Understanding Artificial Intelligence (AI)


Home / Understanding Artificial Intelligence (AI)

How Smart is Artificial Intelligence? Image by artM

Don’t believe everything you see in the movies.

Computers are vast repositories of human knowledge, and can follow complex instructions, but the missing parts (such as human curiosity, motivation to improve outside of commands, and an interest in topics that have not been addressed by programming directives) prevent software programs from achieving true intelligence.

What can computers do?

Computers can accomplish many advanced tasks that produce an illusion of true intelligence. Examples include:

IBM’s Watson Beat Humans at Jeopardy. Image by Raysonho

  • Translation between languages – In order to translate English to Spanish or Spanish to Greek, a computer uses the same process by which it translates human language into data. This process is called natural language processing, and it involves a huge mass of data processing, but is not an indication of intelligence.
  • Playing (and winning) games with humans – Ever since the early days of AI, programmers have been teaching computers to play games and win. From programs for checkers and chess, which require the input of rules and potential moves, to IBM’s Watson, which combined vast data storage and enormous processing power with sophisticated matching algorithms to win at Jeopardy, these winning machines have been astounding humans for decades. Current technology, however, simply assists the machine in providing the best answer or move. Until the computer is interested in the outcome, however, even the most advanced program is no more intelligent than a Rock-em Sock-em robot.
  • Conversations – Chatbots can provide interesting conversations, but there is no intelligence behind the screen. A chatbot program compares the words in your question, comment or response to a database to determine the most appropriate answer, follow-up question or response.
  • Predictive modeling – When a computer predicts future events, the prediction is based on a mathematical model that extrapolates results based on previous events.

What computers cannot do

At the most basic level, a computer can only do what it is told to do. In some cases, the computer is given complex instructions, and even given instructions to improve based on previous outcomes, but no one has yet succeeded in programming a computer to act alone, or to want to accomplish a task for which it has not been programmed. Interest, motivation and curiosity are all missing from AI programs, a lack that prevents them from achieving true intelligence.

Watch Watson perform on Jeopardy:

Leave a Comment