On Tue, Jan 5, 2021 at 12:44 AM <[email protected]> wrote:
> What if not only could a non-simulated brain have unseen function, but a > non-simulated body may also have unseen function, and a natural dataset > from the real world like we get and not human made or limited has unseen > function? > > Without doing that, we still have made impressive AI like GPT-3, IGPT, > JUKEBOX. But could we get more sauce from trying it. Of course we already > know we want to give it real data streams, and a real body, it is part of > the plan. Of course it feels expensive to do so until the end. But our AI > is stuck on a sim for the most part ya. > > My conclusion is a real brain/ body/ data stream is not worth it, as it's > expensive even IF it has some ability to more quickly shed light on how to > make real combustion in an engine happen. Our learning to walk AIs do > pretty good in sims, even if not perfectly simulating friction etc. > > If our goal was to make a computer-simulated combustion engine, yes it may > not work, but we can get it to, just it may take a tad longer since we are > creating the PHYSICS and the MACHINE, and not just the machine in a already > given PHYSICS only. *? . .* > > I really am far in AGI, I see no need to experiment so radically to see if > real axons etc do something naturally useful for pattern finding ex. retain > or pool energy in nodes to act like a short term pool, without us bringing > it to do it by code. I already have that all figured out. I do better by > asking myself how or why would i want it? For example, if i want to eat all > day, do i really want that? Some may argue, ya, it good, but then we are > all big n fat and die young. Even though people may think its obvious and > say "don't eat so much, its just wrong, i don't know why, though, just > don't", asking yourself what the actual reason is actually explains why it > is not so great. > > Now the ?, hmm, when we think about how AGI works, we have the machine & > physics both in mind, no? And if you don't have the physics in mind, you > wouldn't see it anyway. And if the brain has its own level of operation not > dependent on the atomic physics, does it matter then to think about the > physics underneath? We create simple physics sims without atoms. They are > rather pretty close in many ways, despite lacking a massive amount of the > atoms in a real scenario. So, I think, the AGI function is all that > matters, it is the physics/ machine. There may be a physics ni real life we > don't have but the brain is a rebellion to that physics, it acts > differently, it may send a signal/particle to location B only if sees X, or > may decide some nodes get thicker connections, none of this is particle > physics, it's a higher level, like that crappy 3D physics sim in Blender > that works but has 0 atoms. So I think our AGI is the machine, but the > physics is not needed, only the level of the machine physics. And I don't > think making a real brain-chip will show us much insights how to do it, > seems like a slow unintelligence-backed search. > Nature made natural general intelligence without computers, models or abstractions. Just brain physics. I guess it's up to the rest of us to try to do the same thing to make an artificial version of it. The prospect will soldier on without you, I guess. Version 3 has been approved: Hales, Colin (2020): The Model-less Neuromimetic Chip and its Normalization of Neuroscience and Artificial Intelligence. TechRxiv. Preprint. https://doi.org/10.36227/techrxiv.13298750.v3 I am done with the article. If it appears in a journal I'll let the group know. The first basic prototyping results will happen over the next couple of years. The beginnings of a test rig is on the floor next to me. I'm really looking forward to an empirical adventure! cheers Colin ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T2f2a092379e757d2-M02134ecc948723591b89340f Delivery options: https://agi.topicbox.com/groups/agi/subscription
