Machine Learning (“ML”) is the pivot of the popular subject of artificial intelligence that allows for learning more and more of everything familiar to humans like the natural processing of languages and identification of graphic images. Through ML, machine will get smarter and could engage in interactions with human beings more naturally. Likewise, it could do a lot more for humans.
Machine Learning is critical for the development of artificial intelligence
AlphaGo Master competed with a Go chess (Weiqi) contestant in 60 rounds of chess games and successfully won the competition. This unveiled the development of artificial intelligence in specific application has already surpassed the level of humans. AlphaGo is just based on the content of chess manuals accumulated in human history for thousands of years for giving prompt command through computing techniques. The result is amazing and expected to perform much better than humans in the future. The new generation of AlphaGo Zero is based on a thorough understanding of the rules of Go Chess without consulting chess manuals of humans, and has successfully beaten its counterpart, the AlphaGo Master, which was the champion that no human being could compete in just 3 days. This result demonstrates the stunning learning capacity of machine.
If ML is designed to beat humans in games thereby undermines their confidence, it would not be meaningful for the development of this branch of science. Indeed, the purpose of ML is to help to improve human life through the augmentation of human capacity equivalent or superior to humans, including the natural processing of languages that allows for verbal communications with humans, data mining, machine vision, identification of biological features, search engine, medical diagnosis, detection of credit card fraud, analysis of securities market, DNA sequencing, voice and handwriting identification, strategic game, and robots.
ML is a branch of learning from artificial intelligence, which is also another approach to realize artificial intelligence. In other words, ML is the mean to solve the problems in artificial intelligence. After more than 3 decades of development, ML has emerged as an interdisciplinary subject involving probability, statistics, and computational complexity theory. ML allows for the automatic analysis of data for deducing certain rules basing on which the unknown data will be subject to forecasting.
In robotic science, for example, we must let a robot to learn human languages, engage in conversation and interact with humans so that it could identify the object being scanned through machine vision. It will be essential to determine if the object detected is a human or a dog, or just a table. It is also essential to allow the robot to differentiate the changes in the facial expressions of humans and give accurate reciprocal responses, which will be a lot more than just an understanding of languages.
In addition, ML also helps to examine the content of the pictures and films uploaded by humans to the Internet and determine if there is pornographic or violent content and give alert by classification. Or, it could help to detect the suicidal inclination of specific person, or such person in inclined to terror attack. After all, it could help to prompt for preventive and preemptive action to prevent the tragedy.
In commercial application, ML also helps to match the changes in the stock market and the foreign exchange market over the years to deduce the rules of the changes. If specific condition was satisfied, ML will help to prompt timely suggestion to assist investors make favorable choices of investment. In application to medical care, ML allows for the matching of big data of medical history to assist physicians in professional medical diagnosis. All these will help to improve the quality of human life in the future.
ML has been around for decades, but there are just two relatively new trends that triggered the extensive application of ML: the usability of massive training data and the powerful and high performance combined computing capacity given by GPU computation. On the software side, ML relies on a variety of advanced algorithmic methods and big data analysis. On the hardware side, ML relies on high-speed processor and cloud computing to accelerate the speed in data analysis and processing. Scientists in the industry and academic circle have long been using GPU in proceeding ML and have achieved frontier development in various forms of application like graphic image classification, content analysis, voice recognition and natural language processing. This is particularly the case in Deep Learning, which is based on the use of complex, multi-layer “deep” neurotic network to create a system capable of handling misuse detection from massive unmarked training data. This major discovery has already attracted sizable investment and effort in further research. GPU adopted an even bigger training set to train these deep neurotic networks that helps to reduce the time for cloud computing substantially. It also occupies a lesser proportion of the infrastructures at the data center. GPU is also used for the computing of the ML models under training and the classification and forecasting in cloud computing, which allows for supporting the import and export of bigger data volume with the consumption of less energy and lesser proportion of infrastructures.
In the past, the users of GPU accelerators in ML were big Internet and community media firms and top research institutions in information science and ML. As compared with CPU, GPU has emerged as the first choice of data scientists in processing big data, given its thousands of kernels in computation and 10-100 times import and export quantity of application programs.
NVIDIA, a leader of GPU, foresees the future of ML in commercial application thereby launched the NVIDIA DiGiTS DevBox. This is the fastest desk side depth learning system. DiGiTS DevBox provides the degree of freedom in searching a diversity of network structures and the acceleration of big data processing so that you could have a powerful, energy efficient, cool, and quiet solutions in computing by the side of your desk.