Machine Learning And Artificial Intelligence: Essential Guide To Understanding How ML And AI Can Be Applied In Practice And Be Compatible With Human Behaviour In Modern Times

by Chris Neil

Machine learning is a data-driven approach. Hence it implies the availability of large datasets in order to make accurate decisions. In case only a limited dataset is available to solve a particular problem, it is best to use a deterministic approach. When a limited dataset is available it is hard to train a machine learning model and generalize its applicability to other similar problem. The developed model, in this case, is only applicable for the few data it was trained on. Because machine learning methods rely solely only on data and the human expertise or judgment is not taken in consideration, it is the data that dictate if the machine learning method will fail or succeed to perform the task it was designed for. The way a machine learning approach works is that a modeler develops a learning algorithm. Then the modeler feeds the learning algorithm with the data and information. The algorithm learns by itself from the data with no guidance or human interference. It is the algorithm that builds the system. If the data provided for the algorithm is of poor quality and biased then the system is also of poor quality and biased. Hence, cleaning and acquiring the right data to solve a problem with machine learning problem is very crucial. If the data are biases and noisy it is better to stick with a traditional method. Otherwise, the machine learning method will memorize the noise and provide inaccurate results. Could Turning Important Decisions over to AI help Humanity? Some might say that's a terrible idea, but how much worse is it today where we've turned over our societies and civilizations to corrupt leadership, crony capitalism, or had to deal with rogue nation states in other parts of the world with two bit dictators, religious fanaticism, or the quest to destroy another group of peoples' civilization? Perhaps it's time we came up with a special think tank that could go through all the issues concerning our fears, and what we hope to expect from AI decision-making machines. That is to say how to get the best possible answer, all the time with the greatest probability. We all watched IBM's "Watson" supercomputer as it won against the top human Jeopardy players. It did pretty well, most all the time; didn't it? Yes, but it wasn't perfect, and perhaps that's the scary part. In fact, some of the mistakes that it did make were mistakes that even a child wouldn't have missed. Still, if there is a group of humans or a think tank, focus group, or Board of Directors constantly surveying the answers and asking additional questions, then perhaps you don't have to worry about an erroneous answer now and again. In fact, it might make you smile and feel good to be a human at that point. Machine learning is useful anywhere you need to recognize patterns and predict behavior based on historical data. Recognizing patterns could mean anything from character recognition to predictive maintenance to recommending products to customers based on past purchases.

174 pages, Paperback

Reviews

[Ares · Rating 3 out of 5]

This was a lot longer than it needed to be for me. When I learned it for the first time we had long discussions about AI and Ethical AI and this was bought up in this book as well. I really liked this aspect.

But if it was about pure information the book could have been a lot shorter. Elaborating on the key concepts again and then moving over to the practical aspects before ending it with a discussion about Ethics. This book lingered a lot on a few things, it was too long to be a nice surface level introduction but too short to be an in depth or even philosophical work.