Big Data Archives - Business Intelligence Info

Category Archives: Big Data

Intel and partners are using computer vision to help save Antarctica’s penguins from extinction

January 20, 2020 Big Data

Antarctica’s emperor penguin population has suffered such severe climate-related breeding issues that it’s at risk of disappearing by the year 2100, according to a 2019 study by the British Antarctic Survey. In search of a solution, an Intel-led group of tech companies developed a computer vision solution to help ecologists count the remaining penguins faster and more accurately than before.

The coalition consists of Intel’s AI Builders — the company’s curated ecosystem of software vendors — along with Microsoft’s AI for Earth initiative and data science consultancy Gramener. The consultancy sourced a corpus containing photos of Antarctica’s penguin colonies from Oxford University’s Penguin Watch Project, which over the past decade has sourced millions of time-lapse images from camera traps in over 40 locations and recruited online volunteers to annotate them. It then fed the data through a convolutional neural network — a type of AI model most commonly applied to analyzing visual imagery — that preserved spatial information while localizing penguin counts and estimating overall tallies.

The AI model was cascaded and two-stage in design. The first classified images into broad image categories, while the second estimated the density by taking inputs from earlier stages to generate maps. By approximating the number of penguins in clusters of different sizes, this density-based approach managed to handle most of the imaging challenges.

Intel Xeon Scalable processors were used to train the model within a virtual environment on Microsoft’s Azure cloud platform, which was then benchmarked using Intel’s Optimization for PyTorch toolkit. From this benchmarking exercise and other open source tools and libraries, Gramener says it was able to reduce training time from days to hours.

It was quite the feat, all told. Some images exhibited perspective distortion, where penguin faces in the foreground appeared larger than those in the background, and occlusion, where penguins in the background were hidden from view by objects or other penguins in the foreground. There was variability with respect to image quality and shooting angles, owing to unpredictable weather conditions, and there were often differences within the same picture, with some portions containing unusually high clusters of penguins.

Intel and partners are using computer vision to help save Antarctica’s penguins from extinction

Above: The AI system picks out penguins in a crowd.

Image Credit: Intel

“We faced multiple challenges. Almost 70% of images were unusable, as either the lighting wasn’t good or it was foggy, and penguins weren’t visible,” said lead data scientist Soumya Ranjan Mohanty. “These are easy for the human eye to perceive but difficult for an algorithm.”

It isn’t Hyderabad-based Gamener’s first AI project of an ecological nature. The consultancy partnered with Microsoft to build a model capable of classifying thousands of different flora and fauna species, using open source data from iNaturalist. It also developed a two-step solution that weeds out unusable images from camera traps and identifies those that contain animals, which became Microsoft’s Camera Trap API, as well as a system that provides high-resolution land cover information. Separately, Gamener worked with the Nisqually River Foundation to pick out fish species by analyzing video footage from an underwater camera that’s activated when fish pass certain infrared sensors. And in Kenya it worked with local researchers to build a model that could distinguish elephants from other livestock in aerial photographs.

Of course, Gamener is far from the first to apply AI to these kinds of problems. In December 2019, Google partnered with Conservation International and other organizations to process one of the world’s largest and most diverse databases of photographs taken from motion-activated cameras. DeepMind last year detailed ecological research its science team is conducting to develop AI systems that will help study the behavior of animal species in Tanzania’s Serengeti National Park. Microsoft recently highlighted a Santa Cruz, California-based startup called Conservation Metrics that’s leveraging machine learning to track African elephants, akin to a separate effort by an independent team of researchers to develop an algorithm trained on Snapshot Serengeti that can identify, describe, and count wildlife with 96.6% accuracy. And Intel’s own TrailGuard AI system prevents poaching by detecting motion with cameras using an on-device AI algorithm.

Let’s block ads! (Why?)

Big Data – VentureBeat

From Washington state to Washington DC, lawmakers rush to regulate facial recognition

January 19, 2020 Big Data

Amid the start of an impeachment trial; talk of mounting hostility with Iran; new trade deals with China, Canada, and Mexico; and the final presidential debate before the start of the Democratic presidential primary season, you might’ve missed it, but it was also a momentous week for facial recognition regulation.

A bipartisan group in Congress wants action, roughly a dozen state governments are considering legislation, and overseas news broke Thursday that the European Commission is considering a five-year moratorium on facial recognition among potential next steps. This would make the EU the largest government worldwide to halt deployment of the technology.

In Washington, DC this week, the House Oversight and Reform Committee pledged to introduce legislation in the “very near future” that could regulate facial recognition use by law enforcement agencies in the US. Just like in hearings held last summer, members of Congress exhibited a fairly unified, bipartisan position that facial recognition use by the government should be regulated and in some cases limited. There was talk of regulation, but until this week, the future of sweeping facial recognition regulation seemed uncertain.

Congress on Civil Rights, the Constitution, and facial recognition

Lawmakers seem to have a sense of urgency to take action for a variety of reasons, including a lack of standards for businesses; governments; and local, state, and federal law enforcement.

One major area of focus: violation of the 1st amendment right to freedom of assembly, and the idea that facial recognition might be used to identify people at political rallies or track political dissidents at protests.

“It doesn’t matter if it’s a President Trump rally or a Bernie Sanders rally, the idea of American citizens being tracked and cataloged for merely showing their faces in public is deeply troubling,” Jordan said.

“The urgent issue we must tackle is reigning in the government’s unchecked use of this technology when it impairs our freedoms and our liberties. Our late chairman Elijah Cummings became concerned about government use of facial recognition technology after learning it was used to surveil protests in his district related to Freddie Gray. He saw this as a deeply inappropriate encroachment on the freedom of speech and association, and I couldn’t agree more,” Jordan said.

Another reason lawmakers are anxious to regulate facial recognition: Civil rights protections and the great potential for racial discrimination.

Analysis by the Department of Commerce’s National Institute for Standards and Technology (NIST) last month found that some facial recognition systems are anywhere from 10 to 100 times more likely to misidentify groups like the young, elderly, women of color, and people of Asian or African descent.

Facial recognition systems that exhibit discriminatory performance, lawmakers contend, can exacerbate existing prejudices and overpolicing of schools and communities of color.

NIST’s analysis follows studies in 2018 and 2019 by AI researchers that found misidentification issues for popular facial recognition systems like Amazon’s Rekognition. Amazon has not agreed to have its AI analyzed by NIST, director Dr. Charles Romine told Congress this week.

Romine said talks between NIST and Amazon are ongoing on the subject of Rekognition review by the federal government.

If any one company that seemed to be a main source of ire for the committee, it’s Amazon.

Amazon lobbied members of Congress on the subject and has stated a willingness to sell its facial recognition to any government agency. Amazon reportedly marketed Rekognition to ICE officials, but the extent to which facial recognition is sold to government agencies is still unknown. In a shareholder vote last summer, Amazon chose to continue to sell facial recognition services to governments.

One factor that lawmakers say motivates a sense of urgency: China. In the EU, Washington DC, and on the state and local level across the U.S., lawmakers frequently cite China’s use of facial recognition to strengthen an authoritarian state as a future they want to avoid.

Meredith Whittaker is cofounder of the AI Now Institute cofounder and former Google employee. In testimony earlier this week she talked about how facial recognition is often used by those in power to monitor those without power and described the difference between usage in the U.S. and China. Last month, the AI Now Institute called for a ban of business and government use of facial recognition technology.

“I think it is a model for authoritarian social control that is backstopped by extraordinarily powerful technology,” Whittaker said about China’s use of facial recognition software. “I think one of the differences between China and the U.S. is that there, the technology is announced as state policy. In the U.S., this is primarily corporate technology that is being secretly threaded through our core infrastructures without that kind of acknowledgment.”

Washington state’s impact on facial recognition regulation

A handful of cities put facial recognition bans and moratoriums in place in 2019, but state legislatures in 2020 are already moving even faster to regulate the technology. Since the start of the years, 10 state legislatures introduced bills to regulate the use of facial recognition software, according to the Georgetown University Law School Center on Privacy and Technology.

In the state of Washington, the stakes may be unlike anywhere else in the world. Legislation to regulate use of facial recognition in Washington can be particularly influential, since the Seattle area is home to Amazon and Microsoft, two of the largest companies selling facial recognition software to governments. Axon, maker of police body cameras and video cloud storage provider, is also in Washington.

In Washington, state lawmakers this week started a second attempt to pass the Washington State Privacy Act. Known as SB 6281, the bill would regulate data privacy law and require “meaningful human review” of facial recognition results when used by the private sector. In a press conference Monday, the bill’s chief sponsor, Sen. Reuven Carlyle, said Washington is moving forward because there isn’t time to wait for lawmakers in Washington DC to deliver privacy regulation to reign in business use of private data. Carlyle said the bill takes cues from CCPA in California and GDPR in Europe.

A different version of the Washington Privacy Act passed the Washington State Senate with a near unanimous vote in spring 2019 but died in the Washington State House of Representatives.

Lawmakers complained last spring that the legislative process was tainted by lobbying from tech companies like Microsoft, and that companies like Amazon and Microsoft played too much of a role in drafting the 2019 version of the Washington Privacy Act.

Also introduced this week in Washington is SB 6280, a bill to regulate government use of facial recognition. The bill’s chief sponsor, State Senator Joe Nguyen, is a senior program manager at Microsoft, according to his LinkedIn profile. Nguyen is also a cosponsor of the Washington Privacy Act.

Microsoft initially supported the Washington Privacy Act last year but came to oppose amendments to the bill, calling them too restrictive. Microsoft also opposed a moratorium proposed by the ACLU, one of the first such moratoriums to be considered by any state legislature.

Jevan Hutson leads facial recognition and AI policy at the University of Washington School of Tech and Public Policy Clinic. He testified in multiple hearings in Olympia, Washington this week in favor of HB 2363, a bill that would make biometric data the sole property of an individual, and in opposition to the latest iteration of the Washington Privacy Act.

He also introduced a bill known as the AI Profiling Act. The legislation, which he drafted with others at the University of Washington, would outlaw the use of AI to profile people in public places; in important decision-making processes for a number of industries; and to predict a person’s religious affiliation, political affiliation, immigration status, or employability.

His position is that facial recognition may have some legitimate use cases, but it’s also a perfect surveillance tool, and that the root cause motivating facial recognition supporters is to create a new, invasive surveillance capitalism-driven marketplace.

Like last year, he believes the permissive regulatory framework found in the Washington Privacy Act that rejects the idea of a moratorium comes about due to the outsized influence of technology companies in Washington State that stand to profit from the widespread deployment of facial recognition.

He views Microsoft’s involvement and lobbying in 2019, and again in 2020, as an effort to create an initial framework of what facial recognition regulation should look like so they can bring that model to other states and Washington DC.

While speaking at Seattle University last year, Microsoft president Brad Smith said legislation passed in Washington could go on to shape facial recognition policy around the world.

As lawmakers in favor of the bill lay the necessary groundwork to attempt to pass the bill for a second time, politicians and advocates like Hutson argue legislation should take into account the demonstrated harm facial recognition can do and reject the idea that widespread use of facial recognition is inevitable.

“I think legislators and advocates here are seriously concerned and recognize that we need to get out front,” he said. “I think that sort of gets to the question of why now; it’s so important that we act because things will be bought by governments, and businesses will begin to deploy these things if there is not a clear sign from regulators and legislators both at the federal and local level to say, ‘No, this is not a valid market given the dangers that it poses.’”

Hutson also calls action in the near future important in order to shut down the idea that stifling innovation, a common argument against regulation, is always a bad thing. Facial recognition is being used for payments and to arrest people accused of crimes in China, but it’s also being used to track or imprison ethnic minorities, a use case he says could also be considered innovative.

“Innovation in many ways is this sort of false religion right where it’s like innovation in and of itself is a perfect good, and it’s not,” he said. “This is innovation worth stifling. Like I don’t think we should be super innovative with nuclear weapons. We don’t need even more innovative forms of oppression to be legitimized and authorized by the state legislature.”

Finishing thoughts

As laws get hammered out, stories of outrage continue.

In recent days, in Denver, where the city council is considering a facial recognition ban, advocates demonstrated that all nine members of the council met a 92% accuracy rate as people on the local sex offender registry.

There’s also the story Clearview AI, a startup that allows people to upload an image of a person then find where else on the web that person’s images appeared.

“It’s creepy what they’re doing, but there will be many more of these companies. Absent a very strong federal privacy bill, we’re all screwed,” Stanford University professor Al Gidari told the New York Times.

Steps taken this week to introduce legislation are just the beginning.

Bills and regulations continue to percolate through state legislatures and the halls of Congress, and as they do, the string of stories that refresh outage and brought about the initial sense of urgency seem likely to continue.

Let’s block ads! (Why?)

Big Data – VentureBeat

Baidu details its adversarial toolbox for testing robustness of AI models

January 18, 2020 Big Data

No matter the claimed robustness of AI and machine learning systems in production, none are immune to adversarial attacks, or techniques that attempt to fool algorithms through malicious input. It’s been shown that generating even small perturbations on images can fool the best of classifiers with high probability. And that’s problematic considering the wide proliferation of the “AI as a service” business model, where companies like Amazon, Google, Microsoft, Clarifai, and others have made systems that might be vulnerable to attack available to end users.

Researchers at tech giant Baidu propose a partial solution in a recent paper published on Arxiv.org: Advbox. They describe it as an open source toolbox for generating adversarial examples, and they say it’s able to fool models in frameworks like Facebook’s PyTorch and Caffe2, MxNet, Keras, Google’s TensorFlow, and Baidu’s own PaddlePaddle.

While the Advbox itself isn’t new — the initial release was over a year ago — the paper dives into revealing technical detail.

AdvBox is based on Python, and it implements several common attacks that perform searches for adversarial samples. Each attack method uses a distance measure to quantify the size of adversarial perturbation, while a sub-model — Perceptron, which supports image classification and object detection models as well as cloud APIs — evaluates the robustness of a model to noise, blurring, brightness adjustments, rotations, and more.

AdvBox ships with tools for testing detection models susceptible to so-called adversarial t-shirts or facial recognition attacks. Plus, it offers access to Baidu’s cloud-hosted deepfakes detection service via an included Python script.

“Small and often imperceptible perturbations to [input] are sufficient to fool the most powerful [AI],” wrote the coauthors. “Compared to previous work, our platform supports black box attacks … as well as more attack scenarios.”

Baidu isn’t the only company publishing resources designed to help data scientists defend from adversarial attacks. Last year, IBM and MIT released a metric for estimating the robustness of machine learning and AI algorithms called Cross Lipschitz Extreme Value for Network Robustness, or CLEVER for short. And in April, IBM announced a developer kit called the Adversarial Robustness Toolbox, which includes code for measuring model vulnerability and suggests methods for protecting against runtime manipulation. Separately, researchers at the University of Tübingen in Germany created Foolbox, a Python library for generating over 20 different attacks against TensorFlow, Keras, and other frameworks.

But much work remains to be done. According to Jamal Atif, a professor at the Université Paris-Dauphine, the most effective defense strategy in the image classification domain — augmenting a group of photos with examples of adversarial images — at best has gotten accuracy back up to only 45%. “This is state of the art,” he said during an address in Paris at the annual France is AI conference hosted by France Digitale. “We just do not have a powerful defense strategy.”

Let’s block ads! (Why?)

Big Data – VentureBeat

LinkedIn is using AI to spot and remove inappropriate user accounts

January 17, 2020 Big Data

Social networks including Facebook, Twitter, and Pinterest tap AI and machine learning systems to detect and remove abusive content, as does LinkedIn. The Microsoft-owned platform — which has over 660 million users, 303 million of whom are active monthly — today detailed its approach to handling profiles containing inappropriate content, which ranges from profanity to advertisements for illegal services.

As software engineer Daniel Gorham explained in a blog post, LinkedIn initially relied on a block list — a set of human-curated words and phrases that ran afoul of its Terms of Service and Community Guidelines — to identify and remove potentially fraudulent accounts. However, maintaining it required a significant amount of engineering effort, and the list tended to handle context rather poorly. (For instance, while the word “escort” was sometimes associated with prostitution, it was also used in contexts like a “security escort” or “medical escort.”)

This motivated LinkedIn to adopt a machine learning approach involving a convolutional neural network — a class of algorithm commonly applied to imagery analysis — trained on public member profile content. The content in question contained accounts labeled as either “inappropriate” or “appropriate,” where the former comprised accounts removed due to inappropriate content as spotted using the block list and a manual review. Gorham notes that only a “very small” portion of accounts have every been restricted in this way, which necessitated downsampling from the entire LinkedIn member base to obtain the “appropriate” labeled accounts and prevent algorithmic bias.

To further tamp down on bias, LinkedIn identified problematic words responsible for high levels of false positives and sampled appropriate accounts from the member base containing these words. The accounts were then manually labeled and added to the training set, after which the model was trained and deployed in production.

Gorham says the abusive account detector scores new accounts daily, and that it was run on the existing member base to identify old accounts containing inappropriate content. Going forward, LinkedIn intends to use Microsoft translation services to ensure consistent performance across all languages, and to refine and expand the training set to increase the scope of content it is able to identify with the model.

“Detecting and preventing abuse on LinkedIn is an ongoing effort requiring extensive collaboration between multiple teams,” wrote Gorham. “Finding and removing profiles with inappropriate content in an effective, scalable manner is one way we’re constantly working to provide a safe and professional platform.”

LinkedIn’s uses of AI extend beyond abusive content detection. In October 2019, it pulled back the curtains on a model that automatically generates text descriptions for images uploaded to LinkedIn, achieved using Microsoft’s Cognitive Services platform and a unique LinkedIn-derived data set. Separately, its Recommended Candidates feature learns the hiring criteria for a given role and automatically surfaces relevant candidates in a dedicated tab. And its AI-driven search engine leverages data such as the kinds of things people post on their profiles and the searches that candidates perform to produce predictions for best-fit jobs and job seekers.

Let’s block ads! (Why?)

Big Data – VentureBeat

Meet the new twist on data encryption that promises better privacy and security for AI

January 16, 2020 Big Data

Presented by Intel

AI and privacy needn’t be mutually exclusive. After a decade in the labs, homomorphic encryption (HE) is emerging as a top way to help protect data privacy in machine learning (ML) and cloud computing. It’s a timely breakthrough: Data from ML is doubling yearly. At the same time, concern about related data privacy and security is growing among industry, professionals and the public.

“It doesn’t have to be a zero-sum game,” says Casimir Wierzynski, senior director, office of the CTO, AI Products Group at Intel. HE allows AI computation on encrypted data, enabling data scientists and researchers to gain valuable insights without decrypting the underlying data or models. This is particularly important for sensitive medical, financial, and customer data.

Meet the new twist on data encryption that promises better privacy and security for AI

Above: Casimir Wierzynski, Senior Director, Office of the CTO, AI Products Group at Intel

Wierzynski leads Intel’s privacy preserving machine learning efforts, including HE work and in developing industry standards for the technology’s use. In this interview, he talks about why HE is needed, its place as a building block of improved privacy, and how the breakthrough technology helps create new business opportunities and data bridges to previously “untrusted partners.”

What and why

Q: What’s does “homomorphic” mean?

In Greek, homo is the same, morphic is the shape. So it captures the idea that if you do encryption in the right way, you can transform ordinary numbers into encrypted numbers, then do the same computations you would do with regular numbers. Whatever you do in this encrypted domain has the same shape as what you’re doing in the regular domain. When you bring your results back, you decrypt back to ordinary numbers, and you’ll get the answer you wanted.

Q: What big problem is the technology is solving?

We live in amazing times, when we have all these new AI products and services, like being able to unlock your phone with your face, or systems that enable radiologists to detect diseases at earlier stages. These products rely on machine learning systems, which are fed and shaped by data that are very sensitive and personal. It’s important for us as an industry — and I would argue as a society — to figure out how we can keep unlocking all the power of AI while still helping protect the privacy of the underlying data. That’s the overarching problem.

Q: Why is HE creating buzz now?

The technique itself has been around for more than 20 years as a theoretical construct. The criticism has been, okay, you can operate on encrypted data, but it takes you a million times longer than using regular data. It was an academic curiosity. But in the last five years, and especially the last two years, there have been huge advances in the performance of these techniques. We’re not talking about a factor of a million anymore. It’s more like a factor of 10 to 100.

Q: Certainly, there’s no shortage of data privacy and protection technologies…

We’ve all gotten good about encrypting data at rest, when it’s on our hard drives, or sending data back and forth over encrypted channels. But when you’re dealing with ML and AI, at some point those data have to be operated on. You need to do some math on those data. And to do that you need to decrypt them. While you’re using the data, the data are decrypted, and that creates a potential issue. We work to provide better protection on both these fronts, anonymization and encryption.

Above: Simple model of secure inference with HE.

Q: In ML with multiple partners, trust is a big issue. How does HE help here?

Whenever you’re dealing with digital assets, you have this phenomenon: When you share a digital asset with another party, it’s completely equivalent to giving it to them, then trusting they’re not going to do something unintended with it.

Now add the fact that ML is fundamentally an operation that involves multiple stakeholders. For example, one entity owns the training data. Another entity owns data they want to do some inference on. They want to use ML service provided by a third entity. Further, they to use an ML model owned by another party. And they want to run all this on infrastructure from some supply chain.

With all these different parties, and because of the nature of digital data, all must trust each other in a complex way. This is becoming harder and harder to manage.

HE in action

Q: Can you give an example of homomorphic encryption at work?

Say you have a hospital doing a scan on a patient, working with a remote radiology service. The hospital encrypts the scan and sends it over to the radiologist. The radiologist does all the processing in the encrypted domain, so they never see the underlying data. The answer comes back, which is also encrypted, and finally the hospital decrypts it to learn the diagnosis.

Q: In the example above, where does HE actually happen? At the scanner?

It could live in multiple places. There are two major components: encryption and decryption. But then there’s also the actual processing of the encrypted data. Encryption typically happens where the sensitive data are first captured, for example, in a camera or edge device. Processing encrypted data happens wherever the AI system needs to operate on sensitive data, typically in a data center. And finally, decryption happens only at the point where you need to reveal the results to a trusted party.

Powerful with other emerging techs

Q: When you speak and write on HE, you also talk about adjacent technologies. Please explain briefly the roles these other building blocks play in preserving privacy.

There are several very promising and rapidly developing technologies that use tricks from cryptography and statistics to do operations on data that seem magical.

All these technologies can be further accelerated and enhanced by hardware security measures called trusted execution environments, such as Intel SGX.

New bridges to partners in AI and ML

Q: So what opportunities are created when these technologies are combined?

One of the questions we’ve been asking at Intel is, what would happen if you could enable ML in these multi-stakeholder operations between parties that don’t necessarily trust each other, like we described earlier? What would that enable?

You could have banks that normally are rivals. But they may decide it’s in their interest to cooperate around certain risks that they all face in common, like fraud and money laundering. They could pool their data to jointly build anti-money laundering models while keeping their sensitive customer data private.

Another example is in the retail sector. Retailers want to make the most out of the data that they’ve collected on shoppers, so they can personalize certain experiences. What if there were a way to enable that and still provide quantifiable protections around the privacy of that data?

New business opportunities

Q: Are there new revenue models and opportunities being created?

One thing that’s exciting about this area is that you start to enable new business models around data that would have previously been impossible. For example, (HE) can be a way to monetize data while helping maintain security. At the same time, it addresses one of the biggest problems in ML, namely, access to large, diverse data sets for building models. So you can imagine a whole ecosystem that brings together people who hold data with people who need data. And very importantly, it’s all done in a way that preserves security and privacy of the data. That’s an exciting possibility.

Q: How advanced is that concept in implementation? Any other or many real-life instances like that?

I would say it’s beyond theory, and in early stages of commercial deployment.

Data is becoming a much bigger asset class than ever before, sometimes in surprising ways. For example, in corporate bankruptcy cases, creditors can end up owning large data sets unrelated to banking in any way. They’re just leftovers of a loan that went sour. So they’re looking for a way to monetize these assets and provide them to data scientists who seek additional training data to make their AI systems more reliable, all while keeping the underlying data private and secure.

Or imagine a bunch of hospitals that have patient data. For various reasons, they can’t share it. But they know if they could, and could get lawyers in the room to hammer out an agreement, they would potentially be able to jointly build a model with much more statistical power than one any could build individually. Using privacy-preserving ML techniques, they can essentially form a consortium and say: “In exchange for all of us owning this improved model that will improve our patients’ outcomes, we’ll be a part of this consortium, and we can still keep all of our patient data private and secure.”

Key role of developers

Q: Where do developers fit in?

As it is now, if you’re an AI data scientist and you want to build a machine learning model that operates on encrypted data, you have to be some magical person, simultaneously an expert in software engineering and data science and post-quantum cryptography.

One of the major efforts that my team has been working on is making these technologies much more accessible and performant for the data science and developer communities so that they can scale up. This is a priority for Intel. Today we’re offering tools like The Intel HE transformer for nGraph, which is a Homomorphic Encryption (HE) backend to Intel’s graph compiler for Artificial Neural Networks.

Above: The nGraph-HE software stack supports the SEAL encryption library and underlying cryptosystems.

Standards and Intel

Q: Some people will ask: Why is Intel active here? The answer is…

For starters, homomorphic encryption is very compute-intensive. This is an area where Intel can really shine in terms of building optimized silicon to handle this fundamentally new way of computing.

But more broadly, those examples from health care, retail, and finance — sectors representing about a quarter of GDP in the United States. These are very economically important problems. At Intel, we are obsessed with helping customers solve their problems around data. And privacy is at the heart of any data-centric business.

We are in a unique position, because Intel works closely with hyperscale, data-centric users of hardware who are building all kinds of exciting AI applications. At the same time, we are a neutral party with respect to data. To make these technologies perform at scale is going to require the kinds of complex software and hardware co-design that Intel is uniquely positioned to provide. We get to collaborate actively with a fascinating range of players across industry.

Above: The 2019 HomoEncyption.org standards group meeting gathered top private and public sector participants from around the globe in Santa Clara, CA.

Q: Intel has taken a leadership role in developing HE standards. Why are they important? Status?

As with any crypto scheme, people will use it when there’s interoperability and trust in the underlying math and technology. We recognize that to really scale up, to bring all the homomorphic encryption goodness to the world, we need to have standards around it.

As interest in privacy preserving methods for machine learning grows, it’s essential for standards to be debated and agreed upon by the community, spanning business, government, and academia. So in August, Intel co-hosted an industry gathering of individuals from Microsoft, IBM, Google, and 15 other companies interested in this space.

We’ve identified many points of broad agreement. Now we’re trying to figure out the right approach to bring it to standards bodies, like ISO, IEEE, ITU, and others. They all have different structures and timelines. We’re strategizing on how best to move that forward.

Final word

Q: Any thoughts you’d like to leave ringing in people’s ears?

Privacy and security technologies for machine learning like homomorphic encryption are ready for prime time. These are not academic exercises anymore. These are real market-ready ideas. The day is coming when the idea of doing machine learning on someone else’s raw data will seem quite strange. We’re at the start of a new and exciting era where machine learning will enable us to explore new opportunities unlike anything before.

Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. Content produced by our editorial team is never influenced by advertisers or sponsors in any way. For more information, contact sales@venturebeat.com.

Let’s block ads! (Why?)

Big Data – VentureBeat

SenseTime researchers create a benchmark to test face forgery detectors

January 16, 2020 Big Data

Face swapping is a category of deepfakes that extracts the faces of people in existing media and replaces them with other peoples’ features, typically with AI and machine learning. It’s been popularized by apps like MixBooth and SnapChat, and while the underlying techniques have enabled sophisticated image editing for legitimate purposes, they’ve also given rise to concerns about potential misuse or abuse.

Various groups have compiled manipulated media to support the development of face swapping detection methods, but the samples that have released so far are relatively few in number or overly artificial. That’s why researchers from SenseTime Research, the R&D division of Hong Kong-based tech startup SenseTime, partnered with Nanyang Technological University in Singapore to design a new large-scale benchmark for face forgery detection. They call it DeeperForensics-1.0, and they say it’s the largest corpora of its kind with over 60,000 videos containing roughly 17.6 million frames.

According to the researchers, all source videos in DeeperForensics-1.0 were carefully selected for their quality and diversity. They’re ostensibly more realistic than those in other data sets in that they’re closer in kind to real-world detection scenarios, and in that they contain compression, blurriness, and transmission artifacts matching those found in the wild.

To build DeeperForensics-1.0, the researchers collected face data from 100 paid male and female actors of 26 different nationalities and ranging in age from 20 to 45, all of whom were instructed to turn their heads in nine lighting conditions and speak naturally with over 53 expressions. They ran these through an AI framework — DeepFake Variational AutoEncoder, or DF-VAE — using 1,000 YouTube videos as target videos, where each of the 100 actors’ faces was swapped onto 10 targets. And they deliberately distorted each video in 35 different ways to simulate real-world scenarios, such that the final data set contained 50,000 unmanipulated videos and 10,000 manipulated videos.

“We find that the source faces play a much more critical role than the target faces in building a high-quality data set,” wrote the researchers in a preprint paper detailing their work. “Specifically, the expressions, poses, and lighting conditions of source faces should be much richer in order to perform robust face swapping.”

The researchers also created what they call a “hidden” test set within DeeperForensics-1.0 — a set of 400 videos carefully selected to better imitate fake videos in real scenes. Curating the set involved collecting fake videos generated by unknown face-swapping methods and obscuring them with distortions commonly seen in real scenes, and subsequently choosing only videos that fooled at least 50 out of 100 human observers in a user study.

To evaluate the quality of DeeperForensics-1.0 compared with other publicly available data sets, the researchers tasked 100 experts in computer vision with ranking the the quality of a subset of videos contained within it. They report that DeeperForensics-1.0 came out ahead on average in terms of realism for its scale compared with FaceForensics++, Celeb-DF, and other popular deepfake detection corpora.

In future work, the research team intends to expand DeeperForensics gradually and work with the research community toward identifying evaluation metrics for face forgery detection methods.

The fight against deepfakes appears to be ramping up. Last summer, members of DARPA’s Media Forensics program tested a prototypical system that could automatically detect AI-generated videos in part by looking for cues like unnatural blinking. Startups like Truepic, which raised an $ 8 million funding round in July, are experimenting with deepfakes “detection-as-a-service” business models. And in December 2019, Facebook together with the Partnership on AI, Microsoft, and academics launched the Deepfake Detection Challenge, which will offer millions of dollars in grants and awards to spur the development of deepfake-detecting systems.

Let’s block ads! (Why?)

Big Data – VentureBeat

Fight for the Future launches campaign to keep facial recognition off U.S. college campuses

January 14, 2020 Big Data

Fight for the Future, the Boston-based nonprofit promoting causes related to copyright legislation, online privacy, and internet censorship, today announced that it will team up with advocacy group Students for Sensible Drug Policy in an effort to ban facial recognition on university campuses in the U.S. To kickstart the grassroots movement, the organizations this morning launched a website and an organizing toolkit for student groups.

The push is part of Fight for the Future’s broader Ban Facial Recognition campaign, which launched in July 2019 and calls on local, state, and federal lawmakers to prevent government and law enforcement use of facial recognition. While facial recognition isn’t widely deployed on U.S. campuses, Evan Greer, deputy director of Fight for the Future, asserts that it is likely to threaten privacy, civil liberties, and equity as companies increasingly market the tech to schools.

“Facial recognition surveillance spreading to college campuses would put students, faculty, and community members at risk. This type of invasive technology poses a profound threat to our basic liberties, civil rights, and academic freedom,” said Greer. “Schools that are already using this technology are conducting unethical experiments on their students. Students and staff have a right to know if their administrations are planning to implement biometric surveillance on campus … The data collected is vulnerable to hackers and in the wrong hands could be used to target and harm students. And it’s invasive, enabling anyone with access to the system to watch students’ movements; analyze facial expressions; [and] monitor who they talk to, what they do outside of class, and every move they make.”

Facial recognition found itself in the news last year more than perhaps any other application of AI.

A number of efforts to use facial recognition systems within schools have met with resistance from parents, students, alumni, community members, and lawmakers alike. The Lockport City School District in upstate New York abandoned plans to pilot components of a face-recording system after it was revealed that the district planned to flag suspended students. At the college level, a media firestorm erupted after it was revealed that a University of Colorado professor secretly photographed thousands of students, employees, and visitors on public sidewalks for a military anti-terrorism project, and after University of California San Diego researchers admitted to studying footage of students’ facial expressions to predict engagement levels.

Meanwhile, two studies spotlighted the types of biases to which facial recognition algorithms can become susceptible. In early January, a team of MIT scientists alleged that Amazon Web Services’ facial recognition and analysis platform — Rekognition — distinguished gender among certain ethnicities less accurately than did competing solutions. And last December, research published by the National Institute of Standards and Technology (NIST) found that a number of products falsely identified African and Asian faces 10-100 times more often than Caucasian faces.

Perhaps unsurprisingly, a growing number of activists, academics, and lawmakers have called for restrictions or outright bans on facial recognition technology. Last fall, California imposed a three-year moratorium on facial recognition use in law enforcement body cameras, preceding San Francisco’s ban on facial recognition use by police and city departments. Oakland followed suit in June, after which Berkeley passed a ban of its own. And in two House Oversight and Reform committee hearings last summer, some of the most prominent Republicans and Democrats in the U.S. Congress joined together on proposals for legislative reform, following the introduction of a bill that would require businesses to receive consent before using facial recognition.

“Contaminating learning environments with this technology will only harm those it claims to protect. It is untrue that facial recognition makes anyone safer,” said Erica Darragh, who sits on the board of Students for Sensible Drug Policy. She noted that groups at George Washington University and DePaul University have introduced student government resolutions to ban facial recognition. Darragh also said that a dozen major university administrations, including those at Stanford, Harvard, and Northwestern, have been contacted to clarify whether they are using the tech. “Students should not have to trade their right to privacy for an education, and no one should be forced to unwittingly participate in a surveillance program [that] will likely include problematic elements of law enforcement. This automation of racial and political profiling threatens everyone, especially students, faculty, and campus guests of color. Students have an obligation to prevent this technology going mainstream, beginning with university campuses, where we have the most power and we know how to win.”

Let’s block ads! (Why?)

Big Data – VentureBeat

Carnegie Mellon researchers propose AI that surfaces positive online comments

January 13, 2020 Big Data

Researchers at Carnegie Mellon’s Language Technologies Institute say they’ve developed a system that taps machine learning to analyze online comments and pick out those that defend or sympathize with disenfranchised peoples. Although it hasn’t been commercialized, they’ve used it in experiments to search for nearly a million YouTube comments, focusing on the Rohingya refugee crisis and the February 2019 Pulwama terrorist attack in Kashmir. And they hope it’ll form the foundation of a future system that’ll reduce the manual effort necessary to curate comments on publisher websites, social media, and elsewhere.

Tamping down on abusive online behavior is no easy feat, particularly considering the level of toxicity in some social circles. More than one in five respondents to a survey by the Anti-Defamation League, a nonprofit that tracks and fights anti-Semitism, reported having been subjected to threats of violence. And nearly one in five said they’d experienced sexual harassment or stalking and sustained harassment, while upwards of 20% said the harassment was the result of their gender identity, race, ethnicity, sexual orientation, religion, occupation, or disability.

As the researchers explain, improvements in AI language models — which learn from many examples to predict what words are likely to occur in a given sentence — made it possible to analyze such large quantities of text. The study’s contribution was a technique enabling those models to digest short texts originating from South Asia, which can be difficult to interpret because they often contain spelling and grammar mistakes and combine different languages and systems of writing.

Specifically, the researchers obtained embeddings — numerical representations of words — that revealed novel language groupings or clusters. Language models create these so that words with similar meanings are represented in the same way, making it possible to compute the proximity of a word to others in a comment or post.

The team reports that in experiments, their approach worked as well or better than commercially available solutions. Random samplings of the YouTube comments showed about 10% were positive, compared with the 88% found with the AI algorithm.

“Even if there’s lots of hateful content, we can still find positive comments,” said post-doctoral research Ashiqur R. KhudaBukhsh, a contributing author on aforthcoming paper describing the work. He’ll present his findings with coauthors Shriphani Palakodety and Jaime Carbonell at the Association for the Advancement of Artificial Intelligence annual conference next month in New York City.

The study follows the release of a data set by Jigsaw — the organization working under Google parent company Alphabet to tackle cyber bullying, censorship, disinformation, and other digital issues of the day — containing hundreds of thousands of comments and annotations with toxicity and identity labels. It’s intended to help measure bias in AI comment classification systems, which Jigsaw and others have historically measured using synthetic data from template sentences.

In a related development, researchers at the Georgia Institute of Technology and email marketing startup Mailchimp recently proposed RECAST, an interactive tool for examining toxicity detection models by visualizing explanations for predictions and providing alternative wordings for detected toxic speech. The plan to release an open source browser extension in the near future.

Let’s block ads! (Why?)

Big Data – VentureBeat

The White House’s new AI principles won’t solve regulatory problems

January 12, 2020 Big Data

This week, the White House released 10 “AI Principles,” intended as guidance for federal agencies while they consider how to appropriately regulate AI in the private sector. It’s an effort to help reduce the potential harms of AI that have been under scrutiny all over the world, while maintaining the benefits to society AI can bring. This is a moment the industry has been waiting for because of lingering uncertainty around how the US government will work to control this powerful technology and to ensure it doesn’t hurt people more than it helps.

The problem is, while it may be a good thing that the White House is taking an active role in the fight to regulate AI, their emphasis on light-touch regulation means the new rules fail to go far enough.

The principles themselves address some of the concerns raised by the AI ethics community and the academics who study the effects of technology on society. One such principle calls for lawmakers to consider whether the technology will “introduce real-world bias that produces discriminatory outcomes,” echoing the rallying cries of those academics who have warned for years that AI will codify existing societal biases into automated decision systems. These systems have been shown to adversely impact the most vulnerable people in society, including those marginalized by discrimination on the basis of race, gender, sexuality, and disability — and perhaps most alarmingly, our country’s poorest citizens. Unregulated algorithms can automate and thereby govern the human right to life in areas like healthcare, where flaws in algorithms have dictated that black patients receive inadequate care when compared to their white counterparts. In other cases, lawmakers suspect that algorithmic bias may perpetuate gender disparities in access to financial credit and employment.

The guidance also acknowledges that “current technical challenges in creating interpretable AI can make it difficult for agencies to ensure a level of transparency necessary” to foster public trust. It advises agencies to pursue transparency in two forms: disclosing when and where the technology is in use, and making the outcomes transparent enough to ensure that the algorithms comply, at least, with existing laws.

But the true extent of the harm AI does globally is often obscured due to trade secrets and government practices like the Glomar response — the infamous “I can neither confirm nor deny” line. Using these protective measures, entities can hide the extent and breadth of the AI-related programs and products they’re using. It’s entirely likely that many algorithms already in use violate existing anti-discrimination (among other kinds of) laws. In some cases, companies may even choose so-called “black box” model types that obscure the rationale behind decisions at scale, in order to claim ignorance and a lack of control over the actions that result. This legal loophole is possible because some types of AI are so complex that no human could ever truly understand the logic behind a particular decision, making it impossible to understand what happened if something goes wrong.

It is exactly this kind of behavior that has resulted in a massive loss of public trust in the technology industry today, and it’s further evidence that AI-specific regulation is severely needed to protect the public good. It’s been demonstrated time and again that, even with the best intentions, AI has the potential to hurt people in mass quantities, which makes our industry unique in the technology field overall. This incredible power to do harm at scale means those of us in the AI industry have a responsibility to put societal interest above profit. Too few companies currently embody this high degree of responsibility with their actions. It’s of utmost importance that we reverse this trend, or society will never enjoy the benefits that AI promises.

Bias mitigation, public disclosure, and a solution to the problematic “black box” are table stakes for any sufficiently effective regulatory framework for AI. But these “AI Principles” fall woefully short in their attempt to optimize the balance of societal good versus any potential dangers that the technology might one day bring.

In a surprise twist, with the first federal document to address AI lawmaking, the Trump administration focused mainly on the risks of losing out on great power rivalry, market competition, and economic growth. In doing so, the administration dramatically underestimates the ongoing harm facing Americans today, once again sacrificing the wellbeing of the public for unchecked, unregulated industry growth.

Importantly, although this is the first guidance to emerge from the federal government, many cities and states have already had success governing AI, where similar, comprehensive federal bills have stalled and ultimately failed due to congressional deadlock. Several cities have banned intrusive facial recognition practices from use by law enforcement, with many more algorithmically-centered proposals under consideration at the state and city levels.

It’s telling that the new “AI Principles” warn of regulatory “overreach” in one breath while undermining local legislative authority in another. The guidance advises that agencies may use “their authority to address inconsistent, burdensome, and duplicative State laws.” This language subtly indicates to lawmakers that a practice known as federal preemption could be used to undo some of the strong, grassroots, and broadly celebrated local regulations that have been championed by AI experts and civil liberties advocates like the ACLU.

Even more concerning is the fact that these strong local laws are the result of the public democratic will expressed in pockets of the country where technical work is most common (San Francisco, CA; Sommerville, MA [close to MIT]; and a likely proposal in Seattle, WA). These new local laws were enacted as a response to the inherent risks of using predictive technology to gate access to sensitive services like public housing, proactive healthcare, financial credit, and employment, and to a lack of action from Washington. The people who build these technologies know that any algorithm threatening to perpetuate human bias or to provide a “math-washed” license to discriminate must be closely monitored for misbehavior, or never implemented at all.

These AI Principles may be a small step in the right direction, and broadly speaking, they can introduce a degree of enhanced responsibility if correctly implemented by lawmakers who are earnestly seeking to reduce risk. But they are only a starting point, and they actually threaten further harm by raising the issue of federal preemption to undo the incredible work that’s already being done by local legislators. Industry workers with direct knowledge of the benefits and risks of AI have often been the strongest voices in the call for strict regulation, and the White House should take steps to better align its policies with the advice of those working hardest to bring AI to market.

Liz O’Sullivan is the cofounder of ArthurAI and technology director of STOP (Surveillance Technology Oversight Project)

Let’s block ads! (Why?)

Big Data – VentureBeat

ProBeat: A smart speaker does not a smart home make

January 11, 2020 Big Data

To coincide with CES 2020 this week, Strategy Analytics released a survey claiming that smart homes had passed “the tipping point” and that “Most Homes Are Now Smart Homes.” Putting aside that U.S. homes don’t represent the world’s homes, the survey assumed that you had a smart home if you “own at least one smart home device.” First, owning doesn’t mean actively using. Second, a smart home is not defined by owning “smart speakers, interactive security systems, and smart thermostats.” Third, that begs the question: How should we define a smart home?

The smart home definition is important for a few reasons. Technology has been flooding into our homes, and CES 2020 showed this trend is not abating. Internet of things (IoT) devices are not going away, and neither is the dream of a smart home. That doesn’t mean all home devices marketed as “smart” are inherently useful, however, let alone make our homes “smart.” Additionally, there are plenty of valid privacy and security concerns in a potential smart home.

So, let’s try to make sense of all these devices.

IoT, AI, and smart

First and foremost, an IoT device is not a smart device. If your lightbulb, your thermostat, or your fridge is connected to the internet, that doesn’t mean you suddenly have a smart home. Think of it this way: If all you can do with your fancy new light bulb is turn it on and off from your phone, rather than a light switch, that’s not smart. Thermostats are a better example. If you can lower the temperature of your home after you’ve already left for vacation, that’s not smart. That’s an IoT device, and it’s neat, but it’s not smart. If, however, your thermostat can save your home energy and you money by making adjustments based on weather forecasting, occupancy patterns, humidity, and electricity costs, that’s smart.

Next — and this is what really caught my eye in this survey — a smart speaker doesn’t make a smart home. Your smart speaker needs to be doing more for your home than a virtual assistant can do from your phone. If you’re merely asking Alexa for the weather, or Google Assistant to play some music, that’s not much of a smart home.

(Speaking of, Google had a much more muted presence at CES this year. At CES 2019, the company built a freaking massive, rhyming, catchy, musical of a ride for its Google Assistant announcements. At CES 2020, the company dropped its Google Assistant news the old-fashioned way. What can you expect from Google Assistant this year? Among other things, you guessed it: deeper smart home integration.)

I would further argue that artificial intelligence alone is not a good measure of whether a device is contributing to a smart home. What matters is what the AI is enabling.

Pushing the limits of what’s possible

If your smart speaker is controlling your alarm, lights, blinds, fridge, and toaster, we’re getting somewhere. But there’s a difference between you telling your smart speaker to set an alarm, which you can do yourself with an alarm clock, and having it automatically turn off the alarm when you’ve gotten out of bed — something your alarm clock can’t do. If your Google Assistant can lower your blinds automatically when the sun goes down and raise them when your alarm goes off, that’s the inklings of a smart home.

Put another way, AI has to be doing something in the home that wasn’t possible before, or at least not reasonably possible. Sure, I could rig my blinds to move at certain times of the day and set complicated schedules on my thermostat, but neither is optimal.

A smart home contains devices that understand your home, its surroundings, and its inhabitants. And they actively make the home smarter. To be fair, that’s much harder to formulate a survey around.

One final note on the privacy and security front: I suspect that we may one day see smart home devices that are not connected to the internet. First we got IoT devices for the home. Now we’re starting to see smart IoT devices for the home. One day we will get smart devices for the home with built-in offline machine learning models. They will do their processing on-device without requiring a constant connection to the cloud. And that will be the real tipping point.

ProBeat is a column in which Emil rants about whatever crosses him that week.

Let’s block ads! (Why?)

Big Data – VentureBeat

Category Archives: Big Data

Intel and partners are using computer vision to help save Antarctica’s penguins from extinction

From Washington state to Washington DC, lawmakers rush to regulate facial recognition

Congress on Civil Rights, the Constitution, and facial recognition

Washington state’s impact on facial recognition regulation

Finishing thoughts

Baidu details its adversarial toolbox for testing robustness of AI models

LinkedIn is using AI to spot and remove inappropriate user accounts

Meet the new twist on data encryption that promises better privacy and security for AI

What and why

HE in action

Powerful with other emerging techs

New bridges to partners in AI and ML

New business opportunities

Key role of developers

Standards and Intel

Final word

SenseTime researchers create a benchmark to test face forgery detectors

Fight for the Future launches campaign to keep facial recognition off U.S. college campuses

Carnegie Mellon researchers propose AI that surfaces positive online comments

The White House’s new AI principles won’t solve regulatory problems

ProBeat: A smart speaker does not a smart home make

IoT, AI, and smart

Pushing the limits of what’s possible

Recent Posts

Categories

Archives