Re: [agi] Re: (video) How can I improve my AGI architecture?

Fri, 15 May 2020 21:52:27 -0700



On 5/9/20 3:41 AM, [email protected] wrote:
I'm really excited if even just one of yous can advance the AGI design I'm at. I've seen a lot of ANN variants like variants of GANs, LSTMs, Autoencoders, etc etc, they seem to have things like residual connections, layer norm, convolution windows, many feedforward networks stacked, etc, while my design just sticks to a single collection of hierarchies.


Hey Immortal Discoveries... You speak of architecture of an AGI but seem to only present one component of a whole system. You mention that this isn't the full architecture. Do you have a few more components of the architecture?

Is this text compression going to help a unit make better choices? I'm not sure I see how. 

Is the compressor just an example of a “predictive” ability that you deem important for an AGI? How does it interact with other parts of the architecture to make better decisions?

Reminds me of looking at video being processed for a self driving car application. Little boxes were placed around identified objects in the picture. I thought it was impressive that the software could detect so many items in the visual. But, the demo doesn't indicate how you go from finding objects to building a behavior for “driving” a car – knowing when to turn, brake, accellerate, navigate etc.


Object recognition is a natural requirment of an AGI, and the technology is impressive, but it seems to only solve a small part of the decision making task for “intelligence.” I get the same feeling when I consider the “ability” of text compression. It seems like a small piece of a much more complex system.

Care to elaborate on your full architecture?

Stan

Reply via email to