Therefore the agent must ask questions - as would a machine, for example: “Do you mean you can’t access your transfer page or you can’t access your main page? If you can’t access the main page are you looking for an account that’s linked to the main page or an unlinked account? Have you added your social security number?” Let’s say you call a human agent and say, “I can’t access my bank account.” That statement is contextually ambiguous and not clear enough on its own for an agent to understand what corrective action to take. Problem solving is the most complex aspect of any artificial intelligence task and the most critical measure of value. That leads us to the final criterion for realizing true machine intelligence. They must now apply their knowledge to solve problems. It should be able to read a manual and answer questions about that manual.Īs children grow into young adults, they are no longer evaluated on reading comprehension. A machine should be able to pass a fifth grade reading test. It is because of an intuitive and unrivaled capacity to learn.Ĭhildren are evaluated by testing their reading and comprehension skills. The reason why humans are the most advanced and dominant species is not because of keen eyesight, discerning sense of smell or impeccable hearing. Once a machine has mastered understanding, the next step is learning. A search engine would only answer the question of who bought the stock, not who owns the stock.
For example, take the statement, “John bought stocks from Jane yesterday.” If I ask a person, “Who holds stock today?” he or she would likely reply, “John, because he bought stock from Jane yesterday.” If you ask a search engine, “Who owns the stock today?” or “Who owned it last week?” you would not get an answer.
To achieve true artificial intelligence, a machine first must be able to understand and interpret what the user means. A strong exercise in parsing natural language, he could understand that the words used related to a home or travel destination.īut did he really understand what was meant? Did Eugene learn from the interaction, and demonstrate problem solving skills or was he only mimicking understanding? After 75 years, it’s time to return to the spirit of Turing’s question and expand it to “Can a machine understand, learn and solve problems?” Last month, a chatbot named Eugene Goostman reportedly passed the Turing Test by convincingly imitating a 13-year-old boy. In 1950, Alan Turing proposed the question, “Can machines think?” and since then many have strived to pass the Turing Test, which sets out to see whether humans can discern whether they are interacting with a human and a machine.
0 kommentar(er)
