""Our ultimate goal is to build a new generation of computer systems that are substantially more robust, secure, helpful, long-lasting and adaptive to their users and tasks. These systems will need to reason, learn and respond intelligently to things they've never encountered before," said Ron Brachman, the recently installed chief of Darpa's Information Processing Technology Office, or IPTO."
An example of what IPTO/PAL might do:
"If people keep missing conferences during rush hour, PAL should learn to schedule meetings when traffic isn't as thick. If PAL's boss keeps sending angry notes to spammers, the software secretary eventually should just start flaming on its own."
and, this is supposed to be achieved through a proposed technique(s) called REAL-WORLD REASONING based on these three concepts: 1) High-performance reasoning techniques, 2) Expanding the breadth of reasoning and hybrid methods, and 3) Embedded reasoners for active knowledge bases.
Now, in any dictionary, the word 'reason' has to do with mental states, analytic thought, logical deductions and inductions, etc, all of which come around to depend on the thinking process, which is a mental state that ultimately has to do with the human mind. If we are to agree that the human mind is a manifestation of the electro-mechanical-biological human brain, than the approach of rules and logical entities interconnected amongst themselves might some day bring about a machine that might 'act' as the human mind.
However, what is most interesting, it does not seem that there have been any attempts to look at the human process of 'reasoning' and 'thinking' from an angle different than the electro-mechanical-biological viewpoint. A brief reading of the REAL-WORLD REASONING proposal does not reveal any new insights except that it proposes another way based on the information-processing understanding of the information where bits of information are manipulated using relevance judgment for ‘aboutness’ assessment. Perhaps the issue or relevance as used over the past few decades needs to be reassessed?
So, what is uniquely different with the REAL-WORLD REASONING proposal?
The reason why the efforts of AI (artificial intelligence) so far have proven unsatisfactory in emulating the human reasoning and thinking process might have to do with the very fact that so far the approaches have been only mechanistic, thus incompatible with the very nature of the human experience and with the human mind in particular. So, we want computers to think intelligently, reason, learn, and think, and yet we apply mechanistic approaches to attempt to achieve these functions which require intellect?
It would be nice to hear if anyone has come around to know of an effort, practical or theoretical, that attacks the issues of machine 'thinking' and 'reasoning' from a perspective fundamentally different than the information-as-thing understanding (i.e mechanistic). Anyone?