Have you ever thought about an AI-based machine playing poker with you? If your imagination has gone that wild then Facebook is all set to make it a reality with its new general AI framework called Recursive Belief-based Learning (ReBeL) that can even perform better than humans in poker and with little domain knowledge as compared to the previous poker setups made with AI.

With ReBel, Facebook is also going for multi-agent interactions – which means that the general algorithms will soon have the capacity to be deployed on a large scale and for multi-agent settings as well. The potential applications include workings like auction, negotiations, and cybersecurity or the operation of self-driving cars and trucks.

Facebook’s plan of combining reinforcement learning with search for AI model training can lead to some remarkable advancements. This is because Reinforcement Learning is based on agents learning to achieve goals in order to maximize rewards whereas search is basically defined as a process that starts from the plan to the stage of setting the goal.

One such example is of Deepmind’s Alpha Zero that is based on a similar program to deliver state-of-the-art performance in board games like chess, shogi, and Go. However, the combination falls short when it is being applied for games like poker because of imperfect information that can arise as a result of how the situation in the game changes. Actions then take help from probability or the playing strategy.

Hence, proposing a solution to the problem in the form of ReBel, Facebook researchers have now expanded the notion of “game state” while including the agent’s belief which relies on the state they are in while playing – counting the common knowledge and policies of other players as well.

When working, ReBel trains two AI models; one is of a value network and the other is of policy network. There is reinforcement learning happening with search during the self-play which eventually has resulted into a flexible algorithm that now holds the potential to beat human players.

For a high level, ReBel operates with public belief states rather than going for world states. If that has surprised you then public belief states are there to generalize the notion of “state value” in games with imperfect information like Poker. PBS is also more often regarded as a common-knowledge probability distribution over a limited arrangement of possible actions and states, which we sometimes call history as well.

Now in perfect-information games, PBS can be distilled down to histories just like the way it distills down to world states in two-player zero-sum games. Not to forget that a PBS is actually the decisions that a player can and also the outcomes of the possibilities on one hand.

As soon as ReBel starts to work for every new game, it creates a “subgame” in the beginning which is very much similar to the original one, except for the fact that its roots go back to the initial PBS. The algorithm actually wins by repeating the runtime of “equilibrium-finding” algorithm and then take advantage of the trained value network to create estimates on every stage of the iteration. Furthermore, with enforcement learning, the values come out easily and then added back to the network as training examples. The policies in the “subgame” are also added as examples. The process continues to repeat itself until PBS becomes the new subgame root and completes a certain accuracy threshold.

The researchers also benchmarked ReBel, as a part of the experiment, for games of heads-up no-limit Texas hold’em poker, Liar’s Dice, and turn endgame hold’em. They used 128 PCs with eight graphic cards only to generate the stimulated game data and of course place random bets and stack sizes (ranging from 5000 to 25000 chips) to test its abilities.

ReBel was also trained on a game with one of the best heads up poker players in the world Don Kim and the results turned out to be ReBel playing faster than two seconds per hand across 7,500 hands and how it didn’t take more than 5 seconds for any decision. Overall ReBel scored 165 thousandths – which is a pretty good result when compared to the previous poker playing system by the social media giant Libratus that resulted in 147 thousandths.

To prevent cheating, Facebook has decided that they will not release ReBel’s codebase for Poker. The company only open-sourced Liar Dice’s implementation, which according to researchers is easier to understand and adjust.

The post Facebook’s New Algorithm Can Play Poker And Beat Humans At It appeared first on Artificial Intelligence.

The Cloud Native Computing Foundation (CNCF) has accepted a Litmus Chaos application testing tool based on chaos engineering principles as a sandbox level project.

Developed by MayaData, the open source software provides IT teams with a chaos engineering tool that runs natively on Kubernetes.

MayaData COO Uma Makkara says Litmus Chaos was originally created to test OpenEBS, an open source project that makes it easier to access container-attached storage. OpenEBS is the foundation for Kubera, a data management platform the company launched last week.

While there are other tools and services based on the chaos engineering principle, Makkara notes Litmus Chaos makes it possible to run those tests on the same Kubernetes cluster where a microservices-based application built using containers runs.

In theory, microservices-based applications are expected to degrade gracefully in the event a software component or infrastructure element suddenly becomes unavailable. Requests for access to microservices are rerouted to ensure availability. In practice, many microservices-based applications inadvertently have single points of failure. Chaos engineering techniques randomly remove software components and infrastructure elements to test how resilient a microservices-based application really is.

MayaData is contributing Litmus Chaos to the CNCF in the expectation that other organizations building microservices-based applications will want to apply similar techniques to testing their software. Some organizations will prefer to deploy those tools themselves while others may prefer to employ a service that might be constructed on top of Litmus Chaos.

Makkara says chaos engineering tests are also shareable via a Chaos Hub, which eventually will become part of a series of integrated hubs revolving around the Harbor container registry, which is designed to hold multiple types of artifacts such as Helm Charts alongside containers.

Containers make it possible to package code application logic in a more discrete fashion to create a microservice. A microservice packages multiple containers with runtimes in a way that enables them to be more easily executed at the same time. Thanks to that capability, a development team can deploy microservices in a way that allows them to be managed in isolation from one another. Microservices have been around in various forms for some time. However, with the rise of containers, it has become easier to both construct and update microservices. Containers enable any element of a microservice or the entire microservice itself to be ripped and replaced.

The challenge is as applications expand the number of dependencies between microservices grows. Chaos engineering provides a way to discover whether those dependencies are loosely coupled or create a single point of failure that could result in an entire application becoming unavailable.

It’s not clear to what degree IT organizations have embraced chaos engineering techniques. However, microservices based on containers are likely to force the issue. Organizations that build cloud-native applications based on microservices and containers generally employ best DevOps practices that stress continuous testing. That’s critical in the context of any microservices-based application because what works fine today might not work near as well today as more dependencies are introduced.

The post MayaData Donates Chaos Engineering Tool for Kubernetes Apps to CNCF appeared first on Artificial Intelligence.

As soon as hackers take down your account, you normally get to see suspicious posts that might revolve around deals on products or stuff that you would never like buying online. But how about a situation where hackers plan to infiltrate the account of Facebook’s biggest data partners? Yes, we are going to talk about thousands of dollars and credit cards being stolen in a similar case.

Recently, hackers got access to the personal account of LiveRamp’s employee, only with the aim to get control over the Business Manager’s account and hoping to run scam through the ads with other’s money being spent on them.

By doing so, they successfully attacked one of Facebook’s most prominent data partners, however, the damage was still contained. The incident affected a limited number of LiveRamp customers and associated Ad Accounts, while Facebook actively informed the affected parties about it.

Although LiveRamp didn’t tell the exact number of customers who got affected by the hack and stated that the company has their security measures in place, especially for employees who deal with Facebook ads accounts, but there is one thing for sure that thousands of victim’s dollars were spent into tricking users buy fake products. Facebook, on the other hand, did confirm later in November that personal account of an admin for a Business Manager account but didn’t mention LiveRamp directly.

Nevertheless, LiveRamp and Facebook worked together to cut down unauthorized access and restore the functionality back to normal for its users.

This isn’t the first time that hackers targeted the hub of Facebook’s empire – the advertisers. As advertising has been Facebook’s lifeline for a long period of time — considering how it is expected to add up $84 billion in revenue in 2020 with 2.2 billion users, the social media giant is becoming more and more effective with targeted ads. The company is facilitating businesses from around the world in the best way possible and hackers had to pay attention to their success.

Hence, the bad guys knew that they could scam countless people through the tools that marketers use on the social network.

Why Was LiveRamp Worth It?

Besides being a big data partner for Facebook, LiveRamp is a marketing powerhouse that has earned its name for matching data from the real world actions to online identities, helping advertisers more than their expectations. Thus that is also the reason why LiveRamp is favorite of more than 300 businesses and data providers which includes big names like Google, MasterCard, Uber, Snapchat, Spotify and Equifax.

So LiveRamp for Facebook helps advertisers target ads on the basis of data derived from a user’s offline activities and they also integrated Facebook’s Offline Conversions API to help the same advertisers see the effectiveness of their marketing campaigns with knowing how many people actually bought the product.

Liveramp doesn’t run ads on behalf of Facebook itself but it still has access to do so being a Facebook approved partner. Hence, when hackers ran a series of ads on LiveRamp’s customer accounts on Facebook, one of the ads was viewed more than 60,000 times and further directed users to a page that was made to steal the credit card details of users.

Facebook’s Security

Facebook continuously reminds its users of a number of security tools which primarily includes two-factor authentication and login alerts, just so that one should know if a hacker has tried to intrude. The social network even offers Security Center page for business accounts, along with a recommendation that businesses should go for quarterly security cleanups to make sure that employees don’t have unnecessary access.

However, Facebook only goes with the policy of recommending these security measures and not making it a requirement even for its big partners like LiveRamp which is a big problem actually.

Marcin Kleczynski, CEO of cybersecurity company Malwarebytes raised the concern regarding how Facebook doesn’t require separate Business Manager account and instead users can manage their multi-million dollar pages all through their personal profiles.

He further questioned that why Facebook never opted for higher standards when it comes to bigger partners, especially after knowing how people go for poor security habits including reusing the same password everywhere or not turning on two-factor authentication.

Honestly, till the time Facebook doesn’t make important security measurements a requirement, cybercriminals would have a better chance to have access to million-dollar advertising campaigns all by attacking personal profiles.

The post Hackers Attacked LiveRamp – A Big Data Partner of Facebook For A Bigger Advertising Scam appeared first on Artificial Intelligence.

More than 76 percent of enterprise organisations are planning to rearchitect apps for microservices, according to a Forrester report.

The Forester Research, backed by Ensonor and Wipro, found that enterprise adoption of microservices is continuing to increase as businesses transform themselves to compete in the digital era.

The research found that 76 percent of organisations now consider microservices to be a high or critical priority and only four percent do not have this on their agenda at all.

Oliver Presland, VP of Global Product Management at Ensono, said: “There was a time when monolithic architecture was the standard approach to software development in enterprise, but we are seeing an increasing focus from our clients in using a microservices architecture-first approach, either rearchitecting from the ground-up or adapting their existing applications. In a high-velocity world of high customer expectation, technology has become a true competitive advantage and organisations need to act smart and move fast. They need the agility to adapt and pivot. Microservices helps break complex software into smaller, manageable pieces and as our research shows, it’s now the default architecture for delivering that required agility.”

The research found that many organisations are struggling with lengthy deployment cycles (29 percent) and have trouble meeting delivery dates (16 percent). It is little surprise then that coinciding with the switch from monolithic software strategies is the ever-increasing importance being placed on DevOps that speeds up deployment cycles in order to deliver a better experience. 82 percent consider DevOps to be of high or critical importance to their organisations over the next year.

Microservices and DevOps come together to deliver real ROI and operational efficiencies. Organisations like Amazon, Netflix, eBay, Facebook, Uber, Groupon, Google all use microservices architecture.

Additionally, this trend towards microservices coincides with an increase in cloud-first strategies with 69 percent of enterprise organisations currently expanding or upgrading their implementation.

Oliver Presland continues: “New technologies such as serverless, containerisation and microservices are bringing us into a new generation of maximising cloud-first strategies with cloud-native architecture. This exploits the advantages of the cloud computing delivery model and many of our clients are moving to this approach, while optimising their legacy infrastructure.”

The post Enterprises rearchitecting apps for microservices appeared first on Artificial Intelligence.

Familiarity breeding contempt hits home in the results of a new study into the security threat from apps on Google Play. The research, conducted by the University of Sydney and CSIRO’s Data61, has unearthed thousands of dangerous apps hiding in plain sight in the online store, tricking users by mimicking popular alternatives. The study used artificial intelligence to identify likely counterfeits, before testing them for malware and other vulnerabilities.

The study deployed a neural network to examine both the design of icons and the wording in descriptions, reviewing “1.2 million apps” to identify “potential counterfeits for the top 10,000 apps.” It found “2,040 potential counterfeits that contain malware in a set of 49,608 apps that showed high similarity to one of the top 10,000 popular apps in the Google Play Store.” The research also found “1,565 potential counterfeits asking for at least five additional dangerous permissions than the original app and 1,407 potential counterfeits having at least five extra third-party advertisement libraries.”

The use of pre-trained AI algorithms to evaluate style and content “outperforms many baseline image retrieval methods for the task of detecting visually similar app icons,” and on the large dataset of more than1.2 million app icons, the study’s methods achieve “8%-12% higher precision” than alternatives.

“Many counterfeits can be identified once installed,” the authors explain, “however even a tech-savvy user may struggle to detect them before installation,” thus the idea to try the “novel approach of combining content embeddings and style embeddings generated from pre-trained convolutional neural networks to detect counterfeit apps.”

The study found that the 2,040 most dangerous counterfeits “were marked by at least five commercial antivirus tools as malware,” although, encouragingly, 6-10 months since we discovered the apps, 27%-46% of the potential counterfeits we identified are not available in Google Play Store, potentially removed due to customer complaints.”

None of this should come as a surprise-the insecurity of apps on both Android and iOS has been very much in the headlines recently.

Last year, Buzzfeed News reported that “eight apps with a total of more than 2 billion downloads in the Google Play store have been exploiting user permissions as part of an ad fraud scheme that could have stolen millions of dollars.” All eight apps were Chinese in origin, with seven from a single developer, Cheetah Mobile. “The companies claim more than 700 million active users per month for their mobile apps.”

And this month alone, Davey Winder reported for Forbes on the threat from mobile applications, leaving “iPhone and iPad users not as secure as they might imagine, [with] their personal data at risk.” ZDNet has reported that “three-quarters of mobile applications have vulnerabilities relating to insecure data storage, leaving both Android and Apple iOS users open to cyber attacks.” And TechCrunch has reported on vulnerabilities even in U.S. mobile banking apps.

Smartphone users cannot claim that they’re not being warned.

Both Google and Apple continue to fight the good fight to keep their ecosystems secure, and on Android Google Play Protect has been designed to guard against just such vulnerabilities. Google has also said that “in 2018, we introduced a series of new policies to protect users from new abuse trends, detected and removed malicious developers faster, and stopped more malicious apps from entering the Google Play Store than ever before. The number of rejected app submissions increased by more than 55%, and we increased app suspensions by more than 66%.”

The use of AI to moderate content and promote internet safety has been catapulted into the news by social media’s woes in the last 12 months. Projects like Google’s Jigsaw are a sign of things to come. This study is a start on applying the same thinking to a different realm, but one that struggles with the same issues of scale and user naivety.

Ultimately, there’s no substitute for common sense and treating apps from unknown sources as potential threats. And that means checking carefully, not clicking casually. We carry all of the most valuable and private information we have on our smartphones, and we gladly give those devices access to the cloud storage where we store the rest. Our phones know where we live and work, and where we bank and spend. That’s worth remembering before inviting strangers into our virtual homes and giving them permission to roam around simply because they ask nicely.

The post Android Warning: Thousands Of Dangerous Copycat Apps On Google Play, Study Finds appeared first on Artificial Intelligence.

There seems to be an app for everything these days, except for one that can make you a millionaire. Yet.

Less than a decade ago, it was unimaginable that an app might be able to sort through millions of songs to create playlists of ones you’ve never heard of that you might actually enjoy, like Spotify does. Or that Microsoft would be working on a fridge that puts together shopping lists based on what you have run out of, and might even be able to plan meals for you. Or self-driving cars that improve their performance as they spend more time on the roads.

But the emergence of technology from science-fiction to real life is increasingly common and many of these leaps forward share a core feature: they are powered by self-learning software that wrangles huge volumes of data and keeps improving its performance.

It’s called machine learning, a form of artificial intelligence, albeit a limited one (for now). Facebook uses it to deliver more of the content you actually watch on your newsfeed. Google does uses it to serve you more content of the kind you’ve clicked on in the past. The ability of software to work out how to serve its purpose better edges us all ever closer to a world in which genuinely smart machines  will be as common as household pets.

Machine learning is already in some rapidly growing personal finance apps, and many major financial service providers are already experimenting with its potential. For example, applications closed last month for Westpac’s innovation challenge, which called for projects to improve accounting and legal services by tapping into “big data, AI and machine learning”.

Swinburne Business School adjunct professor Steve Worthington told Money the impact of machine learning on personal finance management is only just beginning.

“At the moment in Australia we’ve got what you might call an erosion of trust in the traditional banking players or financial services players as demonstrated by the recent Commonwealth Bank scandal. So you could argue we’re already looking for other ways of handling our personal finances, [so we’re] open to suggestions as to how this might be done.”

Machine-learning powered handling of your personal finances is in its infancy, but the potential is staggering: swapping bank tellers or financial advisers for algorithm-powered chat-bots; fully automated investment services that can move your money through global markets within seconds of financial news; or the ability to instantly access loans based purely on data you’ve already shared online without realising it.

Those involved on the front lines of machine-learning supported financial apps aren’t promising the world yet but they are readying for the seismic shift machine learning will bring to their services.

One the most widely used examples is Acorns, a mobile app that connects to your credit and bank cards and rounds up every purchase you make, investing that little bit extra into a share portfolio. The app uses machine learning throughout almost every function of its services, such as classifying how you spend your money and identifying how financially literate you are. It then uses both insights to customise its messages, to enable you to reach your financial goals.

“We’ve got machine learning that goes through and predicts your future spending habits and future income based on your past,” says Acorns Australia director George Lucas, adding one of its most-clicked upon notifications is simple but effective: when Acorns alerts its users that they’ve been spending too much money on ATMs or their credit card.

Useful, sure. But an artificial intelligence breakthrough? Not yet. However it’s only the beginning for software-driven services that tell you how to invest your money or which insurance products to buy.

Lucas says improving their machine-learning capabilities is critical to their plans, and across the personal finance management industry. “It’s very central. All our work, all our research is on that.”

Another key area where artificial intelligence software is influencing personal finance management is using natural language processing, where a software learns how you talk and recalibrates how it liaises with you to improve communication.

This is the AI element that is most exciting for Alistair Bentley, the founder of Simply Wall Street, an Australian-based service that creates summaries using natural language processing software (another form of limited artificial intelligence) as well as data visualisations for stock investors to track the performance of their portfolios at a glance.

Simply Wall Street is at the beginning of exploring the potential of artificial intelligence, but Bentley urges investors to be wary of the promises of radically new technology.

“The idea that a major robot that picks stocks for you sounds quite dangerous to me,” Bentley says. “You have whole hedge funds with teams of PhDs who are trying to do that. So if anyone’s selling that kind of thing to you, I would say I’d probably run away.”

The rise of machine learning in personal finance is already having a wide array of consequences. These range from the comically eerie – in China you can buy a bucket of KFC using facial recognition software that links to a service offered by Alibaba (which is similar to Amazon but larger) – all the way to verging on the the flat-out dystopian. For example, companies harvesting the ample personal data we share online (including everything you search, click or look at for a second or two) to identify who is worth loaning money through their interests and purchasing behaviours and who is in the network.

“These [major internet] companies have huge amounts of information on us,” Worthing says. He believes the most significant work in AI-powered personal finance will come from large software companies moving into payments and financial services.

The idea of Google or Facebook assigning you a credit score and offering you a loan may seem far-fetched and they’ve not indicated any intention to move into this space yet. But Asian technology powerhouses such as Alibaba and TenCent (China’s most powerful social network) already have.

“It’s coming like a freight train,” says Dr David Tuffley, senior lecturer in applied ethics and sociotechnical studies at Griffith University. He estimates it will take three to  10 years before AI-powered personal finance services are safe for the average consumer.

“At the moment there’s a lot of questions and a lot of unknowns about it, so I’d be a bit wary of it myself. But later on I think I would probably take advantage of it once a lot of the [security and regulatory] issues have been ironed out,” Tuffley says. “These are not mature technologies yet so given the stakes, given the importance of what is at stake, I think people would be unwise to jump in before it’s fully matured.”

The post How artificial intelligence is changing personal finance appeared first on Artificial Intelligence.

More healthcare organizations are looking to use mobile technology to better engage patients, but providers must broaden their strategies to get more value from their efforts.

Data science needs to be incorporated into these efforts, enabling organizations to achieve strategic benefits from their efforts. From our organization’s experience, we found that data science provides the highest value to the enterprise when it’s embedded in applications that assist users in taking action on information.

Predictive modeling is just one way that data science can add value, in addition to guiding users to the best interventions for particular cases and speeding up insight into daily work.

When applications incorporate the best of modern mobile design, the incremental value and simplicity, compared with existing enterprise solutions, improved both adoption and user experience.

Consumers have become comfortable with using apps, but they expect them to actually do useful things, and do them quickly. With an app for ride services, data science can enable apps to use technology that estimates the fare, travel times, how long it will take the driver to arrive and surge pricing—supplying information seamlessly into users’ experience.

Healthcare organizations need to take a strategic lesson from these apps. Here are some ways that enterprise data science can be applied to their apps. The following examples refer to actual work in the Providence St Joseph myHIway app store, an internal platform for publishing hybrid web apps. These apps range from those that are in production, in development or in proof-of-concept.

Predictions
Data science can effectively produce risk scores, the outcome of a model produced by a classification algorithm and applied to new data, or a regression model, through which a specific value is predicted. The Providence St Joseph NoShow app tracks the risk of no-show over time, but also tracks the interventions users perform (for example, call with confirmation or leave a message). Over time, predictive models can incorporate the intervention as an additional predictor, and the model can prioritize instances in which the intervention is most likely to improve the outcome.

Outliers
Data science can help winnow down data to just the most important observations; this could be a simple threshold on a prediction or by more intricate statistical rules. For Providence’s Denials app, for example, it looks at the past six months of denials per charge code, per payer, per facility, and then uses some customized process control methodology to only show those combinations that are surprising or unusual. Combining it with a customizable email capability helps it join with the user to isolate, collaborate and act on important information.

Models
A trained model can itself be a product. Providence’s Search+ app uses the entire corpus of clinical notes across three separate Epic implementations to generate a Word2Vec model mapping words from notes to a high-dimensional vector representation, acting as a layer between a user entering search terms and a SolR search that generates semantically similar terms. Also, for the High-Deductible Health Plan app, it was unable to access eligibility information that would precisely identify each member’s deductible or how close they were to it, so it looks at a year of spending to model the likely deductible per group within each payer umbrella to estimate the current year’s status.

Calculations
The results of a predictive model extend beyond an end product—the Providence No-Show app is an example. Its tracked actions can be used to measure the rate of success (no-shows prevented as a result of making reminder calls), which will be further supplemented using historic charge information to estimate revenue generated per call. This information eventually may be paired with estimates of wages and productivity to suggest the optimal risk threshold for intervention.

Event triggers
Custom software development enables flexible design of data-dependent triggers to evoke some kind of notification, even when users are not actively working with the app. These trigger subscriptions can be tied to observations above a certain risk level, a new outlier being discovered or some calculated amount exceeding a certain threshold.

Streaming rules
It’s often crucial to intelligently handle and react to streaming data by building logical gates to manage the free flow of information that otherwise would be chaotic. In telemetry, benefits are derived by streaming both the transfers and the orders to create a full picture of hospital patients that are being placed on, and then taken off, heart monitors. A combination of business understanding and data sense has to be used to determine relevant cases, and also how a case flows from one class to another, based on the sequence of events.

Implementation lessons
Providence’s experience has been that enterprise products that reflect the collective experience of the last decade of smartphone development are received enthusiastically because they represent a leap from the user experience of prior solutions. That makes sense; data science is only useful when it’s adopted, which is most likely to happen when it is integrated into a product designed with the user at the center.

Success for a data science team is more likely when the output is packaged into a sleek, intuitive and useful application. The benefits of mobile capability are worth the extra work and security considerations, because the impact increases when everyone has access to the data in everyday settings in which it can be acted upon.

Another lesson learned is to be very thoughtful about the presentation of predictive algorithms. Especially when your product recommends action based on an algorithm’s score, trust in the algorithm accumulates slowly but can be quickly eroded. One way trust can be damaged is through a misunderstanding of probability. Probability may better be presented indirectly, through an index rather than a raw number, which also served to redirect focus from the risk calculation to the priority that should be put on intervention. Similarly, showing the variable most responsible for driving an elevated score can be tricky business and should be avoided.

When choosing projects to take on, it is critical to develop an intuition for the amount of value that these apps have to bring to different groups to compensate for the disruption involved in their implementation. For example, physicians at Providence spend almost their entire day inside our electronic medical record, and the only topic found to be sufficiently compelling to lure them to an app is their compensation. On the other hand, care managers are continually jumping between applications, making a solution that combined multiple sources of data especially appealing.

The post HIT Think How apps can incorporate data science to boost benefits appeared first on Artificial Intelligence.