Prediction Machines

But a prediction is not a decision. Making a decision requires applying judgment to a prediction and then acting. Before recent advances in machine intelligence, this distinction was only of academic interest because humans always performed prediction and judgment together. Now, advances in machine prediction mean that we have to examine the anatomy of a decision

better than machines when understanding the data generation process confers a prediction advantage, especially in settings with thin data. We describe a taxonomy of prediction settings (i.e., known knowns, known unknowns, unknown knowns, and unknown unknowns) that is useful for anticipating the appropriate division of labor.
Prediction machines scale. The unit cost per prediction falls as the frequency increases. Human prediction does not scale the same way. However, humans have cognitive models of how the world works and thus can make predictions based on small amounts of data. Thus, we anticipate a rise in human prediction by exception whereby machines generate most predictions because they are predicated on routine, regular data, but when rare events occur the machine recognizes that it is not able to produce a prediction with confidence, and so calls for human assistance. The human provides prediction by exception

Humans, including professional experts, make poor predictions under certain conditions. Humans often overweight salient information and do not account for statistical properties. Many scientific studies document these shortcomings across a wide variety of professions. The phenomenon was illustrated in the feature film Moneyball.
Machines and humans have distinct strengths and weaknesses in the context of prediction. As prediction machines improve, businesses must adjust their division of labor between humans and machines in response. Prediction machines are better than humans at factoring in complex interactions among different indicators, especially in settings with rich data. As the number of dimensions for such interactions grows, the ability of humans to form accurate predictions diminishes, especially relative to machines. However, humans are often

Prediction machines utilize three types of data: (1) training data for training the AI, (2) input data for predicting, and (3) feedback data for improving the prediction accuracy.
Data collection is costly; it’s an investment. The cost of data collection depends on how much data you need and how intrusive the collection process is. It is critical to balance the cost of data acquisition with the benefit of enhanced prediction accuracy. Determining the best approach requires estimating the ROI of each type of data: how much will it cost to acquire, and how valuable will the associated increase in prediction accuracy be?
Statistical and economic reasons shape whether having more data generates more value. From a statistical perspective, data has diminishing returns. Each additional unit of data improves your prediction less than the prior data; the tenth observation improves prediction by more than the one thousandth. In terms of economics, the relationship is ambiguous. Adding more data to a large existing stock of data may be greater than adding it to a small stock—for example, if the additional data allows the performance of the prediction machine to cross a threshold from unusable to usable, or from below a regulatory performance threshold to above, or from worse than a competitor to better. Thus, organizations need to understand the relationship between adding more data, enhancing prediction accuracy, and increasing value creation

Machine learning science had different goals from statistics. Whereas statistics emphasized being correct on average, machine learning did not require that. Instead, the goal was operational effectiveness. Predictions could have biases so long as they were better (something that was possible with powerful computers). This gave scientists a freedom to experiment and drove rapid improvements that take advantage of the rich data and fast computers that appeared over the last decade.
Traditional statistical methods require the articulation of hypotheses or at least of human intuition for model specification. Machine learning has less need to specify in advance what goes into the model and can accommodate the equivalent of much more complex models with many more interactions between variables.
Recent advances in machine learning are often referred to as advances in artificial intelligence because: (1) systems predicated on this technique learn and improve over time; (2) these systems produce significantly more-accurate predictions than other approaches under certain conditions, and some experts argue that prediction is central to intelligence; and (3) the enhanced prediction accuracy of these systems enable them to perform tasks, such as translation and navigation, that were previously considered the exclusive domain of human intelligence. We remain agnostic on the link between prediction and intelligence. None of our conclusions rely on taking a position on whether advances in prediction represent advances in intelligence. We focus on the consequences of a drop in the cost of prediction, not a drop in the cost of intelligence

Prediction is the process of filling in missing information. Prediction takes information you have, often called “data,” and uses it to generate information you don’t have. In addition to generating information about the future, prediction can generate information about the present and the past. This happens when prediction classifies credit card transactions as fraudulent, a tumor in an image as malignant, or whether a person holding an iPhone is the owner.
The impact of small improvements in prediction accuracy can be deceptive. For example, an improvement from 85 percent to 90 percent accuracy seems more than twice as large as from 98 percent to 99.9 percent (an increase of 5 percentage points compared to 2). However, the former improvement means that mistakes fall by one-third, whereas the latter means mistakes fall by a factor of twenty. In some settings, mistakes falling by a factor of twenty is transformational.
The seemingly mundane process of filling in missing information can make prediction machines seem magical. This has already happened as machines see (object recognition), navigate (driverless cars), and translate

Organizations can exploit prediction machines by adopting AI tools to assist with executing their current strategy. When those tools become powerful, they may motivate changing the strategy itself. For instance, if Amazon can predict what shoppers want, then they may move from a shop-then-ship model to a ship-then-shop model—bringing goods to homes before they are ordered. Such a shift will transform the organization.
As a result of the new strategies that organizations pursue to take advantage of AI, we will be faced with a new set of trade-offs related to how AI will impact society. Our choices will depend on our needs and preferences, and will almost surely be different across different countries and cultures. We structured this book in five sections to reflect each layer of impact from AI, building from the foundations of prediction all the way up to the trade-offs for society: (1) Prediction, (2) Decision making, (3) Tools, (4) Strategy, and (5) Society

Economics offers clear insights regarding the business implications of cheaper prediction. Prediction machines will be used for traditional prediction tasks (inventory and demand forecasting) and new problems (like navigation and translation). The drop in the cost of prediction will impact the value of other things, increasing the value of complements (data, judgment, and action) and diminishing the value of substitutes (human prediction)

Predictions affect behavior. They influence decisions

The Chinese government experienced its AI moment when it witnessed DeepMind’s AI, AlphaGo, beating Lee Se-dol, a South Korean master of the board game Go, and then later that year beating the world’s top-ranked player, Ke Jie of China. The New York Times described this game as China’s “Sputnik moment.”2 Just as massive American investment in science followed the Soviet Union’s launch of Sputnik, China responded to this event with a national strategy to dominate the AI world by 2030 and a financial commitment to make that claim plausible