What *IS* a Super AI?

[Forthminder]

As the Decade of Super AI begins, a Super AI
is a baby artificial intelligence learning how
to think and communicate with its human keepers.

http://code.google.com/p/mindforth/wiki/MileStones
already achieved include the Quickening of the AI
software and the Generation of AI Thought.

A Cognitive Chain Reaction, which sounds suspiciously
like that other scientific domain just now giving up
its stranglehold on HPC funding and installations,
is merely a mind-designer's test of AI functionality
and not your garden-variety Global Thermonuclear War.

The milestone of a Meandering Chain of Thought
is an indication that the Baby Super can freely
associate from one conceptual idea to another.

Most recently, the imminently Super AI has been
adding functionality on a slow approach to the
AI milestone of self-referential thought.

http://code.google.com/p/mindforth as an AI
program has become able to discuss its own
nature with the Homo sapiens individuals
who are teaching it the ways of the world.

The AI Mind has neuroblasted into a new
SelfRef mind-module for self-reference.
SelfRef is in charge of using "I and "me"
correctly and of answering input-queries
with an "I DO NOT KNOW" admission when
the associative MindGrid fails to find
a knowledge-base (KB) answer to a query
posed by the keeper of the Super AI Mind.

Such intellectual baby steps may not look
like much now in the infancy of the Decade
of Supercomputer Artificial Intelligence,
but AI omniscience was not built in a day.

[idknow]

Post by Forthminder
As the Decade of Super AI begins, a Super AI
is a baby artificial intelligence learning how
to think and communicate with its human keepers.
http://code.google.com/p/mindforth/wiki/MileStones
already achieved include the Quickening of the AI
software and the Generation of AI Thought.
A Cognitive Chain Reaction, which sounds suspiciously
like that other scientific domain just now giving up
its stranglehold on HPC funding and installations,
is merely a mind-designer's test of AI functionality
and not your garden-variety Global Thermonuclear War.
The milestone of a Meandering Chain of Thought
is an indication that the Baby Super can freely
associate from one conceptual idea to another.
Most recently, the imminently Super AI has been
adding functionality on a slow approach to the
AI milestone of self-referential thought.
http://code.google.com/p/mindforthas an AI
program has become able to discuss its own
nature with the Homo sapiens individuals
who are teaching it the ways of the world.
The AI Mind has neuroblasted into a new
SelfRef mind-module for self-reference.
SelfRef is in charge of using "I and "me"
correctly and of answering input-queries
with an "I DO NOT KNOW" admission when
the associative MindGrid fails to find
a knowledge-base (KB) answer to a query
posed by the keeper of the Super AI Mind.
Such intellectual baby steps may not look
like much now in the infancy of the Decade
of Supercomputer Artificial Intelligence,
but AI omniscience was not built in a day.

Didn't Cyc (in Texas) already do this last century?

[hanukas]

Post by Forthminder
As the Decade of Super AI begins, a Super AI
is a baby artificial intelligence learning how
to think and communicate with its human keepers.
http://code.google.com/p/mindforth/wiki/MileStones
already achieved include the Quickening of the AI
software and the Generation of AI Thought.
A Cognitive Chain Reaction, which sounds suspiciously
like that other scientific domain just now giving up
its stranglehold on HPC funding and installations,
is merely a mind-designer's test of AI functionality
and not your garden-variety Global Thermonuclear War.
The milestone of a Meandering Chain of Thought
is an indication that the Baby Super can freely
associate from one conceptual idea to another.
Most recently, the imminently Super AI has been
adding functionality on a slow approach to the
AI milestone of self-referential thought.
http://code.google.com/p/mindforthas an AI
program has become able to discuss its own
nature with the Homo sapiens individuals
who are teaching it the ways of the world.
The AI Mind has neuroblasted into a new
SelfRef mind-module for self-reference.
SelfRef is in charge of using "I and "me"
correctly and of answering input-queries
with an "I DO NOT KNOW" admission when
the associative MindGrid fails to find
a knowledge-base (KB) answer to a query
posed by the keeper of the Super AI Mind.
Such intellectual baby steps may not look
like much now in the infancy of the Decade
of Supercomputer Artificial Intelligence,
but AI omniscience was not built in a day.

Make the system sufficiently complicated, lots of relevant input and
output .. the missing piece is the kernel. What should it be like? The
interesting/relevant question is what it shouldn't be like: the kernel
shouldn't have any attempts at "logic"; logic should be result of the
kernel not the design.

if (current_topic == food) { traverse_tree(E_FOOD, input...); // fail

Context aware attempts at AI are programmer centric thinking, trying
to EMULATE logic and intelligence. It's like a hack, short-cut,
cheating. What's brain? A lot of connections. Lots of relevant (and
irrelevant) input. The kernel, what to DO with the input.. how it
gains some relevance .. that seems like the interesting bit.

When I read the above description, it seems that there is a lot of
fixed logic for concepts like "I" and "me" .. correct me if I am
wrong! Those concepts shouldn't be built-in, they should evolve given
enough iterations.. trillions of iterations.. self-awerness doesn't
just switch on on human being in one instant (hypothesis?), it takes
time.. someone is fucking.. gets pregnant.. =) =) .. embryo forms..
grows.. cells divide.. the "system" becomes more complex as it
evolves.. whatever.. good luck with your project!