I suppose exploring random topics in math could theoretically lead to advances in AGI with nonzero probability. I'm betting that it won't. You will need a stronger argument to change my mind.
AGI is an engineering problem. How do you run 100-300M lines of code on a 10-20 petaflop computer with 1 petabyte of RAM for less than the global average cost of human labor, USD $5 per hour? How do you reduce the power consumption from 1 MW to 20 watts? We know it is possible because the human body performs 10^18 DNA/RNA/amino acid operations per second on 10^23 bits of DNA at a cost of 10^-16 J per operation. Moore's law predicts the 2080's at the current doubling time of 1.5 years. It won't be done with transistors because they use a million times too much power. The most practical approach for now is to optimize for a million different specialties, most of which can be done with existing hardware and software and specialized machines. I know the holy grail is a humanoid robot with an artificial human mind that can be copied billions of times. Does anyone here really think they can do that in their lifetime? On Tue, Feb 9, 2021, 11:17 AM Ben Goertzel <[email protected]> wrote: > On Tue, Feb 9, 2021 at 5:50 AM stefan.reich.maker.of.eye via AGI > <[email protected]> wrote: > > > IF you have a set of mutually associative combinational operators, can > you then isomorphically map the functions being combined into Hilbert space > vectors to that the (associative) combinations map into vector operations? > > > > And this is a building block of AI? > > I don't need such a mapping for any particular reason within my AGI > approach, I'm just exploring the idea-space that YKY suggested.... > The nature of math is one never knows where a new formalization may > lead... > > I am working on a paper now summarizing some of the math I've found > useful in the OpenCog Hyperon design process, which is different > though related -- looking at how to express our various OpenCog > cognitive algorithms in terms of Galois connections and > chronomorphisms defined over metagraphs ... it's more > discrete/algebraic/graph-theoretic math ... though of course all math > is connected.. > > ben ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T54594b98b5b98f83-Meacded34b43032295783de3a Delivery options: https://agi.topicbox.com/groups/agi/subscription
