Humans are programmed to speak slowly and carefully to voice recognition software. Clear, simple inquiries typically provide the response desired. Speaking with a lack of clarity delivers an unsatisfactory result. Why, then, is Google CEO Sundar Pichai confident that tech companies have the ability to produce AIs with conversational ability in the near future?
Companies like Apple, Google, and Amazon are actively pursuing the title of first provider of next-generation AI that:
Go beyond pre-programmed responses to rehearsed questions.
Can remember the first question, accept a second question related to the first, and act on both.
Memory storage of the user’s personalized data and previous inquiries will allow the response to be more than an educated guess. Understanding the language database lets the advanced AI sort through stored data to find a suitable answer.
Currently, data-mining AIs teach other AIs to speak. DeepMind is used by Google to analyze the customer’s style of asking questions, web searches, and following links. DeepMind then teaches Google Assistant how to understand and use the customer’s vocal inquiry to research and find the correct response.
Facebook uses a data-mining AI called DeepText to search Messenger chat and FB posts. The intent is to provide a real time connection where customers ask AI assistants questions.
Consumers can expect the question and answer experience to improve over time. Actual conversation between humans and AI depends on a major breakthrough in technical capability and is not something to expect in the next few years.