A deep learning system takes textual hints from the context of images to describe them without the need for prior human annotations.

Since its humble beginnings at the turn of the millennium, deep learning, as both a scientific discipline and an industry, has come a long way.

From smartphone assistants to pattern recognition software, security solutions, and other applications, deep learning is becoming a multi-billion dollar business poised for great growth over the few next years.

However, for deep learning agents to reach their full potential, they have to “learn” how to learn on their own.

Herein lies the whole difference between supervised and unsupervised deep learning.

Self-Supervised Deep Learning

The power and appeal of deep learning is all about their ability to recognize different types of patterns like faces, voices, objects, images, and codes.

AI software doesn’t understand what these things really are, and all they see is digital data, and they’re pretty good at that.

The great computer vision capability of deep learning algorithms enable them to tell these things apart, categorize, and classify them.

To do so, however, this software needs to be supervised.

They require human manual input in the form of annotations to guide them before they generalize and build on what they learned into new, similar situations.

Building and labeling large datasets is a complicated and time-consuming task.

Unsupervised machines will be completely autonomous as all they need is data taken directly from their environment. From there, they would take the information to make predictions and yield the expected results.

To design unsupervised, or self-supervised deep learning systems, computer scientists take inspirations from how human intelligence works.

Now, an international team of computer vision scientists has devised a method to enable deep learning software to learn the visual features of images without the need for annotated examples.

Researchers from Carnegie Mellon University (U.S.), Universitat Autonoma de Barcelona (Spain), and the International Institute of Information Technology(India), worked on the study,

Unsupervised Computer Vision Algorithms, it’s a Matter of Semantics

In the study, the team built computational models that use textual information about images found on websites, like Wikipedia, and linked them to the visual features of these images.

“We aim to give computers the capability to read and understand textual information in any type of image in the real world,” said Dimosthenis Karatzas, a research team member.

In the next step, researchers used the models to train deep learning algorithms to pick adequate visual features that textually describe images.

Instead of labeled information about the content of a particular image, the algorithm takes non-visual cues from the semantic textual information found around the image.

“Our experiments demonstrate state-of-the-art performance in image classification, object detection, and multi-modal retrieval compared to recent self-supervised or naturally-supervised approaches,” wrote researchers in the paper.

This is not a fully unsupervised system as algorithms still need models to train on, but the technique shows that deep learning algorithms can tap into the internet to enhance their unsupervised learning abilities.

“We will continue our work on the joint-embedding of textual and visual information,” said Karatzas. “looking for novel ways to perform semantic retrieval by tapping on noisy information available in the Web and Social Media.”

The post New Deep Learning Strategy Could Enhance Computer Vision appeared first on Artificial Intelligence.

In the world of microservices, the goal is to have a small piece of software that performs a well-defined set of tasks. Microservices are software applications that are self-contained. They are small, independently deployable modular services that run a unique process and communicate through a well-defined, lightweight mechanism to serve a specific goal.

Typically, clear boundaries are set with regards to what your microservice can or cannot do. The move to microservices requires not only a change in architecture, but also a solid foundation of trust between the different teams who are working together to develop these microservices. Building this trust gives them the confidence they need to rely on the services’ availability and adherence to the agreed upon service contract for standard APIs.

Without a high level of trust between teams of developers, chaos will quickly ensue in your development organisation. Each team will build whatever they want and/or change APIs without notifying the rest of the organisation. In this state of disarray, functionality will break and your software development will come to a grinding halt.

While trust is extremely important in developing microservices, planning ahead for potential security issues is even more critical. Unfortunately, security considerations are often overlooked in the process of transitioning to microservices. The consequences of security failings can be devastating for your company.

For example, let’s say you are deploying a microservice that accepts input from a user and passes that input to a backend database. If your service is not secure and the input is not validated, hackers will be able to inject malicious code into the system and bring down your service — or perhaps even worse, compromise your entire system.

A plethora of microservices

As your repertoire of services and applications grows, it is probably safe to assume that the number of available microservices will grow as well. As the number of options increases, more and more security issues are bound to come up. In this post, we will review some of the issues and potential solutions that you should take into consideration before your business makes the shift to using microservices.

Reuse of code

Using shared code and libraries can help you jumpstart your move to microservices, but it can also be a double-edged sword. If you choose to borrow and use something off of the internet (i.e., an open-source solution), you are forever tied to that code and all of its shortcomings.

On one hand, reuse of industry standard technology can be a good thing since it has already been tested and used around the world. On the other hand, widespread usage of standard technology can be problematic. If a component vulnerability surfaces, your company and other companies that are using the technology will need to apply emergency patches to mitigate critical flaws in all of your applications. Imagine what it would be like to have to either deploy all of your services again because of a new security package, or to patch the servers with a whole new binary. Such a scenario sounds like a nightmare, doesn’t it?

Take the Heartbleed bug for example. When the bug was discovered in April of 2014, approximately 66 per cent of the internet’s web servers needed to be patched because the software that included the vulnerable libraries was used on almost all of the world’s web servers. This catastrophe required the patching and rebooting of hundreds of thousands of servers in a very short period of time.

If you happen to have a smaller number of servers, the solutions of deploying everything again or patching with a new binary are doable, either manually or with a basic amount of scripting. But when you scale, these are no longer viable options for solving your problem. In these cases with scaling, you need to have a tried and tested method of orchestration already in place in order for you to perform mass operations quickly to alleviate the problems.

Denial of service

Ensuring that your applications are secure is no easy feat, and managing a number of services that have multiple entry points from the outside can be difficult. As the number of services grow, the magnitude of this issue is amplified. Managing security groups appropriately will help you ensure that only the correct ports are exposed. Doing so can help save you and your applications a significant amount of pain and anguish if a malicious party targets your product in an attack.

Setting up a firewall application or an alternative solution in front of your system can correct a problem like this by ensuring that only the appropriate traffic arrives at your application’s front door and that it does not contain malicious codes or threats.

Traffic between microservices

Each microservice passes information from one to another. When the traffic is in a segregated part of your own network, it is safe to assume that your risk of having an eavesdropper is decreased since you are usually behind a corporate firewall, which makes you less susceptible to man-in-the-middle attacks.

When you move to the cloud, this assumption is no longer valid. Traffic between your microservices should be encrypted on the cloud. This means that, in addition to your microservices handling encrypted traffic, they will also need to ensure that the performance of your underlying applications does not suffer as a result of the extra work they have to perform with encrypting and decrypting information.

There are a number of methods that you can use to mitigate this problem. One such solution is creating a Virtual Private Cloud (VPC), which will enable you to segregate your workloads in the cloud without allowing a malicious intruder into the system who could eavesdrop on your traffic. However, this is not a foolproof solution, and it still has a number of attack vectors that you will need to address. Another option is offloading the encryption to an external service (e.g., a load balancing service) that will enable traffic protection with minimal disruption or changes to your current system.

Secret sharing

In order to make sure that your microservices are not open to the world, we suggest that you add authentication between the services in your system. Doing so will ensure that only the correct pieces are allowed to talk with each other and that they have the proper credentials to do so.

Embedding these secrets into your applications is a very bad idea. Best practices for modern architecture strongly advise against storing any credentials on your servers. Of course, this brings up the question of how you will allow applications to authenticate with each other and third-party services if the credentials cannot be stored locally?

There are a few ways to tackle this problem. One strategy is to use third-party tools or the tools and services that are already available from most cloud providers. The concept is pretty simple. When you initiate an authentication request, you ask another service to request a temporary set of credentials on your behalf, which allows you access for a set period of time. This solves the issue of longevity of credentials because they expire after a certain period of time, and because there are no credentials that are embedded in the microservice itself.

Security across the board

It is highly unlikely that those who are producing the microservice will actually belong to a single team. It makes sense that each team will develop, test, and deploy their own microservice (or set thereof). Therefore, the responsibility of securing the service cannot lie solely with a single, traditional operation security team and organisation—it should lie with all of the teams that are producing the software.

In order to ensure that your company will be protected and secure from outside attackers, we recommend that you consider some sort of partnership between the traditional OPSEC team and the developers. Establishing such a partnership will help the participants work together so that the software they are producing is secure by default, hardened, and continuously tested for compliance against a baseline of security requirements before it goes out the door. When the responsibility also lies with the teams who are creating the service, their level of engagement and awareness will definitely increase.

Code changes

The last issue we will address is the lifecycle of the application. No one writes software that is 100 per cent perfect the first time around — there will always be some bugs to fix before the software runs as efficiently as it is supposed to. For example, the foundation of Agile methodology is reiterated all the time, providing only the minimal viable product and improving as you progress with the project.

When you are dealing with microservices, there can be several changes each day. When you upgrade your application by changing or adding functionality, you will need to ensure that your code is (at minimum) the same as it was before, if not even better. This requires scanning the added code for vulnerabilities and weaknesses before the code is even deployed. You will need to tie this into your continuous integration processes so that this is performed as part of your release process.

Summary

Maintaining a secure system is usually a daunting task, and with the shift to microservices, there is an additional vector that needs to be addressed. Each microservice has its respective weak points that need to be secured, hardened, and continuously monitored for vulnerabilities.

The scope of this task should not be underestimated — as your use of microservice architecture grows, security issues become more significant and urgent. To protect yourself and your company, you should address these security issues early and often. We encourage you to use the suggestions from this post to help you with the process of securing your microservices.

The post Security in the world of microservices appeared first on Artificial Intelligence.

Defined as the massive volume of structured and unstructured data that can’t possibly be processed using traditional software or database strategies, Big Data is affecting every corner of the business world. It’s no surprise, really, seeing that more data has been created in the past two years than in the entire history of the human race. By 2020, roughly 1.7 megabytes of new information will be created for every second for every human being and, at that point, the digital universe will be 44 zettabytes strong (up from a current 4.4 zettabytes).

As supply chain managers scramble to wrap their arms around the reams of information now at their fingertips, a growing number of software providers are making the task more manageable and useful. In other words, simply having the data at your avail isn’t enough; it’s about taking that information and transforming it into actionable insights that help drive operational efficiencies across the supply chain.

“Supply chains are more complex than ever, and with these complexities come many challenges,” says Shannon Vaillancourt, president at RateLinx. “Big Data allows companies to diagnose the issue so they truly understand what is causing it.” Of course, capturing the data and then using it to make good decisions are two entirely different things. To help fill that “gap,” Vaillancourt says software developers are focusing on the 5 Vs of Big Data: variety, velocity, veracity, volume and value.

Vaillancourt says the final “v” is extremely important and often overlooked. “Companies need to be looking for software that turns all of their data into value—or, actionable,” he points out. “Actionable data is created through analytics; it’s the analytics that tells the user what to do, and ultimately what action to take.”

Top 20 Supply Chain Managment Sofware Suppliers

SCM (SCE, SCP, Procurement) Total Software Revenue

Following are six big ways that Big Data is affecting the supply chain and helping companies take the right actions.

1. Get better diagnostic information.

To solve problems and circumvent future challenges, companies need good diagnostic data. Big Datagives them that, according to Vaillancourt, while also ensuring that their future strategies are based on solid historical information. “Big Data can help companies diagnose many issues, which will in turn allow them to develop strategies to solve the issues,” he says, “and then ultimately deploy the strategies successfully.”

For example, the organization that wants to leverage Big Data for track and trace of its products can do so by combining the purchase order (PO) details, shipment information and the carrier’s tracking information. Then, once that data is standardized and cleansed, analytics can be applied to it in a way that truly makes the information actionable. “If the analytics notifies the user about a late shipment before the carrier issues the notification,” Vaillancourt explains, “then that user can enact a contingency plan and get the product faster from an alternate source.”

2. Get a clearer “crystal ball” for the future.

Defined as the data mining, statistics, modeling, machine learning, and artificial intelligence used to analyze current data to make predictions about the future, predictive analytics is the modern-day supply chain manager’s crystal ball. “Predictive analytics makes it possible to analyze data and create assumptions as to what will happen to not only predict the future, but influence it as well,” says Marcell Vollmer, chief digital officer at SAP Ariba.

In Kansas City, for instance, a local police department is using data to stop crime before it happens by identifying “hot spots,” patrolling those areas more aggressively and then more closely monitoring the activities of recent parolees. In the business world, predictive analytics is allowing firms to more clearly understand customer needs and adapt their business to accommodate them. Take pricing, for example. Using predictive analytics, companies can predict equilibrium before releasing a new product, thus maximizing the revenue of the solution out of the gate while also understanding future demand. “Data is the new currency,” Vollmer adds, “and predictive analytics is the key to collecting the dividends it pays.”

3. Manage external factors that are beyond your control.

External factors can have a substantial impact on supply chains, yet in many cases these outside forces are hard to control and even detect. “From weather to oil prices to consumer demand, supply chain executives who can quantify and anticipate such impact can better plan their materials and inventory,” says Rich Wagner, CEO at Prevedere. He says retailers are particularly well positioned to leverage this advantage, namely because they’re operating in a dynamic environment where consumers expect quick, accurate deliveries. “If a product is unavailable, manufacturers and retailers alike risk not only losing a customer forever, but also a digital media backlash,” Wagner points out. How can Big Data help? By helping firms better predict demand, and therefore better plan their inventory to mitigate against shortages. The same benefits apply on a global scale, where both supply chains and operations are becoming more interconnected and, subsequently, more impacted by world events. “By coupling Big Data with predictive analytics,” Wagner says, “it’s quite possible to keep a handle on numerous economic and consumer behavior metrics to be better prepared for what’s coming next.”

4. Make more profitable supply chain demand forecasts.

Access to global data, combined with the power of Cloud computing, is giving technology more power to tackle even the toughest supply chain challenges. “With today’s advancements in machine learning, companies can use technology to constantly monitor those external forces,” says Wagner, “and get a real-time view of what’s ahead.” He sees this as a fundamental change in demand planning—compared to traditional forecasts built on past performance with the assumption of stable economic conditions. “Executives know that they can’t rely on precedence and they need insights to make decisions about the future with certainty,” says Wagner. “This desire to be immediately notified of shifts in momentum is now a reality.” For example, one global beverage manufacturer saved about $9 million by improving product distribution through predictive demand models. “The manufacturer realized that external factors (e.g., the architectural billings index) were leading indicators of performance,” says Wagner, “so it adapted its supply chain planning across 400 brands and 21 distributors.”

5. Reduce demand variability and cycle times.

Big Data is turning supply chain managers into “mind readers,” allowing them to predict and react to buyer behaviors in new ways. On the demand side, for instance, Big Data helps companies gain better understanding over consumer behaviors, foot traffic, buyer preferences and the actions that their competitors are taking. “This gives companies a solid offensive footing,” says Dennis Groseclose, president and CEO at TransVoyant, “and allows them to fuse external data and demand patterns to more effectively reduce demand variability.” Having actionable data also helps companies better manage lead times, variability and capacity. This, in turn, helps them better understand manufacturer and carrier behaviors. “With this information in hand, companies can squish planning cycle times down to one month vs. five months,” says Groseclose, “or to one week vs. five weeks.”

6. Prepare for the “SNEW” wave.

Here’s a buzzword you may not have heard of yet: SNEW, or social media, news, event and weather data, is the next acronym that’s either going to make supply chain managers sit up and take notice, or make them roll their eyes and groan. Either way, SNEW data promises to help improve supply chain capabilities and give companies even more data to sift through, digest and make sense of. An existing forecast, for example, can be adjusted accordingly when accurate weather predictions are factored into the equation.

Driven by the Internet of Things (IoT), SNEW uses a “combination of data feeds to determine the best routing, risk management, and other supply chain decisions,” according to Steve Banker, vice president, supply chain management at ARC Advisory Services, who sees SNEW as a potential player in the future of supply chain visibility and risk avoidance (or mitigation).

“This is a new solution to the market, and it’s being driven by the emergence of new technological capabilities,” Banker notes. The integration of social media, news, event and weather data into the manufacturing and distribution process is also getting a boost from the ongoing digitization of the supply chain.

“What we’re looking at is a series of technologies that are either rapidly emerging or already further along in terms of emergence,” says Banker, noting that while IoT is a bit further along in terms of maturity, concepts like SNEW and blockchain (i.e., a digital ledger where transactions made in bitcoin are recorded chronologically and publicly) are still nascent. “Over time,” he concludes, “these innovations will continue to generate Big Data that companies will use for decision making.”

The post 6 Ways Big Data is enhancing the global supply chain appeared first on Artificial Intelligence.

• 44% of contact centres have no intentions to introduce AI.
• 40% of contact centres agree that emotional intelligence is an area they need to explore.
• Emotional intelligence has been eagerly embraced by the contact centre industry.

You’ve no doubt heard about artificial intelligence making inroads into the contact centre, initially in the form of chatbots: AI-powered software applications that can take customer queries, either textual or verbal, and produce useful, relevant answers with consistency and accuracy that can match a human operator.

Depending on whose views you read, AI is either going to enhance the role of human operators, freeing them to handle the more challenging and rewarding queries, or it’s going to relegate them to the ranks of the unemployed.

SMAART Recruitment surveyed over 150 contact centre leaders and decision makers around Australia for its 2017 Contact Centre Leaders Survey. It reported: “Around seven percent of Australian contact centres currently use AI in their interactions with customers, with an additional 11 percent planning to introduce it in the coming 12 months. A high percentage of remaining contact centres are currently exploring the idea, however 44 percent of contact centres have no intentions to introduce AI.”

“Around seven percent of Australian contact centres currently use AI in their interactions with customers.”

You would not expect a recruitment company to be telegraphing mass reduction in employee numbers and a blog discussing the survey’s findings concluded: “Australian contact centres are taking a steady and careful approach to introducing artificial intelligence (AI) into their contact centres … AI presents contact centres with great opportunities and challenges in the coming years.

Those that learn from it and understand it quickly will be the first to benefit.”

However, SMAART Recruitment also focused on another kind of intelligence that will likely be much more challenging for technology to emulate and/or replace: emotional intelligence.

“Increasing staff emotional intelligence (EQ) is one of the key areas of focus for Australian contact centres.”

In another blog reporting additional results from the survey, it said: “Increasing staff emotional intelligence (EI) is one of the key areas of focus for Australian contact centres. In a recent study of contact centre managers and leaders, over 40 percent of respondents indicated that this was an area they needed to explore more and improve in the next 12 months.”

If you look at the definition of EI on Wikipedia, you will see why computer emulation of EI is challenging. It is defined as “the capability of individuals to recognise their own and other people’s emotions, discern between different feelings and label them appropriately, use emotional information to guide thinking and behaviour, and manage and/or adjust emotions to adapt to environments or achieve one’s goal(s).”

Wikipedia says the term first appeared in a 1964 paper by Michael Beldoch and gained popularity in the 1995 book by that title, written by author, psychologist, and science journalist Daniel Goleman.

It goes on to note that the concept has not been universally accepted. “Goleman’s 1995 analysis of EI has been criticised within the scientific community, despite prolific reports of its usefulness in the popular press.”

However there is no doubt EI has been eagerly embraced by the contact centre industry. A Google search will yield many results such as “Does Your Contact Center Lack Emotional Intelligence?”; “Is Your Contact Center Emotionally Intelligent?”,“Emotional Intelligence and Its Importance in the Call Center” and, intriguingly one that seems to be having a bet each way: “AI Can Help Boost Agents’ Emotional Intelligence.”

In our next post we’ll explore some of these ideas in more detail.

The post Contact centre intelligence: emotional or artificial? appeared first on Artificial Intelligence.