Paper by Benjamin Alarie, Anthony Niblett and Albert Yoon: “The set of tasks and activities in which humans are strictly superior to computers is becoming vanishingly small. Machines today are not only performing mechanical or manual tasks once performed by humans, they are also performing thinking tasks, where it was long believed that human judgment was indispensable. From self-driving cars to self-flying planes; and from robots performing surgery on a pig to artificially intelligent personal assistants, so much of what was once unimaginable is now reality. But this is just the beginning of the big data and artificial intelligence revolution. Technology continues to improve at an exponential rate. How will the big data and artificial intelligence revolutions affect law? We hypothesize that the growth of big data, artificial intelligence, and machine learning will have important effects that will fundamentally change the way law is made, learned, followed, and practiced. It will have an impact on all facets of the law, from the production of micro-directives to the way citizens learn of their legal obligations. These changes will present significant challenges to human lawmakers, judges, and lawyers. While we do not attempt to address all these challenges, we offer a short and positive preview of the future of law: a world of self-driving law, of legal singularity, and of the democratization of the law…(More)”
A cautionary tale about humans creating biased AI models
Matt Bencke at TechCrunch: “Most artificial intelligence models are built and trained by humans, and therefore have the potential to learn, perpetuate and massively scale the human trainers’ biases. This is the word of warning put forth in two illuminating articles published earlier this year by Jack Clark at Bloomberg and Kate Crawford at The New York Times.
Tl;dr: The AI field lacks diversity — even more spectacularly than most of our software industry. When an AI practitioner builds a data set on which to train his or her algorithm, it is likely that the data set will only represent one worldview: the practitioner’s. The resulting AImodel demonstrates a non-diverse “intelligence” at best, and a biased or even offensive one at worst….
So what happens when you don’t consider carefully who is annotating the data? What happens when you don’t account for the differing preferences, tendencies and biases among varying humans? We ran a fun experiment to find out….Actually, we didn’t set out to run an experiment. We just wanted to create something fun that we thought our awesome tasking community would enjoy. The idea? Give people the chance to rate puppies’ cuteness in their spare time…There was a clear gender gap — a very consistent pattern of women rating the puppies as cuter than the men did. The gap between women’s and men’s ratings was more narrow for the “less-cute” (ouch!) dogs, and wider for the cuter ones. Fascinating.
I won’t even try to unpack the societal implications of these findings, but the lesson here is this: If you’re training an artificial intelligence model — especially one that you want to be able to perform subjective tasks — there are three areas in which you must evaluate and consider demographics and diversity:
- yourself
- your data
- your annotators
This was a simple example: binary gender differences explaining one subjective numeric measure of an image. Yet it was unexpected and significant. As our industry deploys incredibly complex models that are pushing to the limit chip sets, algorithms and scientists, we risk reinforcing subtle biases, powerfully and at a previously unimaginable scale. Even more pernicious, many AIs reinforce their own learning, so we need to carefully consider “supervised” (aka human) re-training over time.
Artificial intelligence promises to change all of our lives — and it already subtly guides the way we shop, date, navigate, invest and more. But to make sure that it does so for the better, all of us practitioners need to go out of our way to be inclusive. We need to remain keenly aware of what makes us all, well… human. Especially the subtle, hidden stuff….(More)”
Encouraging and Sustaining Innovation in Government: Technology and Innovation in the Next Administration
New report by Beth Simone Noveck and Stefaan Verhulst: “…With rates of trust in government at an all-time low, technology and innovation will be essential to achieve the next administration’s goals and to deliver services more effectively and efficiently. The next administration must prioritize using technology to improve governing and must develop plans to do so in the transition… This paper provides analysis and a set of concrete recommendations, both for the period of transition before the inauguration, and for the start of the next presidency, to encourage and sustain innovation in government. Leveraging the insights from the experts who participated in a day-long discussion, we endeavor to explain how government can improve its use of using digital technologies to create more effective policies, solve problems faster and deliver services more effectively at the federal, state and local levels….
The broad recommendations are:
- Scale Data Driven Governance: Platforms such as data.gov represent initial steps in the direction of enabling data-driven governance. Much more can be done, however, to open-up data and for the agencies to become better consumers of data, to improve decision-making and scale up evidence-based governance. This includes better use of predictive analytics, more public engagement; and greater use of cutting-edge methods like machine learning.
- Scale Collaborative Innovation: Collaborative innovation takes place when government and the public work together, thus widening the pool of expertise and knowledge brought to bear on public problems. The next administration can reach out more effectively, not just to the public at large, but to conduct targeted outreach to public officials and citizens who possess the most relevant skills or expertise for the problems at hand.
- Promote a Culture of Innovation: Institutionalizing a culture of technology-enabled innovation will require embedding and institutionalizing innovation and technology skills more widely across the federal enterprise. For example, contracting, grants and personnel officials need to have a deeper understanding of how technology can help them do their jobs more efficiently, and more people need to be trained in human-centered design, gamification, data science, data visualization, crowdsourcing and other new ways of working.
- Utilize Evidence-Based Innovation: In order to better direct government investments, leaders need a much better sense of what works and what doesn’t. The government spends billions on research in the private and university sectors, but very little experimenting with, testing, and evaluating its own programs. The next administration should continue developing an evidence-based approach to governance, including a greater use of methods like A/B testing (a method of comparing two versions of a webpage or app against each other to determine which one performs the best); establishing a clearinghouse for success and failure stories and best practices; and encouraging overseers to be more open to innovation.
- Make Innovation a Priority in the Transition: The transition period represents a unique opportunity to seed the foundations for long-lasting change. By explicitly incorporating innovation into the structure, goals and activities of the transition teams, the next administration can get a fast start in implementing policy goals and improving government operations through innovation approaches….(More)”
How the Federal Government is thinking about Artificial Intelligence
Mohana Ravindranath at NextGov: “Since May, the White House has been exploring the use of artificial intelligence and machine learning for the public: that is, how the federal government should be investing in the technology to improve its own operations. The technologies, often modeled after the way humans take in, store and use new information, could help researchers find patterns in genetic data or help judges decide sentences for criminals based on their likelihood to end up there again, among other applications. …
Here’s a look at how some federal groups are thinking about the technology:
- Police data: At a recent White House workshop, Office of Science and Technology Policy Senior Adviser Lynn Overmann said artificial intelligence could help police departments comb through hundreds of thousands of hours of body-worn camera footage, potentially identifying the police officers who are good at de-escalating situations. It also could help cities determine which individuals are likely to end up in jail or prison and officials could rethink programs. For example, if there’s a large overlap between substance abuse and jail time, public health organizations might decide to focus their efforts on helping people reduce their substance abuse to keep them out of jail.
- Explainable artificial intelligence: The Pentagon’s research and development agency is looking for technology that can explain to analysts how it makes decisions. If people can’t understand how a system works, they’re not likely to use it, according to a broad agency announcement from the Defense Advanced Research Projects Agency. Intelligence analysts who might rely on a computer for recommendations on investigative leads must “understand why the algorithm has recommended certain activity,” as do employees overseeing autonomous drone missions.
- Weather detection: The Coast Guard recently posted its intent to sole-source a contract for technology that could autonomously gather information about traffic, crosswind, and aircraft emergencies. That technology contains built-in artificial intelligence technology so it can “provide only operational relevant information.”
- Cybersecurity: The Air Force wants to make cyber defense operations as autonomous as possible, and is looking at artificial intelligence that could potentially identify or block attempts to compromise a system, among others.
While there are endless applications in government, computers won’t completely replace federal employees anytime soon….(More)”
How Tech Giants Are Devising Real Ethics for Artificial Intelligence
For years, science-fiction moviemakers have been making us fear the bad things that artificially intelligent machines might do to their human creators. But for the next decade or two, our biggest concern is more likely to be that robots will take away our jobs or bump into us on the highway.
Now five of the world’s largest tech companies are trying to create a standard of ethics around the creation of artificial intelligence. While science fiction has focused on the existential threat of A.I. to humans,researchers at Google’s parent company, Alphabet, and those from Amazon,Facebook, IBM and Microsoft have been meeting to discuss more tangible issues, such as the impact of A.I. on jobs, transportation and even warfare.
Tech companies have long overpromised what artificially intelligent machines can do. In recent years, however, the A.I. field has made rapid advances in a range of areas, from self-driving cars and machines that understand speech, like Amazon’s Echo device, to a new generation of weapons systems that threaten to automate combat.
The specifics of what the industry group will do or say — even its name —have yet to be hashed out. But the basic intention is clear: to ensure thatA.I. research is focused on benefiting people, not hurting them, according to four people involved in the creation of the industry partnership who are not authorized to speak about it publicly.
The importance of the industry effort is underscored in a report issued onThursday by a Stanford University group funded by Eric Horvitz, a Microsoft researcher who is one of the executives in the industry discussions. The Stanford project, called the One Hundred Year Study onArtificial Intelligence, lays out a plan to produce a detailed report on the impact of A.I. on society every five years for the next century….The Stanford report attempts to define the issues that citizens of a typicalNorth American city will face in computers and robotic systems that mimic human capabilities. The authors explore eight aspects of modern life,including health care, education, entertainment and employment, but specifically do not look at the issue of warfare..(More)”
The risks of relying on robots for fairer staff recruitment
Sarah O’Connor at the Financial Times: “Robots are not just taking people’s jobs away, they are beginning to hand them out, too. Go to any recruitment industry event and you will find the air is thick with terms like “machine learning”, “big data” and “predictive analytics”.
The argument for using these tools in recruitment is simple. Robo-recruiters can sift through thousands of job candidates far more efficiently than humans. They can also do it more fairly. Since they do not harbour conscious or unconscious human biases, they will recruit a more diverse and meritocratic workforce.
This is a seductive idea but it is also dangerous. Algorithms are not inherently neutral just because they see the world in zeros and ones.
For a start, any machine learning algorithm is only as good as the training data from which it learns. Take the PhD thesis of academic researcher Colin Lee, released to the press this year. He analysed data on the success or failure of 441,769 job applications and built a model that could predict with 70 to 80 per cent accuracy which candidates would be invited to interview. The press release plugged this algorithm as a potential tool to screen a large number of CVs while avoiding “human error and unconscious bias”.
But a model like this would absorb any human biases at work in the original recruitment decisions. For example, the research found that age was the biggest predictor of being invited to interview, with the youngest and the oldest applicants least likely to be successful. You might think it fair enough that inexperienced youngsters do badly, but the routine rejection of older candidates seems like something to investigate rather than codify and perpetuate. Mr Lee acknowledges these problems and suggests it would be better to strip the CVs of attributes such as gender, age and ethnicity before using them….(More)”
Technology Is Monitoring the Urban Landscape
Big City is watching you.
It will do it with camera-equipped drones that inspect municipal powerlines and robotic cars that know where people go. Sensor-laden streetlights will change brightness based on danger levels. Technologists and urban planners are working on a major transformation of urban landscapes over the next few decades.
Much of it involves the close monitoring of things and people, thanks to digital technology. To the extent that this makes people’s lives easier, the planners say, they will probably like it. But troubling and knotty questions of privacy and control remain.
A White House report published in February identified advances in transportation, energy and manufacturing, among other developments, that will bring on what it termed “a new era of change.”
Much of the change will also come from the private sector, which is moving faster to reach city dwellers, and is more skilled in collecting and responding to data. That is leading cities everywhere to work more closely than ever with private companies, which may have different priorities than the government.
One of the biggest changes that will hit a digitally aware city, it is widely agreed, is the seemingly prosaic issue of parking. Space given to parking is expected to shrink by half or more, as self-driving cars and drone deliveries lead an overall shift in connected urban transport. That will change or eliminate acres of urban space occupied by raised and underground parking structures.
Shared vehicles are not parked as much, and with more automation, they will know where parking spaces are available, eliminating the need to drive in search of a space.
“Office complexes won’t need parking lots with twice the footprint of their buildings,” said Sebastian Thrun, who led Google’s self-driving car project in its early days and now runs Udacity, an online learning company. “Whenwe started on self-driving cars, we talked all the time about cutting the number of cars in a city by a factor of three,” or a two-thirds reduction.
In addition, police, fire, and even library services will seek greater responsiveness by tracking their own assets, and partly by looking at things like social media. Later, technologies like three-dimensional printing, new materials and robotic construction and demolition will be able to reshape skylines in a matter of weeks.
At least that is the plan. So much change afoot creates confusion….
The new techno-optimism is focused on big data and artificial intelligence.“Futurists used to think everyone would have their own plane,” said ErickGuerra, a professor of city and regional planning at the University ofPennsylvania. “We never have a good understanding of how things will actually turn out.”
He recently surveyed the 25 largest metropolitan planning organizations in the country and found that almost none have solid plans for modernizing their infrastructure. That may be the right way to approach the challenges of cities full of robots, but so far most clues are coming from companies that also sell the technology.
“There’s a great deal of uncertainty, and a competition to show they’re low on regulation,” Mr. Guerra said. “There is too much potential money for new technology to be regulated out.”
The big tech companies say they are not interested in imposing the sweeping “smart city” projects they used to push, in part because things are changing too quickly. But they still want to build big, and they view digital surveillance as an essential component…(More)”
Can mobile usage predict illiteracy in a developing country?
Pål Sundsøy at arXiv: “The present study provides the first evidence that illiteracy can be reliably predicted from standard mobile phone logs. By deriving a broad set of mobile phone indicators reflecting users financial, social and mobility patterns we show how supervised machine learning can be used to predict individual illiteracy in an Asian developing country, externally validated against a large-scale survey. On average the model performs 10 times better than random guessing with a 70% accuracy. Further we show how individual illiteracy can be aggregated and mapped geographically at cell tower resolution. Geographical mapping of illiteracy is crucial to know where the illiterate people are, and where to put in resources. In underdeveloped countries such mappings are often based on out-dated household surveys with low spatial and temporal resolution. One in five people worldwide struggle with illiteracy, and it is estimated that illiteracy costs the global economy more than 1 trillion dollars each year. These results potentially enable cost-effective, questionnaire-free investigation of illiteracy-related questions on an unprecedented scale…(More)”.
Enablers for Smart Cities
Book by Amal El Fallah Seghrouchni, Fuyuki Ishikawa, Laurent Hérault, and Hideyuki Tokuda: “Smart cities are a new vision for urban development. They integrate information and communication technology infrastructures – in the domains of artificial intelligence, distributed and cloud computing, and sensor networks – into a city, to facilitate quality of life for its citizens and sustainable growth. This book explores various concepts for the development of these new technologies (including agent-oriented programming, broadband infrastructures, wireless sensor networks, Internet-based networked applications, open data and open platforms), and how they can provide smart services and enablers in a range of public domains.
The most significant research, both established and emerging, is brought together to enable academics and practitioners to investigate the possibilities of smart cities, and to generate the knowledge and solutions required to develop and maintain them…(More)”
What Governments Can Learn From Airbnb And the Sharing Economy
Arun Sundararajan in Fortune: “….Despite some regulators’ fears, the sharing economy may not result in the decline of regulation but rather in its opposite, providing a basis upon which society can develop more rational, ethical, and participatory models of regulation. But what regulation looks like, as well as who actually creates and enforce the regulation, is also bound to change.
There are three emerging models – peer regulation, self-regulatory organizations, and data-driven delegation – that promise a regulatory future for the sharing economy best aligned with society’s interests. In the adapted book excerpt that follows, I explain how the third of these approaches, of delegating enforcement of regulations to companies that store critical data on consumers, can help mitigate some of the biases Airbnb guests may face, and why this is a superior alternative to the “open data” approach of transferring consumer information to cities and state regulators.
Consider a different problem — of collecting hotel occupancy taxes from hundreds of thousands of Airbnb hosts rather than from a handful of corporate hotel chains. The delegation of tax collection to Airbnb, something a growing number of cities are experimenting with, has a number of advantages. It is likely to yield higher tax revenues and greater compliance than a system where hosts are required to register directly with the government, which is something occasional hosts seem reluctant to do. It also sidesteps privacy concerns resulting from mandates that digital platforms like Airbnb turn over detailed user data to the government. There is also significant opportunity for the platform to build credibility as it starts to take on quasi governmental roles like this.
There is yet another advantage, and the one I believe will be the most significant in the long-run. It asks a platform to leverage its data to ensure compliance with a set of laws in a manner geared towards delegating responsibility to the platform. You might say that the task in question here — computing tax owed, collecting, and remitting it—is technologically trivial. True. But I like this structure because of the potential it represents. It could be a precursor for much more exciting delegated possibilities.
For a couple of decades now, companies of different kinds have been mining the large sets of “data trails” customers provide through their digital interactions. This generates insights of business and social importance. One such effort we are all familiar with is credit card fraud detection. When an unusual pattern of activity is detected, you get a call from your bank’s security team. Sometimes your card is blocked temporarily. The enthusiasm of these digital security systems is sometimes a nuisance, but it stems from your credit card company using sophisticated machine learning techniques to identify patterns that prior experience has told it are associated with a stolen card. It saves billions of dollars in taxpayer and corporate funds by detecting and blocking fraudulent activity swiftly.
A more recent visible example of the power of mining large data sets of customer interaction came in 2008, when Google engineers announced that they could predict flu outbreaks using data collected from Google searches, and track the spread of flu outbreaks in real time, providing information that was well ahead of the information available using the Center for Disease Control’s (CDC) own tracking systems. The Google system’s performance deteriorated after a couple of years, but its impact on public perception of what might be possible using “big data” was immense.
It seems highly unlikely that such a system would have emerged if Google had been asked to hand over anonymized search data to the CDC. In fact, there would have probably been widespread public backlash to this on privacy grounds. Besides, the reason why this capability emerged organically from within Google is partly as a consequence of Google having one of the highest concentrations of computer science and machine learning talent in the world.
Similar approaches hold great promise as a regulatory approach for sharing economy platforms. Consider the issue of discriminatory practices. There has long been anecdotal evidence that some yellow cabs in New York discriminate against some nonwhite passengers. There have been similar concerns that such behavior may start to manifest on ridesharing platforms and in other peer-to-peer markets for accommodation and labor services.
For example, a 2014 study by Benjamin Edelman and Michael Luca of Harvard suggested that African American hosts might have lower pricing power than white hosts on Airbnb. While the study did not conclusively establish that the difference is due to guests discriminating against African American hosts, a follow-up study suggested that guests with “distinctively African American names” were less likely to receive favorable responses for their requests to Airbnb hosts. This research raises a red flag about the need for vigilance as the lines between personal and professional blur.
One solution would be to apply machine-learning techniques to be able to identify patterns associated with discriminatory behavior. No doubt, many platforms are already using such systems….(More)”