Artificial Intelligence (AI) has gained tremendous popularity over the recent years. AI can be defined as any device that understands its environment and interacts with it in order to succeed at a predefined goal. The definition of AI has also evolved over time. For example, optical character recognition is no longer regarded as AI since the technology has matured enough.
The main problems of Artificial Intelligence (AI) include reasoning, planning, learning, recommendation and natural language processing. General intelligence (human) is still considered a long-term target. Usually domain specific goals are targeted. For example, as a recent achievement for AI Google’s “Deep Mind” can be given. The researchers have built a system that could play the Chinese game “Go”. The system just learns from experience by playing against older versions of itself after it is initially trained on videos of amateur human players. World’s best human player of “Go” is defeated by this AI system. One thing that differentiates “Go” from Chess is that number of possible moves is many many more times, almost infinite. Therefore, memorizing past games as in chess is not possible.
The use of Artificial Intelligence (AI) in real-time applications such as virtual assistants and dialogue systems should be carefully designed. For example, you may wish to use past customer-agent dialogue transcripts in order to automatically generate a response for a customer request. In this case the AI system could generate reasonable answers from various agent responses, however the persona of the system may create a confusion. Because speaking styles of different agents can be significantly different and this may cause the AI system to respond incoherently. Another disadvantage may be that some of your agents may not reply appropriately to customers, and the AI system could repeat this behavior. In some cases, if the AI system also collects unstructured answers from internet it may end up praising or recommending your competitor product in the Marketplace.
Therefore, companies should realize that AI should be used with caution. In general, an AI-human combination may give better results for some domains at least for the time-being.