Template:M intro technology rumours of our demise: Difference between revisions

Jump to navigation Jump to search

Template:M intro technology rumours of our demise (view source)

Revision as of 19:05, 20 November 2023

1,130 bytes added ,  20 November 2023
no edit summary
No edit summary
Tags: Mobile edit Mobile web edit
No edit summary
Tags: Mobile edit Mobile web edit
Line 186: Line 186:
Now if artificial intelligence is so spectacular, shouldn’t we be a bit more ambitious about what ''it'' could do for ''us''? Isn’t “giving you the bare minimum you’ll take to keep stringing you along” a bit ''underwhelming''?  
Now if artificial intelligence is so spectacular, shouldn’t we be a bit more ambitious about what ''it'' could do for ''us''? Isn’t “giving you the bare minimum you’ll take to keep stringing you along” a bit ''underwhelming''?  


===Digression: the profligacy of Darwin’s Dangerous Idea===
===Digression: Darwin’s profligate idea===
By now most accept the magic of evolution by natural selection. This really is magic: a comprehensive explanation of the origin of reflexive general intelligence — indeed, the sum of all organic “creation”, as it were — can [[reduction|reduce]] to the operation of a simple algorithm that can be stated in a short plain English sentence.
By now most accept the magic of evolution by natural selection. This really ''is'' magic: a comprehensive explanation of the origin of reflexive general intelligence — indeed, the sum of all organic “creation”, as it were — can [[reduction|reduce]] to the operation of a simple mechanical process that can be stated in a short plain English sentence. The economy of ''design'' is this process is staggering. The economy of ''effort'' in its ''execution'' is not.


There is a real cost to that magic, though: cost. Evolution by natural selection is incomprehensibly inefficient. The chain of adaptations that led from amino acid to Lennon and McCartney may have billions of links in it, but the number of adaptations that didn’t — that arced off into one of design space ’s gazillion dead ends and forlornly fizzled out — is orders and orders of magnitude greater. Evolution isn’t directed — that is it's very super power, so it fumbles blindly around, fizzing and sparking, and a vanishingly small proportion of mutations go anywhere. Evolution is a random, [[stochastic]] process.  
Evolution is ''tremendously'' wasteful. The chain of adaptations that lead from amino acid to Lennon and McCartney may have billions of links in it, but the number of adaptations that ''didn’t'' get anywhere — that arced off into one of design space ’s gazillion dead ends and forlornly fizzled out — is orders and orders of magnitude greater. For every serendipitous mutation there are millions and millions of duds.  


Even though it came about through that exact process, mammalian intelligence — call it “natural general intelligence” (NGI) — isn’t like that. It is directed. Because we can hypothesise, we can rule out experiments which plainly won’t work. This is human’s great superpower: it took 5 billion years to get from amino acid to the wheel, but 5000 years to get from the wheel to the Nvidia graphics chip.
Evolution isn’t directed — that is its very super-power — so it fumbles blindly around, fizzing and sparking, and a vanishingly small proportion of mutations do anything useful. Those that do are an accident. Evolution is a random, [[stochastic]] process. It depends on aeons of time and takes colossal resources.


Large learning models are undirected. They work by a stochastic algorithm not dissimilar to evolution by natural. They get better by running that algorithm faster.  
Even though it came about through that exact process, mammalian intelligence — call it “natural general intelligence” (NGI) — isn’t like that. It ''is'' directed. Because animals can narratise and hypothesise, they can remember , they can learn, and they can rule out bad ideas without trying them. They don't have to go through the motions of trying out things that plainly won’t work.
 
All mammals can do this to a degree; humans happen to be particularly good at it.
 
This is our great superpower: it took 5 billion years to get from amino acid to the wheel, but 5000 years to get from the wheel to the Nvidia graphics chip.
 
Large learning models are, like evolution, undirected. They are colossally wasteful of processing power. They work by a stochastic algorithm not dissimilar to evolution by natural selection. They get better by running that algorithm faster, in parallel, with lightning-fast multi-core graphics processors. But they takes colossal amounts of processing power, and we are bumping up against processing limits and energy consumption costs. This is starting to get expensive and hard. And there is an opportunity cost to devoting all our resources for something that, at the moment, creates sophomore mashups we don't actually need.
 
Why burn all this evergy creating premuim mediocre content when we have eight billion idle organic CPUs who can generate first class content? When there is literally millennia of content sitting there unattended?


====LibraryThing as a model====
====LibraryThing as a model====
A better use for this technology — if it is as good as claimed<ref>We would do well to remember Arthur C. Clarke’s law here. The parallel processing power an LLM requires is lready massive. It may be that the cost of expanding it in the way envisioned would be unfeasibly huge — in which case the original “business case” for [[technological redundancy]] falls away. See also the [[simulation hypothesis]]: it may be that the most efficient way of simulating the universe with sufficient granularity to support the simulation hypothesis is ''to actually build and run a universe'' in which case, the hypothesis fails.</ref> — would create [[system effect]]s to ''extend'' the long tail.  
A better use for this technology — if it is as good as claimed<ref>We would do well to remember Arthur C. Clarke’s law here. The parallel processing power an LLM requires is already massive. It may be that the cost of expanding it in the way envisioned would be unfeasibly huge — in which case the original “business case” for [[technological redundancy]] falls away. See also the [[simulation hypothesis]]: it may be that the most efficient way of simulating the universe with sufficient granularity to support the simulation hypothesis is ''to actually build and run a universe'' in which case, the hypothesis fails.</ref> — would create [[system effect]]s to ''extend'' the long tail.  


It isn’t hard to imagine how this might work. A rudimentary version exists in {{Plainlink|https://www.librarything.com/|LibraryThing}}’s recommendation engine. It isn’t even wildly clever '''—''' {{Plainlink|https://www.librarything.com/|LibraryThing}} has been around for nearly twenty years and doesn’t, as far as I know, even use AI: each user lists, by ASIN, the books in her personal library. She can rate them, review them, and the LibraryThing algorithm will compare each users’s virtual “library” with all the other user libraries on the site and list the users with the most similar library. The non-matched books from libraries of similar users are often a revelation.
It isn’t hard to imagine how this might work. A rudimentary version exists in {{Plainlink|https://www.librarything.com/|LibraryThing}}’s recommendation engine. It isn’t even wildly clever '''—''' {{Plainlink|https://www.librarything.com/|LibraryThing}} has been around for nearly twenty years and doesn’t, as far as I know, even use AI: each user lists, by ASIN, the books in her personal library. She can rate them, review them, and the LibraryThing algorithm will compare each users’s virtual “library” with all the other user libraries on the site and list the users with the most similar library. The non-matched books from libraries of similar users are often a revelation.
Retrieved from "https://jollycontrarian.com/index.php?title=Special:MobileDiff/80523"

Navigation menu