دانلود کتاب استراتژی داده و ارزش‌آفرینی هوش مصنوعی برای رهبران داده، نوشته‌ی رهبران داده

We have been living with data and artificial intelligence (AI) for decades. AI needs data as fuel. We know and use data. We know about and have been subjected to AI predictions (e.g., bank loans, insurance claims, credit card transactions, and hospital services). However, two recent events changed the way most people viewed data and artificial intelligence — COVID-19 and generative AI. COVID-19 highlighted the importance, power, and limitations of data. Data are important as the critical ingredient for understanding the actual situation. How contagious is the coronavirus? How deadly is COVID-19? What is the difference between COVID-19 and the seasonal flu? Expert opinion is one thing, but data provide the foundation. Experts need data to form their opinion and justify their recommendations. The public needs data to understand the actual situation and accept/reject assumptions, information, and actions to take. The need for timely and accurate data is felt more acutely now than before. The lack of data perpetuates powerlessness and hearsay. Underestimated data build complacency; overestimated data spread exaggerated fear. Decisions by governments, hospitals, businesses, and citizens need timely, accurate, relevant data. The quality of data could mean the difference between life and death, prosperity and bankruptcy. However, the lack of data could be as destructive as wrong data. Thus, the power of data is felt in times of great stress and uncertainty. In response, in addition to governments and international bodies, some universities, medical centers, and data repositories started collecting and publishing data on COVID-19. But data collected by one party might not be relevant to or sufficient for use by another. Often, the purpose and objectives of different parties may not be the same. The parties providing and using the data may not even know each other. Sooner or later, one may realize that data definitions are not the same, even if different parties use the same variable name. This can happen for data in different systems and even data that come from the same database. Furthermore, some units of measurement are inconsistent. If a data provider cannot provide all the data that one requires, perhaps we could combine our data with data from a third party? But this is easier said than done. Different datasets on the same entities cannot be merged due to a lack of common matching criteria. We cannot see the full picture with isolated pieces of datasets. Thus, we know that important information exists but it is locked and fragmented. There must be a way to see the full picture, but we do not know how, even if we can download related datasets from multiple sources. While we are empowered by data, we are also limited by data and the way they are recorded. Gaining access to data is one thing, but extracting insights from data is another. The devil is in the details and in the implementation of data systems. But the problems with data that we face today seem to be the same as those faced by data leaders 100 years ago, albeit at a greater volume and velocity now. Hence, this book aims to rephrase the problems of our time and provide ideas and solutions that could help forge better management of data at present. In contrast, our ability to extract insights from data has grown by leaps and bounds, especially with the arrival of the machine learning and AI era. In modern societies, we are subjected to the results of machine learning and AI models in our everyday life, even if we do not realize it. A model determines whether someone gets a bank loan or not. A model determines the insurance premiums someone has to pay after another model agrees that he/she is of acceptable risk. A model determines how long a patient can wait to see a doctor at the emergency room. Recently, a US judge decided on a sentence based on the results of a model. We are getting better at extracting insights from data by using machine learning and AI models invented in the last few decades. AI and machine learning have progressed by leaps and bounds, and now there is a new star in the AI space — generative AI. Generative AI, and ChatGPT in particular, is very popular because it is proving the usefulness of AI and anyone can use it. Auto-generated text, sounds, and pictures are useful in many ways to many people, but the excitement is because anyone can use it for free, thereby offering endless potential. We are no longer just subjected to AI, but we could use AI ourselves to do work faster, better, and cheaper. Data leaders who have not been following the progress of machine learning and AI will have a lot to catch up on in order to understand the usefulness, potential, limitations, and dangers of non-generative AI and generative AI. Hence, this book is an attempt to explain in simple terms and in a business context the use and potential of non-generative and generative AI after which the limitations and dangers can be inferred.