Artificial Intelligence  

Certainly a main buzzword for technology and computer science today, 'deep learning' is having a major impact in many sectors.  Why is deep learning so important?

Tthere is no question that deep learning  is one of the most talked about trends in business and computer science today.  Why has this artificial intelligence technology, that can be traced back to the 1970s and 80's so much in the news today? For many the long held promise of artificial intelligence that is actually smart is coming to light through this technology.

Deep learning is used with numerous fields, and it will soon aid in manufacturing, medicine, retail, the home, and beyond.

Gartner calls it “the most significant technology shift of this decade”.

Essentially, deep learning refers to machine learning algorithms, that are programmed to generalize data and use it in a way that is suitable for the application.

Related articles Teaching Computers To See By Seeing Like A Computer  Towards Intelligent Humanoid Robots  What Will Computers Look Like By 2029?

This isn't a new concept, but for decades it has been a pipe dream, limited by computer hardware and insufficient data. Moore's Law has come to fix these shortcomings with server and data storage costs approaching zero and Big Data continuously increasing.

The implications for the not-to-distant future are enormous.

In a shopping recommendation system for a website, a trending product and indications that a user has browsed in the category of that product in the last day, they are very likely to buy the item.

These two variables are so accurate together that they can be combined them into a new single variable, or a feature. Finding connections between variables and packaging them into a new discreet variable is called feature engineering. In deep learning systems the feature engineering can done automatically.  Another common use of the technology today is automatic tagging of images, and work being done to do the same for video.

Deep learning may soon help robots like Baxter 'know' how to handle various objects without the help of human programming.

Image Recognition

In computer vision deep learning is also being found to be increasingly useful. Algorithms fed thousands and thousands of images can independently learn to identify individual components in them. Along with improving search technology, this will eventually be used to make robots navigate their way independently, and interact with objects without human assistance.

It's not about computers recognizing a particular face - we've had that for years. It's about recognizing that an object is a face to begin with. Or a car. Or a cat. In one example, a team was recently able to have their system recognize objects as well as a monkey, and it is only improving.  This rapid progress in automatic image definition is going to allow blind people to see, and get self-driving cars to work properly.

Natural Language Processing & Speech Recognition.

Speech recognition continues to get smarter.  While we aren't quite yet at the level of Samantha, from the film Her, Siri and Google Now are improving all the time.  Just a few months ago the voice search function would never work for my four-year-old.  Now he is browsing videos without help.

Skype's real-time translator is a another example of what lies ahead.

Sentiment analysis—the use of natural language processing, text analysis and computational linguistics to identify and extract subjective information— also help businesses and organizations understand consumer emotions regardless of the language they are written in. Researchers at Stanford University have a demonstration sentiment analysis tool available online.

Surely in the development of social robots, sentiment analysis will play a large role.

Predictive Algorithms

Creating deep neural networks has transformed businesses like Amazon and Netflix through intelligent recommendations, and will lead to better sales automation and lead generation, highly efficient marketing, predictive hiring, algorithmic trading.

In highly trained deep neural networks, there is, and will increasingly be an associated deep predictive tool built into the system. Some would argue that a human's ability to predict the future is one of the key parts of our intelligence.

Facebook's Yann LeCun has said that once deep learning overcomes some technical hurdles, it will open up other areas like, automatically-created high-performance data analytics systems; vector-space embedding of everything (augmented reality); multimedia content understanding, search and indexing; multilingual speech dialog systems; driver-less cars; autonomous maintenance robots and personal care robots.

Looked at it from this way, deep learning is a foundation technology on which so many future developments are built.

Who is Building the Deep Learning Foundation?

Apart from academia, (which has been plundered by the big names in tech), Google is the biggest company in this space, though Baidu will surely follow since they recently poached deep learning pioneer Andrew Ng. IBM, Microsoft, and Facebook have made great strides as well. There are a handful of smaller companies, most notably AlchemyAPI and Cortica.

For instance, Google’s DeepMind team has published their initial efforts to build algorithm-creating systems that it calls “Neural Turing Machines”; Facebook showed off a “generic” 3D feature for analyzing videos; and Microsoft researchers concluded that quantum computing could prove a boon for certain types of deep learning algorithms.

Shivon Zilis, has put together an infographic that shows what she calls the Machine Intelligence Landscape of companies involved with deep learning.

Image Source - Shivon Zilis

Deep learning advances the state of the art in pattern recognition and natural language processing, but what is important when looking at the core feature of deep learning is that it acquires generalized representations grounded in experience.

The next frontier will be building machines that can represent the deepest conceptual structures of our minds, such as what a container is, and can use that ability to understand abstract concepts through metaphor. "If we can make it all the way down so that computers have a grounded understanding of the most fundamental concepts, we will have built an intelligence that is as flexible as our own," writes Jonathan Mugan.

With the increasing exposure, deep learning experts are in high demand, and an ever-increasing range of applications for the technology is being introduced.

Moreover, while a lot of information has come out lately about how organizations are using deep learning, a lot of the work is behind the scenes or not released to the public.  Some of these developments may turn out to have the greatest impact in the future.  For many organizations, deep learning is about much more than tagging images.

In short, deep learning is moving us one step into the future, and giant leap closer to artificial general intelligence. Like it or not, deep learning really means is that we are close to living with smart machines.

By 33rd Square

Embed

 Artificial Intelligence  

According to London-based Dmitry Aksenov, salespeople, call center staff and customer service personnel could all be replaced by computers within the next few years.

Dmitry Aksenov has been working on building computers that “think like human beings” since he was 10 years old. “It is my passion,” he said.

Mr Aksenov, now 21 years old, founded technology company London Brand Management (LBM) in 2011. The company provides an AI service for big brands who want to outsource customer or staff interactions to computers. Customers send questions in to LBM’s system via email or text and it responds within five seconds.

For example, if someone asks LBM's artificial intelligence a question, the system understands it and finds an answer, and gets back in the same tone of voice; it ‘echo’s’ the sentiment.

"Within five years we will have a system that truly knows more than a human could ever know and is more efficient at delivering information."

If the same person asks five more questions, the system understands this person is engaging with the product and can notify one or more of the appropriate people in the organisation who can then follow up the customer’s interest. According to LBM, the only difference between how a real human would do this and our system is the speed of the reply and that our system is tireless – it can work 24/7/365 and can reply to a potential customer’s question with the correct information in milliseconds.

Related articles AI System Breaks CATCHA Turing Test  What Does Watson Tell Us About the Philosophy of AI?  Gary Marcus Urges People To Think About The Threat of Artificial Intelligence

"Our system, like a human, understands natural language such as questions and feedback as well a the product knowledge held by clients and in the public domain. It then finds the required information, performs programmed actions and replies as a human would," states the LBM website.

The service has been used by thousands already. “The only thing that gives away the fact they are talking to a computer is that it responds so fast,” Aksenov told The Telegraph. “No real person could receive, read and respond to a message in three seconds.”

“It not only reads the keywords and understands the kind of information you are trying to learn; it also interprets context, sentiment, and can even understand humour. It also remembers and learns as you talk to it, so it’s capable of having a proper conversation.”

This new technology represents a huge step forward in service automation, Aksenov claims. LBM’s system uses proprietary Natural Language Processing technology to identify the sentiment, meaning and mood of any messages that it receives. It ‘understands’ the received information in the context of each individual user and provides a tailored answer to a specific question.

The company is currently focused on replacing traditional sales and marketing roles but is also moving into the customer care and call center space. New projects for an British cancer hospital and a major Japanese electronics company are already under way. “There are applications for this system in hundreds of industries,” he said.

“Within five years we will have a system that truly knows more than a human could ever know and is more efficient at delivering information,” he said. “It will replace many of the boring jobs that are currently done by humans. Unfortunately, this may take some jobs from the economy by replacing human beings with a machine. But it is the future.”


SOURCE  The Telegraph

By 33rd Square

Embed

 Programming  

Stephen Wolfram has unveiled Wolfram Language, a highly developed knowledge-based programming language that unifies a broad range of programming paradigms and uses its unique concept of symbolic programming to add a new level of flexibility to the very concept of programming.

Stephen Wolfram has introduced the Wolfram Language in this video that shows how the symbolic programming language enables powerful functional programming, querying of large databases, flexible interactivity, easy deployment, and much, much more.

Wolfram Language, is totally symbolic, heavily natural, intensely knowledge-based, and extremely large computer programming language.

"Once things are symbolic," says Wolfram, "its easy to do things like Meta-operations on them."  Along with this, he demonstrates a dynamic version of a plot, generated from a sparse and easily understandable bit of code. "Because everything is a symbolic function, it all just works."

Proceeding, Wolfram demonstrates an example of graphing his web bookmarks:

Related articles Can Robots Invent Their Own Language?  Verbal IQ of a Four-Year Old Achieved in AI System  Modern Technology and Our Means of Understanding Foreign Languages

“I’ve been working towards what is now the Wolfram Language for about 30 years,” Wolfram says in the video above. “But it’s only in recent times that we’ve had what we need to create the whole Wolfram Language.”

Wolfram Language is not yet released, but will be embedded on upcoming Raspberry Pi micro-computers. It’s already widely used within Wolfram’s Mathematica computing environment for scientists, and it is also deployed to Wolfram Alpha’s cloud services as well.

Wolfram Language contains may machine learning algorithms as well, which may impact data classification dramatically.  He calls these superfunctions, that essentially let you treat many aspects of programming as a computational 'black box.'

Wolfram Language Machine Learning example

In these early days it will be interesting to see what emerges from the Wolfram Language.  These are undoubtedly powerful tools.


SOURCE  Venture Beat

By 33rd Square

Subscribe to 33rd Square