W Ross Ashby on `Design for an Intelligence Amplifier'

  W Ross Ashby on `Design for an Intelligence Amplifier'

So, there's this book:

Shannon, C., & McCarthy, J. (Eds.). ([1956] 1972). Automata Studies (5th ed.). Princeton, NJ: Princeton University Press.

Which addresses such questions (analytic & synthetic problems/tasks) as: `How does the brain function? Can we design a machine which will simulate a brain?' And explores Turing machines, and whatnot. 

Turns out, brains are information processing devices. (No real surprises there.) 

Anyway W Ross Ashby has a chapter in it - `Design for an Intelligence Amplifier'. 

The question he addresses (this is in 1956, btw) is: 

Can humans build a machine that can solve problems beyond human-level intellect, or capacities for problem solving? 

(Velikovsky - I certainly hope so, as we humans are not very good at solving global problems, and we now have lots of them...? Climate change, nuclear threats, and biotech-&-AI, for 3 things, as Yuval Noah Harari often notes... Basically, at this point in the game, a superintelligence is "our only hope". Like, Luke Skywalker, for Princess Leia & the Rebellion, or something.)

Ashby (in 1956!) notes: science, maths, society, and the economy all have complex, urgent, and thorny problems. If we can't solve 'em, we're boned. 

He says:

"We have built a civilization beyond our understanding and we are finding it is getting out of hand." (Ashby, 1972, p. 215)

He also notes: Napoleon and Archimedes were geniuses, and, frankly, we need a few more guys like them around, these days. Genii.* 

(*That's my plural of "geniuses". Not sure anyone else ever uses that word, but, there it is. I often don't care how other people do: stuff, I like to do it, using the Frank Sinatra Method. i.e., I did it, My Way.)

Anyway, Ashby makes the great point that, when you take a look at the distribution of human IQ, it's a bit horrifying and scary. There's a whopping vacuum, or some kind of apparent upper limit. 

Namely, where the hell are the 200+ IQ guys? (Or, gals, or, They/Thems...?) 

 Something's gotta be done. We're all way too dumb.

We're now talking about Bounded Rationality, which guys like Herbert Simon and Gazzaniga talk about. Ashby says our muscles and our brains have upper limits. What's more he's right. 

Indeed, I say. We have forklifts and all that guff to throw heavy stuff around: why not brain-machines?

I am becoming convinced of Ashby's point, as are you, especially since I always thought this way in the first place, but if I quote some other guy who said it first, maybe you'll listen and be less skeptical. That's how academia works, apparently. (People also often get offended by my Brutal Truthism School of Philosophy that I just made up now, but it's not my fault; I'm an Aussie, and we often have good Bullshit Detectors, and tend not to pull any intellectual punches.)

Ashby notes, The Ancient Romans solved their physical problems back in the day by having: slaves. 

These days you can't get away with that, so we better get some machine slaves; robots or something.

He also notes, Watt and the steam engine (better than using horses as slaves), and I see, James Lovelock in Novacene had a bit to say about that; mainly that's when the Novacene Era started (with Watt's steam engine.) And, that's also why we've got a lot of problems now - like global warming (I blame: all the steam. Also all the CO2, but whatever.)  

So Ashby says: machines (like, steam engines, and whatnot) give us "power-amplifiers", so, how about some intelligence amplifiers? We could have a thing with an IQ of: a million. 

Basically, he says forget about intelligence for a second, the crux is: "We have problems and we want answers." (p 216)

(VoN - I like how this guy thinks!)

He makes the point: a long enough random sequence of symbols will contain all the answers to anything. 

(At this point I am reminded of the Borges story about The Library of Babel, but hey that's PhilosoPhiction for you. ~VoN)

(When I write `VoN' I mean Velikovsky of Newcastle. Spose that's obvious? Now it is, anyway. Never assume, it makes an ASS out of U and ME)

Then Ashby gets into the idea that if a random kid scribbles/doodles random symbols, he'll sometimes come up with a meaningful set of symbols or formula. (Like "cos2x + sin2x=2"). But mostly nonsense. 

But - after he learns some maths, he'll get more selective about what he scribbles. (Ironically, less productive.)

Here we go! We're getting into what I like: Selection, Variation, and Transmission. (Before the Selection, there was other stuff Transmitted, so we take some of that stuff and select, and maybe even combine them [or their parts] together, etc.) This is the evolutionary algorithm. Creativity and whatnot. But I digress. It's getting exciting now, as the great philosopher, `Fight Club, The Movie', once said.

Ashby notes: genius is all about selection. 

Select the right answer, or solution or whatever, from all the possible ones, most of which are incorrect. 

Ashby asks, in the same way a guy stoking a coal fire that powers a factory amplifies energy, can we do that with selection? 

Ashby says, a sieve selects (filters) stones from garden soil. (Good point.)   

But - if you choose a sieve of a different size mesh, you're selecting the selector. (And will get different sized rocks.)

Or - when a boss hires a Manager of Personnel, the boss is selecting a selector (who: selects and hires people).

Anyway then Ashby gets into systems, and has 2 self-coupled dynamic systems X and S, connect by channels G and U. 

I just like self-coupled systems. (And so should you. So should you. You're in one right now, by reading this. And especially if you give me feedback on this article, by writing a Comment below. Or not.)
Then Ashby has a homeostat of the system, (of 2 self-coupled systems) and with 4 needles N (needles, in: gauges/meters).


Then Ashby gets into increasing "entropy", but not the heat-engine kind; the stochastic process kind.
I blame Shannon, for listening to Von Neumann, who in 1940 suggested Shannon should call his discovery "entropy" (see the great biography of Shannon, A Mind At Play, Soni & Goodman 2019, 45% in Kindle)

Anyway I digress. 
Ashby then makes an AXIOMATIC STATEMENT, which are always fun. I love them. He says, he now better use Set Theory, as he's going to describe some systems of extreme generality. 
I love systems of extreme generality... Probably because it reminds me of my HOLON/parton model of the Meme, the unit of culture . (As: It's also a model of extreme generality... As is, the Darwinian model of evolution.  But anyway - that's just me? That's how I roll.)
Back to Ashby. Check this move out: (Hot-diggity diagrams, Batman!)

omg, what's not to love? Ultrastability. 
Then he hits us with a THEOREM, and then a PROOF, which all gets into Pure Math, so I won't offend you with it here. I don't want to get cancelled, for outraging too many people. 
Then Ashby shows how science works, and I love it so much. You make mathematical models and solve the problem, then apply it in the real world. It's all there on p 226, and I commend it to you. It's how they found Uranus. 
Basically, you gotta reduce the search time. Think of chess problems. Also, get a computer to do it. 
Bear in mind, he's also solving an economic problem, as a toy problem example. So it has real world apps. 
He also notes chess problems are discontinuous, but many real world problems aren't. 
Then, he talks about selection in evolution (bottom of p. 228), and frankly I love it. 
He notes: Very few problems are wholly new, you've gotta use the past knowledge. (Keep the PoMos away from Science, they are trying to ruin it. Stoopid Science-Deniers.)
Then he has this great bit, about finding solutions in a problem-space. Say there's 144 states, and only 40 are solutions to your problem:  
Anyway he winds up saying: Yeah, we should be able to amplify our intellect, with a machine. 
Great chapter, check it out. 
It was 1972, but in 2020 the problems are still hanging around like a bad smell. Let's get right on it.
See also, Nick Bostrom's stuff, like Superintelligence (2015).
And Tegmark's stuff, like Life 3.0 .

Anyway I was re-reading Simon's Sciences of the Artificial (1996) and noticed a ref to this 1972 chapter by Ashby, so I read it. And, wrote this. 
Ashby was a funny guy, check out his quotes (and specifically, his Aphorisms)... 
eg "A Cyberneticist observes what might have happened but did not."
and
" Man pays for his knowledge with humiliation."
Love his work on Systems Theory... and AI, and all that stuff.  
Anyway; there you go.
-------------//---------------


You have been reading / viewing a blog-post by:

Dr J T Velikovsky Ph.D
(aka: Velikovsky of Newcastle)
AI Researcher & Enthusiast & Evolutionary Culturologist & Filmmaker & Writer & Artist & Actor & Muso & Random Guy
(and, also The  StoryAlity  Guy) 
aka Humanimal 

More stuff:

Transmedia Blog: On Writering
IMDb (Movies, Videogames): 
Music: Texas Radio & Zen Stupidity
Youtube channel: https://www.youtube.com/joeteevee (over 100 videos, some are even: good)

Academia page: https://newcastle-au.academia.edu/JTVelikovsky

Researchgate page: https://www.researchgate.net/profile/Jt_Velikovsky

My ouvre...etc etc.


For more, see On Writering and StoryAlity News

-------------//--------------