Daily Archives: July 24, 2017

While online shopping has taken great strides in recent years, the brick-and-mortar retail hasnt managed to keep pace.

Artificial intelligence now permeates every aspect of ecommerce platforms, especially where customer interactions are involved. Smart product suggestions, AI-powered search, cognitive customer service agents are just some of the innovations that have helped make online shopping more personalized and enjoyable for the customerand more profitable for the retailer of course.

Meanwhile, AI advances in brick-and-mortar retail have mostly remained in inventory management and back store operations. The few innovations that have happened in the customer-facing aspects of in-person retail have little or no AI involved, and have failed to make tangible positive impact in the shopping experience and gain wide adoption.

Fortunately, this is something that is fast changing as technological developments enable retailers to gather in-store data and deploy AI-powered solutions. Artificial intelligence can help fix old problems in retail tech as well as introduce new possibilities that were previously inconceivable. Here are some of the trends that are worth watching.

A few years ago, in-store beacons were supposed to be the biggest thing that happened to brick-and-mortar retail, but didnt live up to its hype. Part of the problem with beacons is that they introduce new complexities without solving the real problems customers are facing. Beacons require customers to install an app that does little more than pop up annoying promotions that in no way rivals the personalized suggestions of online shopping platforms.

Now, retailers are experimenting with a new generation of apps powered by machine learning algorithms, whose value go beyond displaying prices and coupons. IBM Watson, a leader in cognitive computing and natural language processing, has partnered with several large retailers to help them better understand and serve the needs of their customers.

An example is Macys On Call, a mobile web application that uses the Watsons cognitive computing power and location-based software to help shoppers get information while theyre navigating the companys stores. The application is able to parse and understand natural language queries about such things as the location of products, departments and services in a particular store, and it responds in a relevant way. As is with all machine learningbased platforms, every customer interaction makes On Call smarter.

Sears Automotive is using the same technology for its Digital Tire Journey in-store web app, which helps shoppers navigate their way through the stores wide assortment of tires using a conversational interface and find whats best for their needs.

While providing value to customers, these apps are enabling retailers to gather a wealth of customter-related data that can in turn be used to fuel other AI-powered solutions.

Retailers annually lose a collective $45 billion to shrinkage, due to non-scans and other errors occurring at the point of sale. This is an especially serious problem at self-checkouts, the technology that was supposed reduce friction and streamline the customer experience but ended up opening a Pandoras box of new problems.

A handful of companies are working toward addressing this problem in real time through artificial intelligence. Everseen, a software company founded in Cork, Ireland, uses computer vision and AI algorithms to analyze video feeds from retailers staffed registers and self-checkout feeds and automatically detect when a product is left unscanned. Whenever Everseen detects unusual activity, it sends a notification to store management via smartwatch, tablet or other mobile device. This will help prevent theft, but it will also help provide assistance at self-checkouts, which are the source of much customer frustration. The companys current AI technology is in use by five of the worlds 10 largest retailers.

StopLift is another company that offers a similar technology. StopLift uses computer vision and video analytics to detect a number of common scams and errors at checkouts. The system compares the items it detects on video to actual POS data to track items that have not been scanned.

Both solutions become better over time as they gather more data and tune themselves to the specifics of each store.

Many believe that in the future, retail will be fully automated by AI, eliminating long lines and obviating the need for checkouts altogether. This means customers can enter a store, grab the items they need and exitwithout getting arrested for shoplifting.

Though the concept is far from mature, a number of companies are making headways in this direction. Last year, Amazon announced Go, a checkout-free retail store that is still in the experimental stages. Go uses computer vision, machine learning algorithms and IoT sensors to understand customers interactions across the store. The technology automatically updates the shopping cart in an associated mobile app whenever a customer picks up or returns an item from a store shelf.

Amazons plan to open its store to the public in 2017 has hit some hurdles. But the complexities have done nothing to deter the online retail giants resolve in creating the store of the future, and its $13.7 billion acquisition of the Whole Foods might have something to do with it.

Neither has Amazons difficulties prevented other companies from making similar moves, including Walmart, the largest retailer in the U.S., which is taking serious strides to incorporate AI in its retail stores.

Everseen, which has been working on a similar concept since 2012, plans to introduce its own checkout-free technology soon. Called 0Line, the solution will provide retailers with an AI-powered network of video cameras, sensors and biometric data to recognize customers. All of this will interact with inventory, POS and a mobile-based payment solution that will enable instant transactions. By the time customers leave the store, their accounts will have been charged and an itemized virtual receipt will be made available to them.

Thanks to a number of developments, AIs reach is fast expanding into every domain of the physical world. These examples show that brick-and-mortar retail is bound for some major transformations. In a few years the in-store shopping experience may look much different from what were used to, maybe even smarter than its online counterpart.

This post is part of our contributor series. The views expressed are the author's own and not necessarily shared by TNW.

Read next: From Uber to Postmates: A tipping guide for the sharing economy

Go here to see the original:

How artificial intelligence is defining the future of brick-and-mortar shopping - TNW

China aims to make the artificial intelligence industry a new, important driver of economic expansion by 2020, according to a development plan issued by the State Council.

Policymakers want to be global leaders, with the AI industry generating more than 400 billion yuan ($59 billion) of output per year by 2025, according to an announcement from the Cabinet late Thursday. Key development areas include AI software and hardware, intelligent robotics and vehicles, virtual reality and augmented reality, it said.

Artificial intelligence has become the new focus of international competition, the report said. We must take the initiative to firmly grasp the next stage of AI development to create a new competitive advantage, open the development of new industries and improve the protection of national security.

The plan highlights Chinas ambition to become a world power backed by its technology business giants, research centers and military, which are investing heavily in AI. Globally, the technology will contribute as much as $15.7 trillion to output by 2030, according to a PwC report last month. Thats more than the current combined output of China and India.

The positive economic ripples could be pretty substantial, said Kevin Lau, a senior economist at Standard Chartered Bank in Hong Kong. The simple fact that China is embracing AI and having explicit targets for its development over the next decade is certainly positive for the continued upgrading of the manufacturing sector and overall economic transformation.

Chinese AI-related stocks advanced Friday. CSG Smart Science & Technology Co. climbed as much as 9.3 percent in Shenzhen before closing 3.1 percent higher, while intelligent management software developer Mesnac Co. surged 9.8 percent after hitting the 10 percent daily limit in earlier trading.

AI will have a significant influence on society and the international community, according to an opinion piece by East China University of Political Science and Law professor Gao Qiqi published Wednesday in the Peoples Daily, the flagship newspaper of the Communist Party.

PwC found that the worlds second-biggest economy stands to gain more than any other from AI because of the high proportion of output derived from manufacturing.

Another report from Accenture and Frontier Economics last month estimated that AI could increase Chinas annual growth rate by 1.6 percentage point to 7.9 percent by 2035 in terms of gross value added, a close proxy for GDP, adding more than $7 trillion.

The State Council directive also called for Chinas businesses, universities and armed forces to work more closely in developing the technology.

We will further implement the strategy of integrating military and civilian developments, it said. Scientific research institutes, universities, enterprises and military units should communicate and coordinate.

More AI professionals and scientists should be trained, the State Council said. It also called for promoting interdisciplinary research to connect AI with other subjects such as cognitive science, psychology, mathematics and economics.

Excerpt from:

China artificial intelligence bid seeks $59 billion industry - The Denver Post

Data big and small have come to education, from creating online platforms to increasing standardised assessments.

Artificial intelligence (AI) enables Siri to recognise your question, Google to correct your spelling, and tools such as Kinect to track you as you move around the room.

Data big and small have come to education, from creating online platforms to increasing standardised assessments. But how can AI help us use and improve it?

Researchers in AI in education have been investigating how the two intersect for several decades. While its tempting to think that the primary dream for AI in education is to reduce marking load a prospect made real through automated essay scoring the breadth of applications goes beyond this.

For example, researchers in AI in education have:

These are new approaches to learning that rely heavily on students engaging with new kinds of technology. But researchers in AI, and related fields such as learning analytics, are also thinking about how AI can provide more effective feedback to students and teachers.

One perspective is that researchers should worry less about making AI ever more intelligent, instead exploring the potential that relatively stupid (automated) tutors might have to amplify human intelligence.

So, rather than focusing solely on building more intelligent AI to take humans out of the loop, we should focus just as much on intelligence amplification or, going back to its intellectual roots, intelligence augmentation. This is the use of technology including AI to provide people with information that helps them make better decisions and learn more effectively.

This approach combines computing sciences with human sciences. It takes seriously the need for technology to be integrated into everyday life.

Keeping people in the loop is particularly important when the stakes are high, and AI is far from perfect. So, for instance, rather than focusing on automating the grading of student essays, some researchers are focusing on how they can provide intelligent feedback to students that helps them better assess their own writing.

And while some are considering if they can replace nurses with robots, we are seeking to design better feedback to help them become high-performance nursing teams.

But for the use of AI to be sustainable, education also needs a second kind of change: what we teach.

To be active citizens, students need a sound understanding of AI, and a critical approach to assessing the implications of the datafication of our lives from the use of Facebook data to influence voting, to Google DeepMinds access to medical data.

Students also need the skills to manage this complexity, to work collaboratively and to innovate in a changing environment. These are qualities that could perhaps be amplified through effective use of AI.

The potential is not only for education to be more efficient, but to think about how we teach: to keep revolution in sight, alongside evolution.

Another response to AIs perceived threat is to harness the technologies that will automate some forms of work, to cultivate those higher-order qualities that make humans distinctive from machines.

Amid growing concerns about the pervasive role of algorithms in society, we must understand what algorithmic accountability means in education.

Consider, for example, the potential for predictive analytics in flexi-pricing degrees based on a course-completion risk-rating built on online study habit data. Or the possibility of embedding existing human biases into university offers, or educational chatbots that seek to discern your needs.

If AI delivers benefits only to students who have access to specific technologies, then inevitably this has the potential to marginalise some groups.

Significant work is under way to clarify how ethics and privacy principles can underpin the use of AI and data analytics in education. Intelligence amplification helps counteract these concerns by keeping people in the loop.

A further concern is AIs potential to result in a de-skilling or redundancy of teachers. This could possibly fuel a two-tier system where differing levels of educational support are provided.

The future of learning with AI, and other technologies, should be targeted not only at learning subject content, but also at cultivating curiosity, creativity and resilience.

The ethical development of such innovations will require both teachers and students to have a robust understanding of how to work with data and AI to support their participation in society and across the professions.

See the original post here:

Artificial intelligence holds great potential for both students and teachers but only if used wisely - The Conversation AU

The Raspberry Pi is one of the most exciting developments in hobbyist computing today. Across the world, people are using it to automate beer making, open up the world of robotics and revolutionise STEM education in a world overrun by film students. These are all laudable pursuits. Meanwhile, what is Microsoft doing with it? Creating squirrel-hunting water robots.

Over at the firms Machine Learning and Optimization group, a researcher saw squirrels stealing flower bulbs and seeds from his bird feeder. The research team trained a computer vision model to detect squirrels, and then put it onto a Raspberry Pi3 board. Whenever an adventurous rodent happened by, it would turn on the sprinkler system.

Microsofts sciurine aversions arent the point of that story its shoehorning of a convolutional neural network onto an ARM CPU is. Itshows how organizations are pushing hardware further to support AI algorithms. AsAI continues to make the headlines, researchers are pushing its capabilities to make it increasingly competent at basic tasks such as recognizing vision and speech.

As people expect more of the technology, cramming it into self-flying drones and self-driving cars, the hardware challenges are increasing. Companies are producing custom silicon and computing nodes capable of handling them.

Jeff Orr, research director at analyst firm ABI Research, divides advances in AI hardware into three broad areas: cloud services, ondevice, and hybrid. The first focuses on AI processing done online in hyperscale data centre environments like Microsofts, Amazons and Googles.

At the other end of the spectrum, he sees more processing happening on devices in the field, where connectivity or latency prohibit sending data back to the cloud.

Its using maybe a voice input to allow for hands-free operation of a smartphone or a wearable product like smart glasses, he says. That will continue to grow. Theres just not a large number of real-world examples ondevice today. Heviews augmented reality as a key driver here. Ortheres always this app, we suppose.

Finally, hybrid efforts marry both platforms to complete AI computations. This is where your phone recognizes what youre asking it but asks cloud-based AI to answer it, for example.

The clouds importance stems from the way that AI learns. AImodels are increasingly moving to deep learning, which uses complex neural networks with many layers to create more accurate AI routines.

There are two aspects to using neural networks. The first is training, where the network analyses lots of data to produce a statistical model. This is effectively the learning phase. The second is inference, where the neural network then interprets new data to generate accurate results. Training these networks chews up vast amounts of computing power, but the training load can be split into many tasks that run concurrently. This is why GPUs, with their double floating point precision and huge core counts, are so good at it.

Nevertheless, neural networks are getting bigger and the challenges are getting greater. Ian Buck, vice president of the Accelerate Computing Group at dominant GPU vendor Nvidia, says that theyre doubling in size each year. The company is creating more computationally intense GPU architectures to cope, but it is also changing the way it handles its maths.

Itcan be done with some reduced precision, he says. Originally, neural network training all happened in 32bit floating point, but it has optimized its newer Volta architecture, announced in May, for 16bit inputs with 32bit internal mathematics.

Reducing the precision of the calculation to 16 bits has two benefits, according to Buck.

One is that you can take advantage of faster compute, because processors tend to have more throughput at lower resolution, he says. Cutting the precision also increases the amount of available bandwidth, because youre fetching smaller amounts of data for each computation.

The question is, how low can you go? asks Buck. Ifyou go too low, it wont train. Youll never achieve the accuracy you need for production, or it will become unstable.

While Nvidia refines its architecture, some cloud vendors have been creating their own chips using alternative architectures to GPUs. The first generation of Googles Tensor Processing Unit (TPU) originally focused on 8bit integers for inference workloads. The newer generation, announced in May, offers floating point precision and can be used for training, too. These chips are application-specific integrated circuits (ASICs). Unlike CPUs and GPUs, they are designed for a specific purpose (youll often see them used for mining bitcoins these days) and cannot be reprogrammed. Their lack of extraneous logic makes them extremely high in performance and economic in their power usage but very expensive.

Google's scale is large enough that it can swallow the high non-recurring expenditures (NREs) associated with designing the ASIC in the first place because of the cost savings it achieves in AIbased data centre operations. Ituses them across many operations, ranging from recognizing Street View text to performing Rankbrain search queries, and every time a TPU does something instead of a GPU, Google saves power.

Its going to save them a lot of money, said Karl Freund, senior analyst for high performance computing and deep learning at Moor Insights and Strategy.

He doesnt think thats entirely why Google did it, though. Ithink they did it so they would have complete control of the hardware and software stack. If Google is betting the farm on AI, then it makes sense to control it from endpoint applications such as self-driving cars through to software frameworks and the cloud.

When it isnt drowning squirrels, Microsoft is rolling out field programmable gate arrays (FPGAs) in its own data centre revamp. These are similar to ASICs but reprogrammable so that their algorithms can be updated. They handle networking tasks within Azure, but Microsoft has also unleashed them on AI workloads such as machine translation. Intel wants a part of the AI industry, wherever it happens to be running, and that includes the cloud. To date, its Xeon Phi high-performance CPUs have tackled general purpose machine learning, and the latest version, codenamed Knights Mill, ships this year.

The company also has a trio of accelerators for more specific AI tasks, though. For training deep learning neural networks, Intel is pinning its hopes on Lake Crest, which comes from its Nervana acquisition. This is a coprocessor that the firm says overcomes data transfer performance ceilings using a type of memory called HBM2, which is around 12times faster than DDR4.

While these big players jockey for position with systems built around GPUs, FPGAs and ASICs, others are attempting to rewrite AI architectures from the ground up.

Knuedge is reportedly prepping 256-core chips designed for cloud-based operations but isnt saying much.

UK-based Graphcore, due to release its technology in 2017, has said a little more. Itwants its Intelligence Processing Unit (IPU) to use graph-based processing rather than the vectors used by GPUs or the scalar processing in CPUs. The company hopes that this will enable it to fit the training and inference workloads onto a single processor. One interesting thing about its technology is that its graph-based processing is supposed to mitigate one of the biggest problems in AI processing getting data from memory to the processing unit. Dell has been the firms perennial backer.

Wave Computing is also focusing on a different kind of processing, using what it calls its data flow architecture. Ithas a training appliance designed for operation in the data centre that it says can hit 2.9 PetaOPs/sec.

Whereas cloud-based systems can handle neural network training and inference, Client-side devices from phones to drones focus mainly on the latter. Their considerations are energy efficiency and low-latency computation.

You cant rely on the cloud for your car to drive itself, says Nvidias Buck. Avehicle cant wait for a crummy connection when making a split second decision on who to avoid, and long tunnels might also be a problem. Soall of the computing has to happen in the vehicle. He touts the Nvidia P4 self-driving car platform for autonomous in-car smarts.

FPGAs are also making great strides on the device side. Intel has Arria, an FGPA coprocessor designed for low-energy inference tasks, while over at startup KRTKL, CEO Ryan Cousens and his team have bolted a low-energy dual-core ARM CPU to an FPGA that handles neural networking tasks. Itis crowdsourcing its platform, called Snickerdoodle, for makers and researchers that want wireless I/O and computer vision capabilities. You could run that on the ARM core and only send to the FPGA high-intensity mathematical operations, he says.

AI is squeezing into even smaller devices like the phone in your pocket. Some processor vendors are making general purpose improvements to their architectures that also serve AI well. For example, ARM is shipping CPUs with increasingly capable GPU areas on the die that should be able to better handle machine learning tasks.

Qualcomms SnapDragon processors now feature a neural processing engine that decides which bits of tailored logic machine learning and neural inference tasks should run in (voice detection in a digital signal processor and image detection on a builtin GPU, say). Itsupports the convolutional neural networks used in image recognition, too. Apple is reportedly planning its own neural processor, continuing its tradition of offloading phone processes onto dedicated silicon.

This all makes sense to ABIs Orr, who says that while most of the activity has been in cloud-based AI processors of late this will shift over the next few years as device capabilities balance them out. Inaddition to areas like AR, this may show up in more intelligent-seeming artificial assistants. Orr believes that they could do better at understanding what we mean.

They cant take action based on a really large dictionary of what possibly can be said, he says. Natural language processing can become more personalised and train the system rather than training the user.

This can only happen using silicon that allows more processing at given times to infer context and intent. Bybeing able to unload and switch through these different dictionaries that allow for tuning and personalization for all the things that a specific individual might say.

Research will continue in this space as teams focus on driving new efficiencies into inference architectures. Vivienne Sze, professor at MITs Energy-Efficient Multimedia Systems Group, says that in deep neural network inferencing, it isnt the computing that slurps most of the power. The dominant source of energy consumption is the act of moving the input data from the memory to the MAC [multiply and accumulate] hardware and then moving the data from the MAC hardware back to memory, she says.

Prof Sze works on a project called Eyeriss that hopes to solve that problem. In Eyeriss, we developed an optimized data flow (called row stationary), which reduces the amount of data movement, particularly from large memories, she continues.

There are many more research projects and startups developing processor architectures for AI. While we dont deny that marketing types like to sprinkle a little AI dust where it isnt always warranted, theres clearly enough of a belief in the technology that people are piling dollars into silicon.

Ascloud-based hardware continues to evolve, expect hardware to support AI locally in drones, phones, and automobiles, as the industry develops.

In the meantime, Microsofts researchers are apparently hoping to squeeze their squirrel-hunting code still further, this time onto the 0.007mm squared Cortex M0 chip. That will call for a machine learning model 1/10,000th the size of the one it put on the Pi. They must be nuts.

We'll be covering machine learning, AI and analytics and specialist hardware at MCubed London in October. Full details, including early bird tickets, right here.

See the original post:

What sort of silicon brain do you need for artificial intelligence? - The Register

In todays age of flying cars, robots, and Elon Musk, if you havent heard of artificial intelligence (AI) or machine learning (ML) then you must be avoiding all types of media. To most, these concepts seem futuristic and not applicable to everyday life, but when it comes to marketing technology, AI and ML actually touch everyone that consumes digital content.

But how exactly are these being deployed for marketing technology and digital media? We hear about AI being applied in medical and military fields, but usually not in something as commonplace as media. Utilizing these advanced technologies actually enables martech and adtech companies to create highly personalized and custom digital content experiences across the web.

The ultimate goal of all marketers is to drive sales through positive brand-consumer engagements. But a major problem is that marketers have so much content (oftentimes more than they even realize) and millions of potential places to show it, but dont know how to determine the optimal place for each piece of content to reach specific audiences.

With all of these possible placements, it would be incredibly inefficient, if not impossible, for a human being to amass, organize, and analyze this data comprehensively and then make the smartest buying decision in real time based on the facts. Trying to test an infinite number of combinations of creative ideas and placements is like solving a puzzle that keeps adding more and more pieces while you are trying to assemble them.

So how can marketers put this data to work to efficiently and distribute their content across the digital universe using the right messaging to drive the best results?

Human beings can make bad decisions based on incomplete data analysis. For example, someone might block a placement from a campaign based one or two prior experiences with incomplete or statistically insignificant data, but it actually may perform very well. An optimization engine can leverage machine learning to understand the variance in placement performance by campaign and advertiser vertical holistically. This is why computers are simply better than humans at certain tasks.

This does not discount the value of humans, for superior customer service and relationships will always be critical. But the combination of human power plus machine learning will yield a much better result, not only in marketing technology but across all industries that are leveraging this advanced technology.

Machine learning and AI address the real inefficiencies present in digital media and have made tremendous progress pushing the industry toward personalization. Delivering personalized content experiences to todays consumer is incredibly important, especially given the always-on, constantly connected, multi-device life that we all lead.

The power of machine learning and artificial intelligence lies in their ability to achieve massive scale that is not otherwise possible, while also maintaining relevancy. This demand for personalization escalates the number of combinations that would need to be tested to an unimaginable degree. For example, if a marketer wants to build a campaign with a personalized experience based on past browsing behavior, it becomes difficult to glean insight from the millions of combinations of the context in which their advertisement will appear and the variety of different browsing behaviors people exhibit. Even with fast, granular reporting, it is impossible to make all the necessary adjustments in a timely manner due to the sheer volume of the dataset.

Furthermore, it is often impossible to draw a conclusion from the data that can be gathered by running a single campaign. A holistic approach that models the interaction between users and a variety of different advertising verticals is necessary to have a meaningful predictor of campaign performance. This is where the real impact of a bidder powered by machine learning lies, because individual marketers are not able to observe these trends due to the fact that they may only have experience running campaigns in a specific vertical.

An intelligent bidder determines how each placement has performed in previous campaigns. If one specific placement performed poorly for multiple advertisers with similar KPIs, similar advertisers in the future will not waste money testing that placement. The learning happens very quickly and precisely. Instead of humans taking these learnings and adjusting the algorithms, the technology is making the changes as they are detected.

By leveraging the billions of historical data points from digital campaigns, predictions are made for future campaigns and then real-time performance data is applied to revisions. This is not a one-off process. The technology is constantly taking insights from user behavior and feeding them back into the algorithms, enabling personalized content experiences at scale.

The advertising industry has faced major challenges in relevancy for consumers and brand safety for marketers. Lack of relevancy in advertising has led to the advent of ad blockers and poor engagement, causing brands to become even more unsure of where their budgets are going and how users are responding to content. The controversy around brand safety further calls into question not only how budgets are being spent, but potential negative consequences for a brands image.

Machine learning holds the promise of overcoming these challenges by delivering better, smarter ads to engaged consumers and restoring trust for brands in advertising spend and the technology that executes content and media.

Kris Kalish is the Director of Optimization at Bidtellect, a native advertising platform.

Link:

AI is impacting you more than you realize - VentureBeat