Monthly Archives: July 2017

For every exciting opportunity promised by artificial intelligence, theres a potential downside that is its bleak mirror image. We hope that AI will allow us to make smarter decisions, but what if it ends up reinforcing the prejudices of society? We dream that technology might free us from work, but what if only the rich benefit, while the poor are dispossessed?

Its issues like these that keep artificial intelligence researchers up at night, and theyre also the reason that Google is launching an AI initiative today to tackle some of these same problems. The new project is named PAIR (it stands for People + AI Research) and its aim is to study and redesign the ways people interact with AI systems and try to ensure that the technology benefits and empowers everyone.

Google wants to help everyone from coders to users

Its a broad remit, and an ambitious one. Google says PAIR will look at a number of different issues affecting everyone in the AI supply chain from the researchers who code algorithms, to the professionals like doctors and farmers who are (or soon will be) using specialized AI tools. The tech giant says it wants to make AI user-friendly, and that means not only making the technology easy to understand (getting AI to explain itself is a known and challenging problem) but also ensuring that it treats its users equally.

Its been noted time and time again that the prejudices and inequalities of society often become hard-coded in AI. This might mean facial recognition software that doesnt recognize dark-skinned users, or a language processing program which assume that doctors are always male and nurses are always female.

Usually this sort of issue is caused by the data that artificial intelligence is trained on. Either the information it has it incomplete, or its prejudiced in some way. Thats why PAIRs first real news is the announcement of two new open-source tools called Facets Overview and Facets Dive which make it easier for programmers to examine datasets.

In the screenshot above Facets Dive is being used to test a facial recognition system. The program is sorting the testers by their country of origin and comparing errors with successful identifications. This allows a coder to quickly see where their dataset is falling short, and make the relevant adjustments.

Currently, PAIR has 12 full-time staff. Its a bit of a small figure considering the scale of the problem, but Google says PAIR is really a company-wide initiative one that will draw in expertise from the firms various departments.

More open-source tools like Facets will be released in the future, and Google will also be setting up new grants and residencies to sponsor related research. Its not the only big organization taking these issues seriously (see also: the Ethics and Governance of Artificial Intelligence Fund and Elon Musk-funded OpenAI), but its good to see Google join the fight for a fairer future.

Follow this link:

Google wants to make sure AI advances don't leave anyone behind - The Verge

Image: iStock/surfleader

According to Samsung, global internet traffic surpassed one zettabyte or one billion terabytes in 2016. That number is huge, but it doesn't begin to approach the total data that companies are storing.

Even more concerning is the possibility that, at most companies, data "under management" is a misnomer.

Key areas of data management challenge are:

IT departments struggle in these areas for the following reasons:

The question now is: can machine learning, artificial intelligence (AI) and analytics provide assistance in the area of data managementespecially for the large amount unstructured data?

SEE: As EU's General Data Protection Regulation (GDPR) looms, tech vendors ready pitches (ZDNet)

Here is where machine learning, AI and analytics can help:

Sorting through dark data

Every corporate system, and every business department, has troves of data that have accumulated but that people know nothing about. By using machine learning and combining its power with algorithms that address how to sort and handle different types of emails, documents, images, etc., stored on servers, machine learning, AI and analytics can go to work on this unplumbed data and pre-sort it for you. A knowledgeable human can then review what the automation recommends as a data classification scheme, tweak it, and perform the scheme. Part of the process could also address data retention, with the analytics producing a set of recommendations on which data could potentially be purged from files.

Deciding what to throw away

Machine learning, analytics, and AI can objectively identify data that is seldom or never used, and recommend that you throw it away, but it doesn't have the same discernment abilities that employees do. For instance, these processes can pick out pieces of data or records that haven't been accessed for more than five years, indicating that the data could be obsolete. This saves an employee time hunting down this potentially obsolete data, because now all they need to do is to determine whether there is any reason to keep it.

Aggregating data

When analytics developers determine the kinds of data they need to aggregate for queries, they often produce a repository for the application, and then pull in various types of data from different sources to make up an analytics data pool. To do this, they must develop integration methods to access the different sources from which they pull data. Machine learning can make this still very manual process more efficient by automatically developing "mappings" between data sources and the application's data repository. This cuts down integration and aggregation times.

Organizing data storage for best access

Over the past five years, data storage vendors have made significant inroads into automating storage management, thanks to the development of lower cost solid state storage. These technology advances have enabled IT departments to use "smart" storage engines that use machine learning to see which types of data are used most often, and which are seldom or never used. The automation can be used to automatically store data in fast or slow storage, based on the business rules inserted into machine algorithms. The automation saves storage managers from having to address storage optimization manually.

Data management is a major IT challenge that is not close to resolution in most organizationsand it is going to get worse as the data continues to stream in.

CIOs, data architects, and storage managers need to highlight the issue to C-level executives, but data management projects are not easy "sells."

Nevertheless, by pointing out the value of faster times to market for analytics and potential person power and storage cost reductions for data management, IT managers at least have viable entry points into C-level discussions about how to increase strategic agility and reduce cost of operations at the same time.

Read the rest here:

How AI and machine learning can help solve IT's data management problem - TechRepublic

Next month, Comcast will host PHLAI, a technical conference for engineers and professionals interested in and working with machine learning and artificial intelligence.

Well bring together local practitioners in A.I. and machine learning to discuss past experiences and common technological goals aimed at making peoples lives better. Attendees will learn about new ideas and best practices from experts in the field and hear about the latest developments in machine learning and artificial intelligence.

As an example, I was fortunate to be part of the team here at Comcast that used A.I. to launch our Xfinity X1 voice remote. That device has changed the way people watch television, and to date weve deployed more than 14 million of them in homes all across our service area from San Francisco to Philadelphia. And it keeps getting smarter, faster and more accurate every day, all thanks to machine learning.

The PHLAIconference will take place onTuesday, Aug.15 at Convene Cira Centre in Philadelphia.

Featured speakers will include:

And as part of the event, we also want to hear from those who are solving their own problems with A.I.

Practitioners can share their stories and submit proposals until July 14.

Attendees can register here(its free). We hope to see you there!

Jeanine Heck serves as Executive Director in the Technology and Product organization of Comcast Cable. In this role, Heck brings artificial intelligence into XFINITY products. She was the founding product manager for the X1 voice remote, has led the launch of a TV search engine, and managed the companys first TV recommendations engine.

See more here:

Calling all AI experts - Technical.ly

A $27 million fund aimed at applying artificial intelligence to the public interest has announced the first targets for its beneficence: $7.6 million will be split unequally between MITs Media Lab, Harvards Berkman Klein Center and seven smaller research efforts around the world.

The Ethics and Governance of Artificial Intelligence Fund was created by Reid Hoffman, Pierre Omidyar and the Knight Foundation back in January; the intention was to ensure that social scientists, ethicists, philosophers, faith leaders, economists, lawyers and policymakers have a say in how AI is developed and deployed.

To that end, this first round of fundings supports existing organizations working along those lines, as well as nurturing some newer ones.

The lions share of this initial round, $5.9 million, will be split by MIT and Harvard, as the initial announcement indicated. Media Lab is, of course, on the cutting edge of many research efforts in AI and elsewhere; Berkman Klein focuses more on the legal and analysis side of things.

The funds focuses are threefold:

Those two well-known organizations will be pursuing issues related to those (theyre already working together anyway), but the seven smaller efforts are also being more modestly funded.

Digital Asia Hub, FAT ML and ITS Rio will be hosting conferences and workshops to which experts across fields will be invited, advancing and enriching the conversations around various AI issues. ITS Rio also will be translating debates on the topics a critical task, since there are important thinkers worldwide and these conversations shouldnt be limited by something as last-century as native language.

On the research side, AI Now will be looking at bias in data collection and healthcare; the Leverhulme Center will be looking at interpretability of AI-related data; Data & Society will be conducting ethnographically-informed studies on the human element of AI and data for example, how demographic imbalances in who runs real estate businesses might inform the systems they create and use.

Access Now (which doesnt really fit in either category) will be working to create a set of guidelines for businesses and services looking to conform to major upcoming data regulations in the EU.

For this initial cohort, we looked for projects that fit our goal of building networks across fields, and that would complement the work of our anchor partners at the Media Lab and Berkman Klein, said Knights VP of Technology and Innovation, John Bracken, in an email to TechCrunch.

We think its vital that civil society has a strong voice in the development of artificial intelligence and machine learning. We see these projects as part of a growing set of researchers, engineers, and policy makers who will be part of ensuring that these new tools are developed ethically.

Although the funds are in the public interest, they arent just handouts; I asked Bracken whether there were any concrete expectations for the organizations involved.

Absolutely, he said. The discussion around artificial intelligence is no longer a far-off, speculative thing. Each of the grants were making have deliverables planned for the next twelve months, and well be showcasing them as they launch.

Well hear about them soon, no doubt.

A few million bucks may seem like a drop in the bucket among the herds of unicorns we track here at TechCrunch, but on the other hand it may seem cheap when the studies and events being funded come to fruition and result in the kind of productive dialogue this fast-moving field needs.

View original post here:

Ethics and Governance AI Fund funnels $7.6M to Harvard, MIT and independent research efforts - TechCrunch

Healthcare inequalities are systemic and closely intertwined with social inequalities. In the US, black men and women can be expected to live a decade less than their white counterparts, and are also much more likely to die from heart disease, various types of cancer, and stroke. Rates of diabetes in Hispanic Americans are around 30% higher than in whites. Gay, lesbian, and bisexual adults are twice as likely to suffer with mental-health problems. Access to and quality of healthcare is similarly dismal when it comes to diversity, starkly cutting across racial, social, and economic divides.

If developed and used sensitively, artificial intelligence systems could go a long way to mitigating these inequalities by removing human bias. A careless approach, however, could make the situation worse.

AI has the potential to revolutionize healthcare, ushering in an age of personalized, accessible, and lower-cost medicine for all. But theres also a very real risk that those same technologies will perpetuate existing healthcare inequalities. A large part of this risk comes from existing biases in healthcare data.

AIs transformative potential comes from its ability to interrogate, parse, and analyze vast amounts of data. From this information, AI systems can find patterns and links that would have previously required great levels of expertise or time from human doctors. For this reason, AI is particularly useful in diagnostics, creating personalized treatment plans, and even helping doctors keep up to date with the latest medical research.

If we want to use AI to facilitate a more personalized medicine for all, it would help if we could first provide medicine that works for half the population.But this use of data risks exacerbating existing inequalities. Data coming from randomized control trials are often riddled with bias. The highly selective nature of trials systemically disfavor women, the elderly, and those with additional medical conditions to the ones being studied; pregnant women are often excluded entirely. AIs are trained to make decisions using this skewed data, and their results will therefore favor the biases contained within. This is especially concerning when it comes to medical data, which weighs heavily in the favor of white men.

The consequences of this oversight are pernicious. Women are far more likely to suffer the deleterious side effects of medication than men. Pregnant women get sick, but the consequences of taking many medications when pregnant are chronically understudied, or worse yet, unknown entirely. Women are far less likely to receive the correct treatment for heart attacks because their symptoms do not match typical (read: male) symptoms.

If evidence-based medicine is already far less evidence-based for anybody who is not a white male, how can the use of this unmodified data do anything other than unwittingly perpetuate this inequality? If we want to use AI to facilitate a more personalized medicine for all, it would help if we could first provide medicine that works for half the population.

The effects of this data can be even more insidious. AI systems often function as black boxes, which means technologists are unaware of how an AI came to its conclusion. This can make it particularly hard to identify any inequality, bias, or discrimination feeding into a particular decision. The inability to access the medical data upon which a system was trainedfor reasons of protecting patients privacy or the data not being in the public domainexacerbates this. Even if you had access to that data, the often proprietary nature of AI systems means interrogation would likely be impossible. By masking these sources of bias, an AI system could consolidate and deepen the already systemic inequalities in healthcare, all while making them harder to notice and challenge. Invariably, the result of this will be a system of medicine that is unfairly stacked against certain members of society.

This is especially true of less-connected communities. There is already an unhealthy digital divide where poorer and older members of society dont have access to the digital technologies that can be used to improve healthcare. This also means theyre not producing the data that comes with its use, and as this chasm grows, the system will stack against older and poorer patients even further than it currently does. Even if they were to readily gain access to these technologies in the next decade, it would be too late, as the systems will already be calibrated for younger, more urban bodies.

If we dont closely monitor AIs use in healthcare, theres a risk it will perpetuate existing biases and inequalities by building systems with data that systemically fails to account for anyone who is not white and male. At its core, this is not a problem with AI, but a broader problem with medical research and healthcare inequalities as a whole. But if these biases arent accounted for in future technological models, we will continue to build an even more uneven healthcare system than what we have today.

You can follow Rob on Twitter. Learn how to write for Quartz Ideas. We welcome your comments at ideas@qz.com.

Continue reading here:

If you're not a white male, artificial intelligence's use in healthcare could be dangerous - Quartz