The statement that unsupervised GAN's are the way of the real brain is an absurdity. Human beings do not yet know how the brain works. I do believe that the mind must have some way to project objects (like images of objects) onto other objects (like other images of objects or within an object space), but I cannot say absolutely that that is how the brain works. It is part of my theory or belief about how the mind works. This idea that I call 'projection' is like generation, so I totally agree that the use of generation and discrimination must be important methods for us to use (or consider using) in AI programs. I can imagine a CNNets that are able to recognize cats, boxes, jumping, and what out is might be able to generate combinations of such things in order to guess what a 'cat jumping out of a box' might look like. That makes a great deal of sense to me. But notice how I used combinations of discrete objects in order to imagine how a computer program might generate such a thing. The demand that we only use CNNs and GANs in order to fill the theory in seems pretty weak to me. The fact is that Deep Learning combines discrete steps with CNNs and everything that we are considering uses discrete methods in detailing what is to be learned and the steps that need to be taken to modify the nets. Therefore a more sophisticated theory of AI is really pointing at hybrids (of some sort). From there it makes perfect sense to ask if there is anything that might be done to make the seams a little more seamless and the gaping gaps a lot less gaping.
On Tue, Jun 11, 2019 at 12:03 PM <keghnf...@gmail.com> wrote: > Generative Neural Networks, GAN. > This give give a relation from stating image or data to another. > > Latent Space Human Face Synthesis | Two Minute Papers #191: > https://www.youtube.com/watch?v=aR6M0MQBo2w > > > A programmer select two images or data points. > A programmer put in 50 percent value into a GAM and train it to be 50 > percent transformation > between to faces. This 50 percent value is called a "latent value" > > Latent value can used for mapping distance in weight space. > > https://towardsdatascience.com/graduating-in-gans-going-from-understanding-generative-adversarial-networks-to-running-your-own-39804c283399 > > > The latent value can be used to make movement vectors through weight > space: > https://poloclub.github.io/ganlab/ > > Unsupervised GAN's are the way of the brain, artificial or real: > > https://www.academia.edu/37275998/A_Nice_Artificial_General_Intelligence_How_To_Make_A_Nice_Artificial_General_Intelligence > > > > *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + delivery > options <https://agi.topicbox.com/groups/agi/subscription> Permalink > <https://agi.topicbox.com/groups/agi/T395236743964cb4b-Mbe82375be668c59a8d768a16> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T395236743964cb4b-M3d4aeabeaede3bdbe71d4aa2 Delivery options: https://agi.topicbox.com/groups/agi/subscription
