immortal: i am not claiming that wikify is AI, i do not think anyone is claiming that? are you saying that people should only start threads about finished AI products here? also, we do not know how it works since stefan is keeping it secret, so assuming it is brute-forcing something is silly, imo. are you simply trying to ask if wikify's searching algorithm could be used as a component for an AGI? if so, that was not clear at all.
i am also not sure why you are disagreeing with me about the question of designing a platform to host an AGI, you said it yourself, one problem is simply the amount of data required. at a certain point, scaling vertically (adding more cpu cycles/ram to use) has diminishing returns in terms of cost and even hits physical limits at times, such as how most x86_64 processors are sold in the 3-4Ghz range due to cost/heat considerations which forced the 2x growth "law" to halt in its tracks. at some point, you have to scale horizontally, so how do you thread your sequentially-computed AGI across a cluster of computation nodes? you appear to be saying that it is not important to consider this at all, which i suppose I disagree with. are you saying you'd prefer to find a way to simulate a human brain in order to do whatever you are considering instead of sequential computation? parallelism is cool, but again, not sure how you plan to accommodate parallel computational threads on a single server at a certain point. i would be more interested in simulating brains with qubits or something to achieve more granular states of information transmission, but I do not know too much about how the brain actually works right now and i think the most stable qubits working together to do math as a computer is like ~116, possibly it has gone up in the last 6 months or so. my personal ideal would be to just grow a brain analogue and see if i could bake the AGI into it somehow. bio-printing is coming a long way, they have printed functional uteruses in rats and things, with findings published just as of this year: https://www.nature.com/articles/s41587-020-0547-7 the technique they use is one where they print a cell scaffold and then seed it with cells from a donor to grow a replacement organ. right now, they are using cells from the eventual host to start growth, too. On Tue, Aug 18, 2020 at 4:08 PM <[email protected]> wrote: > On Tuesday, August 18, 2020, at 10:57 AM, ella wismer wrote: > > immortal: do you mean wikify or designing platforms to run said A[G]I, > unclear > > But wikify is barely AI at all - it just stores data, and it is a trie or > what exactly? Explain it all, every bit, clearly, if u can :) > > No, not designing platform to run AGI, rather AGI, - I already have a PC > to run AGI. Computation cost / storage is based on how much data it has to > work with, more data improves AGI but takes time to eat it, same for > semantics, backoff, averaging next occurrence liklihood, etc, because it > takes time to update weights. I can make AGI that has all those merging > mechanisms but has little data and is fast because the network would be > extremely small model of the data it has seen and we get it soon with few > weights to update and few update epochs. If AGI searches by brute force > alone it is super slow and runs faster on my PC than a real brain because > it is sequential. > *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/T6322565b7d29a2a0-M7fe6d9ebe75999491dcedebc> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T6322565b7d29a2a0-Me011513e9565f911784cd991 Delivery options: https://agi.topicbox.com/groups/agi/subscription
