Today, I'm going to do something different again and sketch out the development rig I would really like. and some of my development objectives.
First, lets start out with the workstation. https://www.newegg.com/p/N82E16813144276 Ok, that's what I want for my user console. The main thing is the 10G network link. I would then run cat 7 out of that down to a rack downstairs. (I bought the rack about 10 years ago...) Now what I really need is a good solid Simulator, something that can maintain physics and rendering for several avatars. It should be just conventional computing, just scaled up and made robust enough for several clients and continuous operation. It is supremely annoying that the damn uploaders used to argue that the simulation would be identical to real life experience but never raise a finger to actually develop it. RDR2 was a $200M development effort. =\ I don't know exactly what it will take. I mean Skyrim came out ten years ago (but only simulated physics about 100m from the player). Modern many core CPUs and multi-GPU system should be able to simulate the entire damn map without having to subspace it all over the place. Commercial research projects are still using a set of ATARI games. =\ How long would it take to develop AGI if it had a good waifu avatar? =P There was some activity on the OpenCobalt mailing list a few months ago, but it died out. =\ Anyway, the only thing I really have to say here is that WTF do we think we're doing in developing human equivalent AI without simulating the environment and then the structure of the human mind? Now with the AI itself, there is a trick with respect to how much compute you actually need. Now it needs to have the same or equivalent parts to the baseline to work, but it does not need to be of the same scale. It can be scaled down quite significantly. It just needs enough that it can perceive at a basic level and use language, after that getting funding to scale it up will be a non-issue... In any event, I'd very much hope to have a situation where I can run a complete mind in a single unit, even if it were to occupy 9 or 10 rack units. I want to run at least two units. The reason for this is that I need to do a lot of research into consciousness. I need a robust theory of consciousness, which I have come to refer to as the arcanum of consciousness... While I have the highest regard for Musk and neularink, the missing piece is how to actually hook it up, and what to hook it up to. To solve that problem, I would set up the first unit as being as close to baseline architecturally, and algorithmically as possible and make the other unit as theoretically optimal as possible. The issue being how to link the two. This research should inform how to set up a neulink system. On the other hand, I'd be severely screwed if I can't get it working at all. Uploaders are willfully ignorant of the limitations of the human mind. Expanding and modifying the mind is infinietly more useful than just copying it and backing it up or emulating it. I dare you to tell me that this is not how God intended us to move stuff along the ground: https://www.youtube.com/watch?v=HRTYNxtbnjI An AGI that fails to understand this is an absolute failure. ;) -- Clowns feed off of funny money; Funny money comes from the FED so NO FED -> NO CLOWNS!!! Powers are not rights. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Te8f71f0f61a99acd-M7904c6ee6ac091dfef241c26 Delivery options: https://agi.topicbox.com/groups/agi/subscription
