Recently, a robot called ‘Watson’ that was developed through a collaboration between Tokyo University and IBM is receiving attention for its ability to study and learn from a large amount of medical journals and for the large contributions it made in the treatment of female leukemia patients.
AlphaGo, the Google Go game program, also showed us that the era of artificial intelligence is not as far away as we once thought with its face off against Go champion Saedol Lee earlier this year.
Artificial intelligence is no longer just something we see in science fiction movies. But when did artificial intelligence truly get its start?
The first use of the term artificial intelligence can be traced to a conference held at Dartmouth University in 1956. During the Cold War, great strides were made in the development of artificial intelligence and artificial intelligence began receiving attention from the public when the Deep Blue super computer, developed by IBM, defeated the world chess champion. With the advent of the Internet and the mass amount of information that is now attainable, artificial intelligence has seen a great amount of development in the field of engineering.
Today, we will discuss elements of artificial intelligence known as ‘Machine Learning’ and ‘Deep Learning’, and take a look at the artificial intelligence engines that make them possible.
One area of artificial intelligence known as machine learning allows machines to read text and learn from it. This technology analyzes vast volumes of big data and forecasts what will happen in the future.
Machine learning is remarkable in that it can be designed with complex algorithms and programming to support a wide range of industries. There are 3 methods for creating machine learning algorithms.
The first method is the supervised learning method, which outputs data that has been previously entered by a user. Because the data entered into the system is of precise set values, even though the data that is output is greater in volume, it is predictably accurate and precise as well.
The next method is the unsupervised learning method. This method allows a computer to learn information on its own and then produce output data. Unsupervised learning requires a high level of arithmetic capabilities and it is often used in the field of data mining.
Finally, the reinforcement leaning method allows a computer to learn information on its own and receive feedback after outputting data to create algorithms. AlphaGo implemented this method during its Go match against the global Go champion.
Machine learning uses the methods described above to learn information and perform pattern analysis to make decisions based on continuously changing conditions.
Deep learning is commonly thought of to be the same as machine learning. Deep learning is made up of a multi-level artificial network and is one of the many types of learning methods included in machine learning. It is unique in that it can make its own assessments of a diverse range of environments and circumstances without direct prompting from an engineer.
As with machine learning, deep learning processes vast volumes of data to produce reliable results. Through understanding of data patterns, deep learning can analyze not only shapes and status but also abstract objects as well.
Deep learning technology was demonstrated at the CVPR (Computer Vision and Pattern Recognition) 2015 convention showing its ability to recognize the content of an image and provide an explanation in sentence form. This technology is currently being developed in several countries around the world.
Google AlphaGo and Facebook are also being used in fields such as face recognition and these technologies are expected to be implemented in self-driving cars to detect objects and obstacles.
As the artificial intelligence industry develops, various types of artificial intelligence engines are also being developed. Many companies are using open source solutions in order to overcome the limits of artificial intelligence development. The following technologies are open source artificial intelligence engines that are open to the public.
① Google TensorFlow
TensorFlow is a Google machine learning engine opened last year that offers an open source library. As most users are currently using C++ and Python coding languages, TensorFlow is a machine learning technology with a high level of accessibility.
TensorFlow also makes it possible to analyze data on mobile devices, which makes it easier to use commercially. However, the process is limited due to the fact that data is not opening provided and users must fully understand and construct the system directly.
TensorFlow uses both Google Photo and AlphaGo directly. Recently, it even composed a song on the piano. Google is also developing machine learning hardware similar to TPU (Tensor Processing Unit) and making efforts to develop artificial intelligence in a wide range of fields.
② Microsoft DMTK (Distributed Machine Learning Toolkit)
Last year, Microsoft made their open source DMTK (Distributed Machine Learning Toolkit) available to the public. Just as the name states, DMTK is a machine learning distribution system that makes it possible to perform tasks on multiple nodes at the same time.
Also, the C++ based software development kit makes it easy for users to construct artificial intelligence systems.
When Microsoft made DMTK public, they also made the MPI and ZMQ libraries available and included an important algorithm in the source code, which allows the user access to a vast number of parameters.
Currently, Microsoft’s artificial intelligence technology based cloud platform, Azure, possesses the second largest market share in the world.
③ Facebook Torch
I think we have all experienced being tagged on Facebook with our faced being recognized in a picture automatically. This feature is made possible through a technology called Facebook DeepFace. DeepFace has an accuracy rating of 97.25%.
Torch is an open source machine learning engine that implements Facebook deep learning with artificial intelligence capabilities. This platform is particularly suitable for startups looking to capitalize on image analysis and advertizing on Facebook.
At the beginning of this year, Facebook announced the release of an artificial intelligence engine called UETorch that is a combination of Torch and the global gaming engine, Unreal. UETorch is expected to improve the security of technology through implementation on artificial intelligence based virtual gaming environments.
While machine learning technology is becoming more popular, high-performance hardware is also receiving a lot of attention. Intel is focusing on the artificial intelligence hardware market with the development of their Intel Xeon Phi Coprocessor, which combines the strong points of both CPUs and GPUs.
With the development of this processor, competition will increase with the world’s top GPU manufacturer, NVIDIA. Artificial intelligence is making waves in other diverse fields as well such as Comcast’s machine learning content recommendation system.
LG Electronics and 6 other Korean companies invested USD 3M each to establish an ‘Intelligence Information Research Center’ that opened in October of this year. The goal of this research center is to support a national project in Korea where USD 75M will be invested to develop the artificial intelligence industry and focus on the implementation of artificial intelligence.
LG CNS has also announced that they have succeeded in developing an Android SDK for the human form robot known as ‘Pepper’ and they will be supplying it to the Pepper manufacturer.
As artificial intelligence continues to grow in popularity and become more a port of our daily lives, there are also risks that should be considered. For example, when a self-driving car has a critical malfunction, how will the issue be managed and how can people’s personal information be protected?
Of course, artificial intelligence will make our lives easier but there is a need for policies and technological development to address the potential dangers of implementing artificial intelligence into our lives.
As artificial intelligence continues to be developed, it will be fascinating to see how it changes our world. Keep a look out for where you will see artificial intelligence emerge next!
Written by Yonghoon Kim, LG CNS Student Reporter
 Deep Blue was a chess-playing computer developed by IBM. It is known for being the first computer chess-playing system to win both a chess game and a chess match against a reigning world champion under regular time controls. [Source: https://en.wikipedia.org/wiki/Deep_Blue_(chess_computer)] [back to the article]