Open in TheBrain

NOVAMENTE

<p class="default">10:39 PM, August 16, 2007&#160;

&#160;

Patternist Philos of Mind&#160;

&#160;

&#160;

The phenomenal self, a key aspect of intelligent systems is the result of an intelligent system&#160;

recognizing itself as a pattern in its (internal and external) behaviors.&#160;

&#160;

A baby knows nothing not even what itself is and what its mother or the bed is.&#160;

Must gradually build up repres of world.&#160;

&#160;

Intelligence is the ability to achieve complex goals in a complex environment&#160;

&#160;

Goals are achieve via recog. probabilistic patterns of the form&#160;

"Carryin out procedure P in contex C will achieve goal G"&#160;

&#160;

his book: The Hidden Pattern&#160;

A Patternist Philosophy of Mind&#160;

&#160;

Novamente's Atom Space&#160;

nodes and links: similar to&#160;

&#160;

ATTENTION VALUES low LT import High LT import&#160;

low ST import useless not used now but need it (mother phone #)&#160;

&#160;

high ST import very important&#160;

&#160;

attention,semantics including prob. truth value&#160;

&#160;

links may denote generic assoc: Hebbian link: coffee cup to coffee&#160;

&#160;

logical representaiton with truth values and associations.&#160;

&#160;

Novamente machine&#160;

&#160;

&#160;Sloman cognitive architecture&#160;

&#160;Stan Franklin&#160;

&#160;Aaron&#160;

&#160;Minsky&#160;

&#160;Patty Maes&#160;

&#160;

&#160;Percepts go to Active Memory&#160;

&#160;

&#160;Assigning a LT import value (must cycle thru and decide)&#160;

&#160;

&#160;Philos of Cognitve Process Integration&#160;

&#160;Combine various algorithms, carrying out var. cog functions such that they&#160;

AMELIORATE rather than exaggerate each others' intrinsic combinatorial explosions.&#160;

&#160;

&#160;Can you hook algorithms together so that they CALM DOWN the combin explosion.&#160;

&#160;

&#160;Probab. logic engine (best integration of prob. theory + logic&#160;

&#160;

&#160;How to QUELL combinatorics&#160;

&#160;

&#160;Moses Looks: Moshe's PhD thesis: make inferences about which prob.&#160;

&#160;Probab. evolutionary learning ;&#160;

&#160;

&#160;Control Processes&#160;

&#160;

&#160;Trick: how to get system to recognize itself in the world.&#160;

&#160;

&#160;COMBINE EXPLAIN COMBINE EXPLAIN LOOP&#160;

&#160;

&#160;WHAT COULD I BE???&#160;

&#160;

&#160;SELF REFLEXIVE&#160;

&#160;

&#160;Atom Table MindAgends&#160;

&#160;

&#160;Must WORK ON AGI to achieve AGI (to easy to cheat with single domain)&#160;

&#160;

Q: relation to Jeff Hawkins: hierarchy with Bayes Net.&#160;

A: hierach. pattern recog. great but only one tool.&#160;

&#160;

Q: What not use&#160;

A:&#160;

&#160;

Q: Each percept has a time stamp has a separate time server.&#160;

&#160;

Q: Fetch and Tag: give partial rewards;&#160;

&#160;

Bioinformatics system: great discoveries but limited domain.&#160;

&#160;

&#160;</p>

Attachments