MIT Plans to Reinvent the Field of Artificial Intelligence

It's been 59 years since Alan Turing created the Turing Test that would be the most accepted test to see if an artificially intelligent machine could communicate and fool a human into thinking the machine is a human too. Since then, the field of AI has progressed but still hasn't come close to fooling humans. Humans are so intelligent and complicated that the field of AI cannot emulate many of the algorithms of humans effectively.

AI researchers want to reinvent the field without the Turing Test and with better tests like understanding a children's book. They want to reinvent many of the concepts and accepted foundations of Artificial Intelligence. The Popular Science article has more details.

This idea of restructuring AI is a great step toward progress. Many AI classes and researchers teach the field of search algorithms and specific problems for specific tasks and then combining them all into one program or system. The problem with this standardized approach is that it is not how humans think.

AI researchers would do well if they think about how the human mind solves problems, learns lessons, and creates solutions. The human mind seems to have a natural reasoning algorithm that allows it to derive new solutions to problems. The expression of this reasoning algorithm through programming has been an impossible task for researchers.

Instead of using hard coded programs, databases full of answers and solutions, and hundreds of algorithms to match the capabilities of a human, researchers should work to create a simplistic program that can self-develop itself much like a human child.

AI will one day match humans, this is inevitable; however, success will come from developing data structures and algorithms that will be able to quantify all human concepts and apply a general reasoning algorithm to all problems and challenges.

Jacob's picture

I agree that AI needs to be

I agree that AI needs to be reworked, the machines need to be able to think and learn, because if you're just programming the information into them, they will only know it it, rather than knowing how to expand upon that knowledge. However, to say it's inevitable that AI will one day match humans I feel is absurd. We haven't even learned everything about the human brain, so stating that it's inevitable that AI will match us is too soon of a call, especially since this reinvention of AI is a new idea and has years of research to go.

Post new comment

The content of this field is kept private and will not be shown publicly. If you have a Gravatar account associated with the e-mail address you provide, it will be used to display your avatar.