CEOs are confronted daily with the speed of technology evolution, under pressure to transform organisations to ensure they stay competitive, relevant and prepared for the next wave of change.

Business leaders are also challenged to be informed about new technology and resources and to determine the best for their organisations. All while also maintaining company culture and improving productivity.

Artificial intelligence (AI) is an area where management is certainly encouraged to have a strategy. But it can be bewildering to stay on top of what’s happening and try to understand how it can be applied in a business setting. The boardroom can feel quite far from the lab!

Several recent AI breakthroughs are already available for businesses to apply. Here are three recent AI advancements — all in the area of deep learning, the most advanced branch of AI– and how they can be applied to business.

Deep learning in visual recognition

Before 2012, AI was effective in certain scenarios such as detecting faces and recognising the side-view of cars. However, at that time, the typical amount of visual data available was small– only a few thousand images had been labelled and categorised.

2012 was the breakthrough year in visual recognition demonstrated by ImageNet- a robust data set of more than 14 million images in more than 21,000 categories, developed from 2007-2010 by a team of scientists including me.

This development showed that computers can leverage huge amounts of training data to improve visual recognition performance towards that of humans. By 2015, computers were able to surpass human performance in recognising 1,000 object categories in ImageNet.

There are a number of exciting ways that industries such as transportation and healthcare are using this technology: to better perceive the environment surrounding an autonomous vehicle, and better identify anomalies such as tumours and plan the subsequent treatment.

For traditional businesses, particularly those that are ‘document-heavy’ such as banking, insurance and law, visual recognition powered by deep learning can identify huge numbers of documents at speed, digitise and categorise them.

Where a human must add information manually, a computer can review and file several thousand (or more) documents an hour. It can flag a human reviewer to verify the information only for those that might be distorted or particularly complex. This frees up time for human workers to focus on more critical and creative tasks.

Reinforcement learning

In 2016, world champion Go player Lee Sedol was beaten at the game by Google DeepMind’s AlphaGo computer program. Go is a strategic game that requires players to place their game pieces (stones) in the right area of the board to stop the other player from advancing.

The technology was able to do this because it had practiced millions of times to improve its win rate through reinforcement learning. It looks at which actions software should take in a given environment or situation to result in a reward or positive outcome (in the case of AlphaGo, winning the game).

For businesses, this breakthrough can support improvement in areas that involve resource allocation.

For example, large technology companies with large data centres need to ensure consistent quality while also reducing power consumption. Reinforcement learning can automatically allocate which machines should perform a task, while also changing the appropriate cooling settings at the same time. As in the game of Go, it’s about ‘where to place to stone’ for the best outcome.

Reinforcement learning can also be valuable for any company that has a logistics component within its business. In the case of delivering or transporting goods or people (shipping companies, rideshare services or food delivery providers, for example), organisations have typically made resource allocation decisions by looking at past patterns and experience.

However, by adding reinforcement learning, these businesses can now predict future resource allocation, ensuring the best action for any given situation (i.e. more drivers or riders in times of heavy traffic, high demand).

Deep learning models for language recognition

Like visual recognition, it’s only in recent years that AI has become able to understand the raw text in multiple languages, pair similar words together, and identify when the same words have multiple meanings (such as ‘Let’s go out’ vs. ‘Let’s play Go’).

This capability opens a wide range of possibilities for business leaders who want to better understand how customers, staff or other stakeholders interact with written information about or relevant to the organisation.

One area of business where this can be particularly valuable is marketing. Marketers are adept at collecting data on consumers’ digital footprint– what they search, where they browse and shop, etcetera. Traditionally, a marketer sees that a customer visits travel, fashion website, and finance websites, setting broad categories for her interests.

With improved language recognition, marketers can gather more insights from the text a consumer is seeing when she lands on a web page. The technology can ‘read’ specific words, drawing out correlating terms such as ‘fashion show’, ‘Paris’, ‘air miles rewards’. With this insight, marketers can more effectively target content and messages.

What it all means

The technology reviewed here already features in business-ready solutions, and executives in the midst of making decisions about AI should confirm that their solution providers are using the most cutting-edge AI technology available.

There will be many more advancements in AI, but business leaders cannot wait to make decisions to ensure they have the ‘latest’ thing. Rather, they need to work with technology and talent that have the flexibility to adapt and upgrade as AI advances.

Senior management themselves should find ways, through lots of reading and listening, and working with educated members of staff and outside experts; to stay abreast of what’s happening in AI in the lab, so that when that technology reaches the boardroom, the C-suite can make the most effective decisions about implementation to drive the business forward.

The post How AI will revolutionise the boardroom through these 3 breakthroughs appeared first on Artificial Intelligence.

Humans have misconceived robots. They are often observed as employment thieves, undermining labourers with redundancy. Artificial intelligence just aggravates their apparent danger. Machine intelligence can supplement human intelligence. We, humans, are innovative, creative, vital and strategic. Robots are more qualified for tasks that people dislike and find difficult. The assessment of tremendous amounts of information and search for patterns in that information may include interminable repetition. It would debilitate any human brain, but not robots.

However, AI’s progression can’t be achieved in confinement. The growth of intelligent automation (IA) and its continued development should be perceived alongside AI’s advancement to get a clearer perspective of where we remain as far as technological advancement.

With intelligent automation technologies, businesses can change their procedures, accomplishing higher speed and accuracy, yet in addition, automating anticipations and decisions based on structured and unstructured sources of data. The convergence of modernized AI and Robotic Process Automation (RPA), dramatically elevates the business value and competitive edge for organizations, significantly giving rise to intelligent automation processes.

Amplified Value of RPA Along with AI
As it stands now, RPA and AI have individually established themselves as pronounced and extensive technologies revolutionizing various businesses verticals. However, any technology alone has varied limitations and that is why experts suggest to couple diverse technologies for enhanced prospects. “RPA is a great technology; however, the fact is that no tool today can provide transformation across the breadth of processes. Most organizations today are using patches to piece together a broader solution to meet their needs. As a result, the number of organizations that have scaled up automation are low as compared to the total number of early adopters of automation. RPA needs to be coupled with Artificial Intelligence (AI) capabilities like Intelligent Character Recognition (ICR), Natural Language Processing (NLP), Smart Analytics, and Machine learning to provide broader level of automation. In the near future, we will see more integrated tools that can provide real transformation to organizations,” says Siddhartha Singh, CEO, Quale Infotech Pvt. Ltd.

Businesses that eventually adopt intelligent automation technologies will lead the route in their sphere of work. All things considered, it is substantially ahead of direct process automation. IA innovation has the ability to comprehend procedures that are applicable to a business’s usual way of doing things and can execute themselves in accordance. However, to make the most out of IA, it should be in sync with the defined orchestration architecture, where machine-settled decisions are evaluated by humans to yield better results.

Besides, investors, today, clearly know the potential in intelligent automation. Venture capital investment in companies related to artificial intelligence and robotics has developed more than 70% in each of the last two years, surpassing US$600 million since 2011.

With increased investment and business interest, the rapid development of intelligent automation is introducing an altogether new era of innovation and productivity. As its applications set new norms of quality, effectiveness, speed, and functionality, organizations that efficiently deploy IA might outperform contenders that don’t. If exploited earnestly, the overall impact of intelligent automation across a business could match that of the enterprise resource planning wave of the 1990s.

IA Futurism in Digital Era
“Intelligent automation overcomes the breakneck pace of digital change. It has the power to make things simpler. It can help integrate new products, services, tools, business models, alliances, ecosystems and more. Transformational leaders use intelligent automation to create a new digital world where they are masters of competitive advantage,” says Ashish Sukhadeve, CEO, Analytics Insight.

However, barring the innovation and enthusiasm brought in by modern digital technologies, the unpredictability of the present digital landscape makes comprehensive analysis and processing through manual workflows impracticable. Further, the reliability on manual concentrated procedures makes results rather slow, error-prone, despite the inefficiency associated with it. Intelligent automation exceeds these obstructions and empowers a steady, precise analysis of the method of operations, subsequently revoking the present and future challenges of a human operator.

IA futurism is not a vision or far-off tale; it is rather a workplace where the technology impacts the digital space whose potentials are yet to be fully realised by businesses. The prospective dawn of intelligent automation technology is a new breakthrough, outshining other automation technologies, while transforming the organizational operations into future-ready workstation offering demonstrable profits. IA, undeniably, is something to consider, even as an experiment, in an effort to follow the right allocation of future investments with minimal risk.

The post Forget AI; Intelligent Automation is the New Breakthrough appeared first on Artificial Intelligence.

In particular, the more potential inputs you have to an algorithm, the more out of control your scaling problem gets when analyzing its problem space. This is where MACH, a research project authored by Rice University’s Tharun Medini and Anshumali Shrivastava, comes in. MACH is an acronym for Merged Average Classifiers via Hashing, and according to lead researcher Shrivastava, “[its] training times are about 7-10 times faster, and… memory footprints are 2-4 times smaller” than those of previous large-scale deep learning techniques.

In describing the scale of extreme classification problems, Medini refers to online shopping search queries, noting that “there are easily more than 100 million products online.” This is, if anything, conservative—one data company claimed Amazon US alone sold 606 million separate products, with the entire company offering more than three billion products worldwide. Another company reckons the US product count at 353 million. Medini continues, “a neural network that takes search input and predicts from 100 million outputs, or products, will typically end up with about 2,000 parameters per product. So you multiply those, and the final layer of the neural network is 200 billion parameters … [and] I’m talking about a very, very dead simple neural network model.”

At this scale, a supercomputer would likely need terabytes of working memory just to store the model. The memory problem gets even worse when you bring GPUs into the picture. GPUs can process neural network workloads orders of magnitude faster than general purpose CPUs can, but each GPU has a relatively small amount of RAM—even the most expensive Nvidia Tesla GPUs only have 32GB of RAM. Medini says, “training such a model is prohibitive due to massive inter-GPU communication.”

Instead of training on the entire 100 million outcomes—product purchases, in this example—Mach divides them into three “buckets,” each containing 33.3 million randomly selected outcomes. Now, MACH creates another “world,” and in that world, the 100 million outcomes are again randomly sorted into three buckets. Crucially, the random sorting is separate in World One and World Two—they each have the same 100 million outcomes, but their random distribution into buckets is different for each world.

With each world instantiated, a search is fed to both a “world one” classifier and a “world two” classifier, with only three possible outcomes apiece. “What is this person thinking about?” asks Shrivastava. “The most probable class is something that is common between these two buckets.”

At this point, there are nine possible outcomes—three buckets in World One times three buckets in World Two. But MACH only needed to create six classes—World One’s three buckets plus World Two’s three buckets—to model that nine-outcome search space. This advantage improves as more “worlds” are created; a three-world approach produces 27 outcomes from only nine created classes, a four-world setup gives 81 outcomes from 12 classes, and so forth. “I am paying a cost linearly, and I am getting an exponential improvement,” Shrivastava says.

Better yet, MACH lends itself better to distributed computing on smaller individual instances. The worlds “don’t even have to talk to one another,” Medini says. “In principle, you could train each [world] on a single GPU, which is something you could never do with a non-independent approach.” In the real world, the researchers applied MACH to a 49 million product Amazon training database, randomly sorting it into 10,000 buckets in each of 32 separate worlds. That reduced the required parameters in the model more than an order of magnitude—and according to Medini, training the model required both less time and less memory than some of the best reported training times on models with comparable parameters.

Of course, this wouldn’t be an Ars article on deep learning if we didn’t close it out with a cynical reminder about unintended consequences. The unspoken reality is that the neural network isn’t actually learning to show shoppers what they asked for. Instead, it’s learning how to turn queries into purchases. The neural network doesn’t know or care what the human was actually searching for; it just has an idea what that human is most likely to buy—and without sufficient oversight, systems trained to increase outcome probabilities this way can end up suggesting baby products to women who’ve suffered miscarriages, or worse.

The post Deep Learning breakthrough made by Rice University scientists appeared first on Artificial Intelligence.