machine learning

Crowdbreaks: Tracking Health Trends using Public Social Media Data and Crowdsourcing

Curated on May 19, 2018November 15, 2018 by admin

Paper by Martin Mueller and Marcel Salath: “In the past decade, tracking health trends using social media data has shown great promise, due to a powerful combination of massive adoption of social media around the world, and increasingly potent hardware and software that enables us to work with these new big data streams.

At the same time, many challenging problems have been identified. First, there is often a mismatch between how rapidly online data can change, and how rapidly algorithms are updated, which means that there is limited reusability for algorithms trained on past data as their performance decreases over time. Second, much of the work is focusing on specific issues during a specific past period in time, even though public health institutions would need flexible tools to assess multiple evolving situations in real time. Third, most tools providing such capabilities are proprietary systems with little algorithmic or data transparency, and thus little buy-in from the global public health and research community.

Here, we introduce Crowdbreaks, an open platform which allows tracking of health trends by making use of continuous crowdsourced labelling of public social media content. The system is built in a way which automatizes the typical workflow from data collection, filtering, labelling and training of machine learning classifiers and therefore can greatly accelerate the research process in the public health domain. This work introduces the technical aspects of the platform and explores its future use cases…(More)”.

Creating a Machine Learning Commons for Global Development

Curated on May 7, 2018August 3, 2018 by Stefaan Verhulst

Blog by Hamed Alemohammad: “Advances in sensor technology, cloud computing, and machine learning (ML) continue to converge to accelerate innovation in the field of remote sensing. However, fundamental tools and technologies still need to be developed to drive further breakthroughs and to ensure that the Global Development Community (GDC) reaps the same benefits that the commercial marketplace is experiencing. This process requires us to take a collaborative approach.

Data collaborative innovation — that is, a group of actors from different data domains working together toward common goals — might hold the key to finding solutions for some of the global challenges that the world faces. That is why Radiant.Earth is investing in new technologies such as Cloud Optimized GeoTiffs, Spatial Temporal Asset Catalogues (STAC), and ML. Our approach to advance ML for global development begins with creating open libraries of labeled images and algorithms. This initiative and others require — and, in fact, will thrive as a result of — using a data collaborative approach.

“Data is only as valuable as the decisions it enables.”

This quote by Ion Stoica, professor of computer science at the University of California, Berkeley, may best describe the challenge facing those of us who work with geospatial information:

How can we extract greater insights and value from the unending tsunami of data that is before us, allowing for more informed and timely decision making?…(More).

Privacy and Freedom of Expression In the Age of Artificial Intelligence

Curated on April 30, 2018August 3, 2018 by Stefaan Verhulst

Joint Paper by Privacy International and ARTICLE 19: “Artificial Intelligence (AI) is part of our daily lives. This technology shapes how people access information, interact with devices, share personal information, and even understand foreign languages. It also transforms how individuals and groups can be tracked and identified, and dramatically alters what kinds of information can be gleaned about people from their data. AI has the potential to revolutionise societies in positive ways. However, as with any scientific or technological advancement, there is a real risk that the use of new tools by states or corporations will have a negative impact on human rights. While AI impacts a plethora of rights, ARTICLE 19 and Privacy International are particularly concerned about the impact it will have on the right to privacy and the right to freedom of expression and information. This scoping paper focuses on applications of ‘artificial narrow intelligence’: in particular, machine learning and its implications for human rights.

The aim of the paper is fourfold:

1. Present key technical definitions to clarify the debate;

2. Examine key ways in which AI impacts the right to freedom of expression and the right to privacy and outline key challenges;

3. Review the current landscape of AI governance, including various existing legal, technical, and corporate frameworks and industry-led AI initiatives that are relevant to freedom of expression and privacy; and

4. Provide initial suggestions for rights-based solutions which can be pursued by civil society organisations and other stakeholders in AI advocacy activities….(More)”.

Artificial Unintelligence

Curated on April 23, 2018August 3, 2018 by Stefaan Verhulst

Book by Meredith Broussard: “A guide to understanding the inner workings and outer limits of technology and why we should never assume that computers always get it right.

In Artificial Unintelligence, Meredith Broussard argues that our collective enthusiasm for applying computer technology to every aspect of life has resulted in a tremendous amount of poorly designed systems. We are so eager to do everything digitally—hiring, driving, paying bills, even choosing romantic partners—that we have stopped demanding that our technology actually work. Broussard, a software developer and journalist, reminds us that there are fundamental limits to what we can (and should) do with technology. With this book, she offers a guide to understanding the inner workings and outer limits of technology—and issues a warning that we should never assume that computers always get things right.

Making a case against technochauvinism—the belief that technology is always the solution—Broussard argues that it’s just not true that social problems would inevitably retreat before a digitally enabled Utopia. To prove her point, she undertakes a series of adventures in computer programming. She goes for an alarming ride in a driverless car, concluding “the cyborg future is not coming any time soon”; uses artificial intelligence to investigate why students can’t pass standardized tests; deploys machine learning to predict which passengers survived the Titanic disaster; and attempts to repair the U.S. campaign finance system by building AI software. If we understand the limits of what we can do with technology, Broussard tells us, we can make better choices about what we should do with it to make the world better for everyone…(More)”.

AI And Open Data Show Just How Often Cars Block Bus And Bike Lanes

Curated on March 27, 2018August 3, 2018 by Stefaan Verhulst

Eillie Anzilotti in Fast Company: “…While anyone who bikes or rides a bus in New York City knows intuitively that the lanes are often blocked, there’s been little data to back up that feeling apart from the fact that last year, the NYPD issues 24,000 tickets for vehicles blocking bus lanes, and around 79,000 to cars in the bike lane. By building the algorithm, Bell essentializes what engaged citizenship and productive use of open data looks like. The New York City Department of Transportation maintains several hundred video cameras throughout the city; those cameras feed images in real time to the DOT’s open-data portal. Bell downloaded a week’s worth of footage from that portal to analyze.

To build his computer algorithm to do the analysis, he fed around 2,000 images of buses, cars, pedestrians, and vehicles like UPS trucks into TensorFlow, Google’s open-source framework that the tech giant is using to train autonomous vehicles to recognize other road users. “Because of the push into AVs, machine learning in general and neural networks have made lots of progress, because they have to answer the same questions of: What is this vehicle, and what is it going to do?” Bell says. After several rounds of processing, Bell was able to come up with an algorithm that fairly faultlessly could determine if a vehicle at the bus stop was, in fact, a bus, or if it was something else that wasn’t supposed to be there.

As cities and governments, spurred by organizations like OpenGov, have moved to embrace transparency and open data, the question remains: So, what do you do with it?

For Bell, the answer is that citizens can use it to empower themselves. “I’m a little uncomfortable with cameras and surveillance in cities,” Bell says. “But agencies like the NYPD and DOT have already made the decision to put the cameras up. We don’t know the positive and negative outcomes if more and more data from cameras is opened to the public, but if the cameras are going in, we should know what data they’re collecting and be able to access it,” he says. He’s made his algorithm publicly available in the hopes that more people will use data to investigate the issue on their own streets, and perhaps in other cities….Bell is optimistic that open data can empower more citizens to identify issues in their own cities and bring a case for why they need to be addressed….(More)”.

How Refugees Are Helping Create Blockchain’s Brand New World

Curated on March 21, 2018August 3, 2018 by Stefaan Verhulst

Jessi Hempel at Wired: “Though best known for underpinning volatile cryptocurrencies, like Bitcoin and Ethereum, blockchain technology has a number of qualities which make it appealing for record-keeping. A distributed ledger doesn’t depend on a central authority to verify its existence, or to facilitate transactions within it, which makes it less vulnerable to tampering. By using applications that are built on the ‘chain, individuals may be able to build up records over time, use those records across borders as a form of identity—essentially creating the trust they need to interact with the world, without depending on a centralized authority, like a government or a bank, to vouch for them.

For now, these efforts are small experiments. In Finland, the Finnish Immigration Service offers refugees a prepaid Mastercard developed by the Helsinki-based startup MONI that also links to a digital identity, composed of the record of one’s financial transactions, which is stored on the blockchain. In Moldova, the government is working with digital identification expertsfrom the United Nations Office for Project Services (UNOPS) to brainstorm ways to use blockchain to provide children living in rural areas with a digital identity, so it’s more difficult for traffickers to smuggle them across borders.

Among the more robust programs is a pilot the United Nations World Food Program (WFP) launched in Jordan last May. Syrian refugees stationed at the Azraq Refugee Camp receive vouchers to shop at the local grocery store. The WFP integrated blockchain into its biometric authentication technology, so Syrian refugees can cash in their vouchers at the supermarket by staring into a retina scanner. These transactions are recorded on a private Ethereum-basedblockchain, called Building Blocks. Because the blockchain eliminates the need for WFP to pay banks to facilitate transactions, Building Blocks could save the WFP as much as $150,000 each month in bank fees in Jordan alone. The program has been so successful that by the end of the year, the WFP plans to expand the technology throughout Jordan. Blockchain enthusiasts imagine a future in which refugees can access more than just food vouchers, accumulating a transaction history that could stand in as a credit history when they attempt to resettle….

But in the rush to apply blockchain technology to every problem, many point out that relying on the ledger may have unintended consequences. As the Blockchain for Social Impact chief technology officer at ConsenSys, Robert Greenfeld IV writes, blockchain-based identity “isn’t a silver bullet, and if we don’t think about it/build it carefully, malicious actors could still capitalize on it as an element of control.” If companies rely on private blockchains, he warns, there’s a danger that the individual permissions will prevent these identity records from being used in multiple places. (Many of these projects, like the UNWFP project, are built on private blockchains so that organizations can exert more control over their development.) “If we don’t start to collaborate together with populations, we risk ending up with a bunch of siloed solutions,” says Greenfeld.

For his part, Greenfeld suggests governments could easily use state-sponsored machine learning algorithms to monitor public blockchain activity. But as bitcoin enthusiasts branch out of their get-rich-quick schemes to wrestle with how to make the web more equitable for everyone, they have the power to craft a world of their own devising. The early web should be a lesson to the bitcoin enthusiasts as they promote the blockchain’s potential. Right now we have the power to determine its direction; the dangers exist, but the potential is enormous….(More)”

The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation

Curated on March 1, 2018August 3, 2018 by Stefaan Verhulst

Report by Miles Brundage et al: “Artificial intelligence and machine learning capabilities are growing at an unprecedented rate. These technologies have many widely beneficial applications, ranging from machine translation to medical image analysis. Countless more such applications are being developed and can be expected over the long term. Less attention has historically been paid to the ways in which artificial intelligence can be used maliciously. This report surveys the landscape of potential security threats from malicious uses of artificial intelligence technologies, and proposes ways to better forecast, prevent, and mitigate these threats. We analyze, but do not conclusively resolve, the question of what the long-term equilibrium between attackers and defenders will be. We focus instead on what sorts of attacks we are likely to see soon if adequate defenses are not developed.

In response to the changing threat landscape we make four high-level recommendations:

1. Policymakers should collaborate closely with technical researchers to investigate, prevent, and mitigate potential malicious uses of AI.

2. Researchers and engineers in artificial intelligence should take the dual-use nature of their work seriously, allowing misuserelated considerations to influence research priorities and norms, and proactively reaching out to relevant actors when harmful applications are foreseeable.

3. Best practices should be identified in research areas with more mature methods for addressing dual-use concerns, such as computer security, and imported where applicable to the case of AI.

4. Actively seek to expand the range of stakeholders and domain experts involved in discussions of these challenges….(More)”.

How AI-Driven Insurance Could Reduce Gun Violence

Curated on February 28, 2018August 3, 2018 by Stefaan Verhulst

Jason Pontin at WIRED: “As a political issue, guns have become part of America’s endless, arid culture wars, where Red and Blue tribes skirmish for political and cultural advantage. But what if there were a compromise? Economics and machine learning suggest an answer, potentially acceptable to Americans in both camps.

Economists sometimes talk about “negative externalities,” market failures where the full costs of transactions are borne by third parties. Pollution is an externality, because society bears the costs of environmental degradation. The 20th-century British economist Arthur Pigou, who formally described externalities, also proposed their solution: so-called “Pigovian taxes,” where governments charge producers or customers, reducing the quantity of the offending products and sometimes paying for ameliorative measures. Pigovian taxes have been used to fight cigarette smoking or improve air quality, and are the favorite prescription of economists for reducing greenhouse gases. But they don’t work perfectly, because it’s hard for governments to estimate the costs of externalities.

Gun violence is a negative externality too. The choices of millions of Americans to buy guns overflow into uncaptured costs for society in the form of crimes, suicides, murders, and mass shootings. A flat gun tax would be a blunt instrument: It could only reduce gun violence by raising the costs of gun ownership so high that almost no one could legally own a gun, which would swell the black market for guns and probably increase crime. But insurers are very good at estimating the risks and liabilities of individual choices; insurance could capture the externalities of gun violence in a smarter, more responsive fashion.

Here’s the proposed compromise: States should require gun owners to be licensed and pay insurance, just as car owners must be licensed and insured today….

The actuaries who research risk have always considered a wide variety of factors when helping insurers price the cost of a policy. Car, home, and life insurance can vary according to a policy holder’s age, health, criminal record, employment, residence, and many other variables. But in recent years, machine learning and data analytics have provided actuaries with new predictive powers. According to Yann LeCun, the director of artificial intelligence at Facebook and the primary inventor of an important technique in deep learning called convolution, “Deep learning systems provide better statistical models with enough data. They can be advantageously applied to risk evaluation, and convolutional neural nets can be very good at prediction, because they can take into account a long window of past values.”

State Farm, Liberty Mutual, Allstate, and Progressive Insurance have all used algorithms to improve their predictive analysis and to more accurately distribute risk among their policy holders. For instance, in late 2015, Progressive created a telematics app called Snapshot that individual drivers used to collect information on their driving. In the subsequent two years, 14 billion miles of driving data were collected all over the country and analyzed on Progressive’s machine learning platform, H20.ai, resulting in discounts of $600 million for their policy holders. On average, machine learning produced a $130 discount for Progressive customers.

When the financial writer John Wasik popularized gun insurance in a series of posts in Forbes in 2012 and 2013, the NRA’s argument about prior constraints was a reasonable objection. Wasik proposed charging different rates to different types of gun owners, but there were too many factors that would have to be tracked over too long a period to drive down costs for low-risk policy holders. Today, using deep learning, the idea is more practical: Insurers could measure the interaction of dozens or hundreds of factors, predicting the risks of gun ownership and controlling costs for low-risk gun owners. Other, more risky bets might pay more. Some very risky would-be gun owners might be unable to find insurance at all. Gun insurance could even be dynamically priced, changing as the conditions of the policy holders’ lives altered, and the gun owners proved themselves better or worse risks.

Requiring gun owners to buy insurance wouldn’t eliminate gun violence in America. But a political solution to the problem of gun violence is chimerical….(More)”.

What if technology could help improve conversations online?

Curated on February 21, 2018August 3, 2018 by Stefaan Verhulst

Introduction to “Perspective”: “Discussing things you care about can be difficult. The threat of abuse and harassment online means that many people stop expressing themselves and give up on seeking different opinions….Perspective is an API that makes it easier to host better conversations. The API uses machine learning models to score the perceived impact a comment might have on a conversation. Developers and publishers can use this score to give realtime feedback to commenters or help moderators do their job, or allow readers to more easily find relevant information, as illustrated in two experiments below. We’ll be releasing more machine learning models later in the year, but our first model identifies whether a comment could be perceived as “toxic” to a discussion….(More)”.