Lecture 12 Flashcards
Can a Machine Think?
AI Super-Powers: China, Silicon Valley, and the New World Order by Kai-Fu Lee
“After thirty years of pioneering work in artificial intelligence at Google China, Microsoft, Apple and other companies, Lee says he’s figured out the blueprint for humans to thrive in the coming decade of massive technological disruption: ‘Let us choose to let machines be machines, and let humans be humans.’”— Forbes
“Kai-Fu Lee believes China will be the next tech-innovation superpower and in his new (and first) book, AI Superpowers: China, Silicon Valley, and the New World Order, he explains why. Taiwan-born Lee is perfectly positioned for the task.”— New York Magazine
“Both a provocative and readable distillation of the conventional wisdom on AI supremacy, as well as a challenge to it.”— Financial Times
Recent advances in algorithms and processing power, combined with the growth in available data, are enabling
the creation of machines with unprecedented capabilities
While these technologies might not redefine what it means to “think,” they are starting to perform
activities long thought to be the sole purview of humans, sometimes at higher levels of performance than people can achieve
Conventional software:
hard-coded by developers with specific instructions on the tasks it needs to execute. It works well in many situations but has limitations, as programmer cannot account for every scenario (i.e. if environment changes the program will malfunction)
AI Software:
leverages an algorithm to learn from data and adapt to new circumstances without being reprogrammed. It is based on the concept of giving the algo “experiences” (training data) and a generalized strategy for learning, then let the algo identify patterns, associations, and insights from the data (train the system vs program it!
Current State of AI
Systems enabled by machine learning can provide customer service, manage logistics, analyze medical records, or even write a news story
Data scientists have made breakthroughs that enable machines to recognize objects and faces, to beat humans in challenging games such as chess, to read lips, and even to generate natural language
Digital giants such as Google, Facebook, Intel, and Baidu as well as industrial companies such as GE are leading the way in these innovations, seeing machine learning as fundamental to their core business and strategy
Machine learning
encompasses various techniques that recognize patterns and associations in huge amounts of complex data. Regression, support vector machines, and k-means clustering have been in use for decades while others have become viable only now that vast quantities of data and unprecedented processing power are available (i.e. artificial “neural networks,” which are inspired by the connectivity of neurons in the human brain)
Reinforcement learning is
another technique used to identify the best actions to take now in order to reach some future goal. These type of problems can be useful for solving dynamic optimization and control theory problems (common issues in fields such as engineering and economics). Reinforcement learning algorithms that use deep neural networks have made breakthroughs in mastering games such as chess and Go (“deep reinforcement learning”)
Type of Problems AI Can Solve
Classification
Prediction
Generation
Business Opportunities for Machine Learning
Machine learning can be useful in a number of settings combined with conventional optimization and statistical methods
Machine Learning Has Broad Potential Across Industries and Use Cases
Top Use Cases:
Radical Personalization
Predictive Analytics
Strategic Optimization
Real-time Optimization
Radical Personalization
One of the most exciting and yet to be fully exploited capabilities of machine learning
Enormous potential for industries collecting a wealth of data about individuals, as machine learning requires large and granular data sets to train
Radical Personalization is ideal for organizations that can create
huge value by tailoring their offerings to suit the preferences, characteristics, and needs of each person they serve (i.e. health care, CPG, media, finance, education)
Radical Personalization
Netflix’s recommendation engine currently influences about
80% of content hours streamed. The company estimates that using personalization has increased subscriber retention and engagement to such a degree that it is worth some $1 billion annually
Predictive Analytics
Helps classify customers or observations into groups for predicting value, behavior, risk, or other metrics
It can be used to triage customer service calls, to segment customers based on risk, to identify fraud in banking and cybersecurity, and to diagnose diseases
i.e. One media company used machine learning to discover the factors that were most predictive of customer churn and identified that 2% of customers were causing almost 20% of overall churn
