Deep Learning Leads to More AI Investments

The world of artificial intelligence has gotten a boost in recent months, driven by the entry of new players, major investment announcements, and a growing number of proven AI capabilities.

What’s emerging is a robotics and AI marketplace that’s increasingly focused on two elements: deep learning and human-robot collaboration (HRC).

Deep learning builds on neural network studies and currently represents the most ambitious and promising avenue for AI research. In addition, it’s poised to make our robots smarter and more capable than they have ever been before.

The pace of advancement in AI is “actually speeding up,” as Jeff Dean, a senior fellow at Google, told Bloomberg Business in late 2015. This is being driven by lower cloud-computing costs, which make deep-learning approaches to AI problems more feasible.

Thanks to deep learning, AI is “finally getting smart,” as MIT Technology Review put it.

Meanwhile, HRC combines the best of intelligent robotics with the best of human capabilities.

The market for AI, particularly machine learning and in Asia, will grow from $419.7 million in 2014 to $4.05 billion by 2020, according to research firm Markets and Markets.

As it becomes clearer to all just how important deep learning and HRC are to the short- to medium-term future of robotics, new players are entering these markets with enthusiasm and considerable financial backing.

Accenture invests in Irish lab

For example, Accenture is investing $25 million in a “Centre of Innovation” and an AI-focused laboratory in Dublin.

“Artificial intelligence will disrupt businesses and industries on a global scale, and we see this shift going well beyond deploying analytics, cognitive computing, or machine-learning systems in isolation,” said Paul Daugherty, chief technology officer at Accenture.

“We are investing early to drive more innovation at Accenture, recruit top talent in every location we operate in, and infuse more intelligence across our global business to help clients accelerate the integration of intelligence and automation to transform their businesses,” he added.

Meanwhile, the Accenture Technology Labs University Grant on AI at University College Dublin will fund research that explores the human-machine interface.

The research goals that Accenture has set reveal just how wide-ranging it expects the implications of these technologies to be. The company has set the following objectives:

• Create more intelligent tools to advance capabilities in cognitive computing, machine and deep learning, natural-language processing, data augmentation, and predictive analytics.
• Integrate and apply AI into front- and back-office operations, including customer support, procurement, supply chain, and warrantee services.
• Embed AI capabilities into architectures, tooling, and service management analysis.
• Design and scale AI capabilities for Accenture Consulting around the world.

“We believe that the real power of artificial intelligence is to augment what humans are great at and make them better at what they do,” Daugherty told Silicon Republic.

“It’s all about making humans more effective,” he said. “For workers with augmented technology in the form of virtual reality technologies, augmented vision combined with machine learning to help lower-skilled workers do more advanced jobs very productively. [It] creates additional opportunities for people, and that’s why we are excited by artificial intelligence.”

Accenture’s AI researchers are working with drug manufacturers to speed up the clinical trial analysis phase. Similarly, startup MedyMatch Technology Ltd. is using vision and cognitive analysis to improve diagnosis.

Big data gets a lot of deserved press coverage, but it is AI — in the form of algorithms — that turns all that data into tangible business benefits, according to Peter Sondergaard, head of research at analyst firm Gartner Inc.

“Anybody can gather data, anybody can store it,” Sondergaard said at the Gartner Symposium 2015 in Barcelona, Spain. “You may be able to do good analysis, which is worth a little more, but anybody can hire somebody to do data analysis, no matter how big the data set.”

NEC, Kodak double down on AI

In November 2015, NEC Corp. announced that it plans to double the size of its AI-focused workforce to around 1,000 over the next five years. The company intends to redeploy existing staff and make new hires in India and Singapore, all with the aim of doubling its AI-related sales to some 60 billion yen ($484 million) by fiscal 2020.

Meanwhile, Kodak Alaris has launched the AI Foundry, a business built on a proprietary, self-learning, technology that’s designed to help organizations make sense of the unstructured business data found in emails, chat logs, text messages, and so on.

“AI Foundry was established to harness the power of artificial intelligence to understand amorphous data and provide smarter ‘context aware’ information management,” said General Manager Steve Butler.

Japan tries to catch up

The Japanese government is also getting in on deep learning with the establishment of an AI research center and the announcement of plans for a second.

Japan has a sterling reputation in robotics, but it lags behind the U.S. in deep learning. As Japan Today reported, the Japanese government is keen to bridge this gap.

Nvidia has released hardware intended to help AI developers.

In November 2015, graphics chipmaker Nvidia Corp. unveiled a hyperscale data center platform with the goal of encouraging developers to build networks and smart apps rooted in AI techniques.

When the company said in February that it had beaten beat fourth-quarter earnings targets, it attributed this success to strong customer interest in its deep-learning technology.

Proven technology

In terms of proven applications for deep learning, we have seen an impact across several industries, and at least one longstanding AI challenge has been met.

Cybersecurity strategies based on deep learning have proven themselves in the realm of cloud computing, for example, helping companies to secure the ever-growing universe of data stored in the cloud.

A “cyber defense platform” combines Tel Aviv-based Deep Instinct‘s deep-learning technology with Redwood City, Calif.-based FireLayers‘ policy-based platform that is designed to monitor cloud applications and data.

Deep learning has also been shown to make driverless cars better at spotting pedestrians. A new pedestrian-detection algorithm combines traditional computer-vision classification architecture (known as cascade detection) with deep learning.

“Such a vision-based safety system has remained elusive in cars because computers typically face a tradeoff between analyzing video images quickly and drawing the right conclusions,” said IEEE Spectrum.

“On the one hand, a simple ‘cascade detection’ computer vision algorithm can quickly detect many pedestrians in certain images, but [it] lacks the sophistication to distinguish between pedestrians and similar-looking objects in the toughest cases,” said the report.

“On the other hand,” the IEEE noted, “machine-learning algorithms called deep neural networks can handle such complex pattern recognition but work too slowly for real-time pedestrian detection.”

Deep learning has also proven its worth to hedge fund managers, farmers, medical imaging specialists, and search-and-rescue drone operators.

Google Inc. has plans to incorporate deep learning capabilities into mobile devices.

Google’s AI beats Go

In perhaps the most headline-grabbing moment of all, in early 2015, a deep learning based AI developed by Google broke new ground by beating a top human player at the game of Go.

“Traditional AI methods — which construct a search tree over all possible positions — don’t have a chance in Go,” said Google on its blog. “So when we set out to crack Go, we took a different approach.”

“We built a system, AlphaGo, that combines an advanced tree search with deep neural networks,” the company said. “These neural networks take a description of the Go board as an input and process it through 12 different network layers containing millions of neuron-like connections.”

More prospects

With cloud storage costs falling and labs around the world in a race to develop new deep learning-based AI, more new entrants and further R&D investments are likely over the coming year.

As robots get smarter, they become more capable as our collaborators in the home and in our factories.

The increasing focus on deep learning and HRC is encouraging for many reasons, including increased efficiency, more natural communication between humans and robots, and taking advantage of the best capabilities of humans and machines.

These parallel trends offer a tantalizing glimpse of a robotic future in which truly intelligent collaborative robots, or cobots, come to enhance human labor, not to replace it.