Artificial intelligence (AI) is frequently the focus of technology news. There is a marked tendency, however, to treat AI as if it were a singular thing.
The Varieties of AI
It isn't. A recent article in Wired points out that there are many varieties of artificial intelligence. If you hear of deep neural networks, for example, you are reading about a pattern recognition system. Facebook, Microsoft, and Google all use neural networks.
Another form of AI uses probabilistic programming. This type of AI uses the hypotheses of programmers, and then use data generated by AI to test and refine those hypotheses.
Probabilistic programming can be used in translation, for example. It can also be used to cull meaning from raw data. Ben Vigoda, the chief executive officer and founder of Gamalon, a probabilistic programming AI company, believes that form of AI crunches information and data more rapidly that neural networks. It can also learn from much smaller data streams.
Evolutionists try to replicate the rules of natural selection.
Symbolists code knowledge and the rules with which that knowledge can be assimilated by machines.
The future likely belongs to AI blends.
Using probabilistic programming allows scientists to create algorithms for machine learning just like computer scientists create computer code into programs. But it can also provide researchers with insight into why and how an AI system makes the decision it does. As a result, they can refine its decision-making processes.
Neural networks excel at speech and image recognition. Those processes are not in competition with the kind of research into decision-making that probabilistic programming allows. Companies are increasingly combining the two.
What Businesses Use AI For
How do businesses use AI? A recent Forbes article pinpointed four ways in which organizations will deploy AI, both now and in the future.
The first is chatbots used for customer service. Those are already operating in the websites of many companies and can be particularly useful for frequently asked customer questions.
The second is crunching of data. This can be specifically created databases, such as responses to customer surveys.
Intriguingly, though, AI can also be used to crunch data for specific purposes. A company looking for energy sources, for example, may have tons of data on drilling sites, practices, and history. But the data may exist in far-flung locations and sources, like technical manuals and maps long out of use. AI can be fed this data to find usable information and patterns.
The third is automation of manual processes. Factories, which are increasingly using robots, are often given as examples. Forbes observes that the first demarcation is likely to be routine work versus nonroutine work. Box scores and financial summaries in journalism, for example, are routine - they involve reporting facts and figures on what has happened. Analysis and reporting are nonroutine work unlikely to be done by robots.
The fourth is using unstructured data. Customer service centers, for example, may collect data on customer and employee personalities with the eventual aim of matching customers with employees who suit their own personalities.
AI types differ by methodology. In the future, it is likely that multiple methods will determine a given AI system. Companies use AI for a variety of purposes as well, and the purposes are likely to grow in the future.