia commons)

And, here is the new one:

Uber logo

(Justin Ormont/Wikimedia commons)

More notably, Uber also has new icons for its apps (left icon for users, right icon for drivers):

New Uber icons


As Dezeen notes, the new icons have not met with universal applause, with some wags commenting that the user icon resembles an anus.

Is the new icon a #fail? Clearly, resembling an unfortunate orafice would not be a good move. However, netizens do love to dump on big companies and their doings.

Goodbye CFL! February 2, 2016

Posted by Cameron Shelley in : STV202 , add a comment

No, the Canadian Football League is continuing operations. In this case, CFL stands for Compact Fluorescent Lamp. General Electric has recently announced plans to cease production of its CFL goods by the end of 2016. If GE is phasing them out, then other manufacturers may not be far behind.

The spiral-shaped CFL was invented in 1976 in order to provide a cheaper source of household illumination, as a response to the OPEC oil crisis. Its use has expanded in the last decade as governments mandated higher energy efficiency in lighting.

However, it seems the the swirly CFL has been overtaken by a new technology, LED lighting. LED bulbs are yet more efficient than CFLs, last longer, do not contain mercury, and have become competitive on price. GE expects sales of LED bulbs to surpass sales of CFL bulbs in 2016:

Industry wide, the breakdown for light bulb sales in 2015 was incandescents at 38%, halogen at 26%, CFLs at 15% percent, LEDs at 15%, and linear fluorescent at 6%. Last year, LED sales grew by more than 250% and the growth trend is predicted to continue. Internally, GE expects that during the first half of 2016, CFLs will outpace LED bulb sales and by the end of the year it will be close to 50/50 with LEDs. The company expects that there will still be some of its CFLs available in 2017 as the phase-out takes its course.

So, the CFL has become obsolete.

It is interesting to note that sales of incandescent bulbs continue to be healthy, totaling more than CFL and LED bulbs combined. Although the technology is older, it seems that the lighting is still popular. Perhaps this is because of its familiarity and warm colour spectrum.

Also, the article notes that GE has other reasons to favour LED bulbs. LED bulbs are compatible with GE’s move into “smart” technology:

“LEDs give us the real estate in the product to put in Bluetooth and Wi-Fi,” Patell [Daraius Patell, North America Consumer Lighting Leader at GE] says. “It’s really the vehicle to go connected.”

GE’s interest in LED bulbs has much to do with its plans to sell consumers a lineup of networked, smart objects for their households.

So, the obsolescence of CFLs has less to do with their intrinsic qualities and more to do with their compatibility with other products.

06 Spiral CFL Bulb 2010-03-08 (white back).jpg

06 Spiral CFL Bulb 2010-03-08 (white back)” by Sun LadderOwn work. Licensed under CC BY-SA 3.0 via Wikimedia Commons.

When is a Mustang a Mustang? January 29, 2016

Posted by Cameron Shelley in : STV202 , comments closed

One concept that often arises in discussions of good design is the essence of a design. Modernist designers, like Dieter Rams, often talk about the essence of a design, that is, those features that are essential to some category of functional object. For them, good design means reducing a given object to only those functions that are essential to it.

The main problem with this approach is knowing what is essential. For example, what functions are essential to a toaster? The ability to scorch bread would certainly qualify, but what about a pop-up mechanism? Such mechanisms have become nearly universal and increase safety of and even interest in toasting, yet they could be (and have been) viewed as extraneous. And then there are features that allow you to scorch a selfie onto your toast!

This issue came to mind when I read an article in Gizmag about a new model of the venerable Ford Mustang, the quintessential pony car. The typical Mustang—which recently celebrated its 50th anniversary—has always been a high-powered V8, four-seater muscle car. Yet, Ford recently introduced a four-cylinder, turbo-charged, EcoBoost version. Is such a car still a Mustang?

The author of the review puts the matter succinctly:

The Mustang is a motoring icon, but even icons need to change and evolve with the times. For Ford, that meant turning to EcoBoost power and (finally) dropping the live rear end for a fully independent setup. But has it worked, and is it sacrilege to stick a Mustang badge on a turbocharged four-cylinder car?

Good design usually allows for innovation. However, changing a design fundamentally could mean abandoning the original type, perhaps as obsolete.

In the end, the reviewer seemed satisfied with the result. The new model looks, sounds, and performs much like a typical Mustang, close enough that the issue of whether or not it counts as a Mustang at all does not recur. Even the fact that the rear seats are not truly big enough to fit people with legs and heads seems reassuring somehow.

This consideration does not, of course, solve the problem of knowing what the essence of a design is. However, it underlines some of the usual qualities that people look for: It must be true to form and it must be sufficiently different from other designs to prevent confusion.

2015 Ford Mustang.jpg

2015 Ford Mustang” by A. BaileyOwn work. Licensed under CC BY-SA 3.0 via Wikimedia Commons.

Uber rides high January 27, 2016

Posted by Cameron Shelley in : STV 201, STV202, STV302 , comments closed

Ride hailing services, epitomized by Uber, have shaken up the ride-for-hire industry, and more!

In Europe, Uber’s entry into the French market has sparked a violent reaction. French taxi drivers and other protestors have set up roadblocks, set tires on fire, and mounted a drive-slow campaign. Their beef is that it is unfair for the government to require them to buy expensive taxi licenses while allowing Uber partners to work without them.

Uber has struggled in Germany. Their difficulties stem not from taxi protests exactly but from strict German regulation. Not only do taxi drivers there require special licenses but also stringent security checks and health exams. In the meantime, native ride-hailing services such as MyTaxi have thrived because they cooperated with regulators:

“In every country, you have to be prepared to change your setup,” said Mr. Mewes, whose start-up now operates across several European countries after being acquired by Daimler, the German automaker, in 2014. MyTaxi currently has 45,000 drivers using its online platform, with roughly half of them based in Germany. “That’s why Uber failed here. They aren’t willing to change when they enter a new country.”

Native services have now occupied much of the market, meaning that Uber may have a hard time making up for lost ground in Germany.

Stateside, the city of Seattle has voted to allow Uber drivers to unionize. The local App-Based Drivers Association was lobbying for the move with local teamsters. The measure is not popular with the ride-hailing services, of course, but may also contravene federal labour laws. If Uber drivers are independent contractors, then collective bargaining on their part could be considered a form of collusion under anti-trust laws. In the end, it is not clear that unionizing will be very popular among Uber’s partners anyway.

Of course, this measure raises the issue of whether or not Uber partners are contractors or employees. Some labor boards have ruled that they are. However, several US states are passing laws to support Uber. Ohio, Florida, North Carolina, Arkansas, and Indiana are moving ahead with legislation that designates drivers in network-based services as contractors. Representatives of Uber and Lyft were evidently involved in the drafting of those bills. Such legislation would tend to limit the damage to Uber’s business model if it loses a class-action lawsuit in California brought by drivers wanting to be designated as employees in that state.

Whatever its legal standing, Uber’s relationship with its partners continues to develop. The Guardian reports that Uber is conducting a trial in Texas wherein it uses data from drivers’ smartphone to track their driving behaviour. A smartphone’s GPS locator and gyroscope can deliver a lot of information about when and where a driver accelerates dangerously, for example. Uber says that it wants the data to help protect passengers from dangerous driving by its partners.

Uber also takes measures to protect its drivers from dangerous passengers. In North Carolina, for example, Uber has experimented with leaving a children’s toy called BopIt in the back seats of cars. The idea is that drunk passengers find playing with the toy so engrossing that they are kept from distracting or abusing their drivers.

Interactions with vomitose passengers is apparently one of the biggest problems for Uber partners. One driver in New York left a woman in labour without a ride to the hospital because of fear that she would puke on the upholstery. The driver even charged her $13 for his time! Uber apologized to the woman, noting that the practice is against company policy. It is also illegal in the State of New York to refuse a pregnant woman a ride to hospital.

The article goes on to note that Uber drivers have great incentives not to take vomit-likely passengers. As contractors, the time and expense for cleanup comes out of their own pockets. Uber can charge passengers who damage vehicles up to $200, which it passes on to drivers. However, drivers find it difficult to claim this money in practice:

Rideshare Dashboard, a website that offers resources and advice for Lyft and Uber drivers, advises drivers to take pictures of the car’s interior immediately after a passenger gets sick. “Only if there is sufficient vomit will they pay you the full amount,” the site says. “If its just some splatter or some random chunks, they will only pay you half or less.”

Giving money to Uber is easy. Getting money from Uber, not so much.

So, the gigging drive-for-hire business can be messy. However, this fact has not deterred Ford. The carmaker has recently announced that it plans to introduce a service to compete with Uber and the like. It seems that the new technology may profoundly shake up not only the legal landscape but the car industry itself, turning carmakers into ride providers.

Marvin Minsky dies, 24 Jan. 2016 January 26, 2016

Posted by Cameron Shelley in : Events , comments closed

The New York Times has published a short article on the passing of Marvin Minsky, one of the “founding fathers” of Artificial Intelligence, in Boston at the age of 88.

Minsky was noted for many things, including his work in robotics, computer vision, and neural networks. His main interest was always intelligence. In 1985, he published Society of Mind, a theory accounting for how intelligence could arise in humans (or other machines) from a network of agents, none of which were intelligent on their own. As a grad student entering AI at the time, I can recall the interest that this work sparked in the field.

Minsky had an interesting perspective on intelligence. In addition to the conventional view that intelligence was something that a system had to positively possess, he took the view that intelligence was also centrally an issue of not being stupid. For example, he was very interested in the Frame Problem, that is, how intelligent agents deal effectively with all the logical but irrelevant implications of their beliefs and actions.

I remember encountering Minsky at a computational humor conference in the Netherlands in which he espoused the view that humour is a way that people have of dealing with such issues. Much humour relies on conceptual incongruities, which are occasions in which our conceptual repertoire displays breakdowns of the sort that reveal the challenge presented by the Frame Problem. We had an interesting conversation about it although, as a young graduate student, I was too scared of him to pursue the matter at length. My loss.

Anyway, Marvin Minsky will long be remembered as a seminal contributor to Artificial Intelligence, a field that continues to grow in scope and relevance.

Marvin Minsky at OLPCb.jpg

Marvin Minsky at OLPCb” by The original uploader was Sethwoodworth at English Wikipedia, taken by Bcjordan – Transferred from en.wikipedia to Commons by Mardetanha using CommonsHelper.. Licensed under CC BY 3.0 via Wikimedia Commons.

Self-driving cars are coming! January 25, 2016

Posted by Cameron Shelley in : STV 201 , comments closed

A recent piece in the New York Times outlines some of the pitfalls (potholes?) in the road to the self-driving car. Some of these are technical. There are still many conditions where current self-drivingish cars fail to work properly, as in the recent “offramp” bug with Tesla’s Model S, in which the car would try to follow every offramp from a highway that it was driving along.

Of particular interest to Canadians may be the fact that current systems do not work well in heavy rain or snow.

Other issues have to do with each car’s relationship with its driver, who may well decide to check Facebook or take a nap while self-driving mode is on.

In a related issue, many laws require there to be a driver in every car in motion and that the driver maintain control of the car at all times. Clearly, this condition cannot apply to self-driving cars. In response, President Obama recently launched a $4 billion program to smooth the path of self-driving cars. Part of the program is to develop model laws that states could adopt that would make self-driving cars legal.

Then there is the issue, discussed earlier, of how self-driving cars will affect cities, where many of them will live. The matter has also been addressed by Clive Thompson, who is optimistic that self-driving cars will reduce traffic and the need for parking space. Let’s hope the optimism is warranted: As Thompson points out, cities have many hectares of excess parking space that could be put to better use, such as infill.

Some of the projections that he cites having worrying elements. For example, researchers are creating simulations of cities populated by self-driving cars in which traffic flows much more efficiently than it does now:

Robot cars could also drive much more closely to one another, packing far more vehicles onto a street. (Computer scientist Peter Stone even created software that would let robot cars do away with traffic lights; instead of stopping at an intersection, they would simply weave around one another, navigating street corners nearly 10 times faster than cars do today.)

What about the cyclists, pedestrians or busses? Road design is already biased in favour of maximizing car traffic. Would a big push to introduce self-driving cars thrust these other road users further to the margins?

A lot will depend upon our attitudes and policies. Thompson points to the increasing trend away from individual car ownership and towards more urban and walkable lifestyles. This, plus policies such as so-called congestion charges in big cities, should facilitate introduction of self-driving cars in a way that is positive and equitable.

Smart clothes on the way January 22, 2016

Posted by Cameron Shelley in : STV302 , comments closed

The HVAC systems in big buildings never seem to work right. In summer, office workers wear sweaters to ward off arctic blasts from the overhead A/C vent. In winter, they sit in the lee of USB fans trying to resist melting. There are various reasons why people experience thermal discomfort, especially in such places. Money is saved by installing only few thermostats, meaning that the climate is good only in their vicinities. Air circulation is cut off of by cubicle walls, etc.

Research funded by the US Department of Energy promises to change that. A team of researchers at the University of California is mimicking the coloration-changing biology of squids to come up with clothing designs that thermoregulate themselves. The method is secret. Another group at Cornell is designing undergarments with a system of microtubes that helps to transfer heat where it is needed and way from where it is not. Another group at SRI is studying ways for clothing to remove heat from hot spots like the soles of the feet.

Besides thermal comfort, such systems may deliver energy savings. After all, making people comfortable mostly means regulating the temperature right around their bodies, a job that is crudely managed through regulation of whole spaces. Could smart clothing keep people warm in winter, for example, while lowering the temperature of their interior surroundings? If so, then heating in winter (and A/C in summer) could require less energy to operate.

Of course, no object is truly smart until it is connected to the Internet. As soon as smart clothing appears, people will want to control it through an app. What is the point of smart undergarments unless you can set them to warm up in the morning before you put them on, like your smart coffee maker does today?

Putting things on the Internet raises security issues. Hackers have already taken control of building heating systems remotely, for example. Plus, people seem increasingly obsessed with tracking every statistic about themselves, which would extend to the environment in their smart clothes as well, I assume. So, the day is coming when Vladimir Putin will be able to toast your gonads by hacking into your pants! Better change the password on your underwear as soon as you get it home.

DNN: 21 Jan 2016 January 21, 2016

Posted by Cameron Shelley in : STV202 , comments closed

In a development that may interest the UWaterloo community, a pest-control company named Bird-X has developed what they call the ProHawk drone. According to Gizmag, the drone is designed to resemble a bird of prey and to emit scary bird calls as it flies around its designated patch:

Fitted with GPS, the ProHawk UAV can be programmed to follow a specific flight path along waypoints and features a built-in sonic sound unit. This device stores a range of predator calls, prey bird distress cries and Canada goose cries, which are emitted en route to clear the area of unwanted pests.

The unit can also be flown manually if desired.

Canada geese have long been an ornament (or ordeal) on the Waterloo University campus. The University even has a special Goose Watch service to track nests and aggressive fowls. Perhaps the University could obtain special permission to fly the drone around campus in order to keep these geese at bay.

If so, then it should consider insurance. As noted by the Globe and Mail, insurance companies are beginning to adjust policies to cover consumer drones. Previously, drones were considered light aircraft, a category that is manifestly inappropriate for such lightweight devices. So, policies are now available to cover misadventures for little drones such as colliding with things or crashing into people.

Even so, some activities remain uninsurable:

Mr. Lessard said most operators essentially need the same kind of insurance against physical damage to people or property. Limitations to coverage include using the drone to “take pictures of someone through the window of a hotel or something like that, and that person is being sued,” Mr. Lessard said. “We’re not covering these kind of things.”

It remains unclear how collisions with angry geese might be considered.

In the meantime, enjoy the sight—and sound—of the ProHawk.

  • Regulus by Ben @ Binary Moon
  • Created with WordPress
  • Top
  • -->