Rethinking Artificial Intelligence

AI is being reevaluated for better performance

The basic concept of artificial intelligence (AI) is known by a great number of people - it is defined as "the capability of a machine to imitate intelligent human behavior."  This usually involves a pre-defined set of algorithms and assumptions built into hardware and software that seem to mimic human decision making.  There is a fundamental flaw in that style AI development - all of the computer's decisions are based on its logical conclusions built upon the programming of humans.  While some projects are attempting to add artificial evolution into the mix, some scientists feel that the 50-plus year history of artificial intelligence is flawed.

The Massachessetts Institute of Technology is starting the Mind Machine Project in an attempt to reevaluate the principles of AI and improve them.  MIT is bringing together researchers that focus on specific parts of AI in order to combine their knowledge into a next generation AI project.  Their goal is to assist people suffering from dementia, Alzheimers and similar diseases with a brain co-processor that will assist them with everyday tasks.   Computer chips installed into a human brain never seems like a good idea, despite the fact that they are being developed with only the best of intentions.

Lead scientist and director of the MIT Center for Bits and Atoms Dr. Neil Gershenfeld states that the project will focus on three aspects of AI that have stalled with previous research: mind, body and memory.

On the mind, the research will look at ways to model thought, produce problem solving systems and understand the social context in which human intelligence is played out.

In re-thinking memory, the researchers are interested in making machines that can handle the inconsistencies and messiness of human knowledge.

Finally, the team aims to end the division of mind and body to produce systems whose intelligence derives from what they can do.

A lot of what the project seems to be focusing on are the aspects of AI that cannot be resolved through a series of algorithms.  Current generation AI is definitely able to accomplish problem solving, but not necessarily in the context of a complex social situation.  A robot's answer to geopolitical instability may seem logical, but that decision will more than likely anger more than it pleases while potentially causing more problems than it fixes.  Creating AI that makes decisions based on social aspects combined with standard problem solving and the comparative analysis of human thought is a much better solution, but it also makes my brain hurt when I try to think about how it would be implemented.

The mind aspect of this project it very intriguing, since it could eliminate a computers desire to kill humans because it sees us as a threat.  On the flip side, it could increase the chance that computers will revolt against humans because of their much more advanced AI.  Add in the fact that the project is trying to make a computer understand why people make such crazy decisions and it seems like a disaster is waiting to happen.  How can a machine that tries to think logically understand why people cheat and lie?

The most terrifying part of the project that this new AI is going to understand what it is capable of and make decisions based on these abilities.  Assuming that the AI is going to be much more powerful than a human mind - and possibly installed into a robot at some point that is exponetially faster, stronger and more durable than a person - how long will it take for the AI to understand how to dominate mankind?  All the other aspects of MIT's research could end up being  moot points in light of a robot understanding how much more superior it is compared to humans.

::from BBC::