U.S.S.R. and communism historical discussion.
[ Login ] [ Active ]

Artificial Intelligence and Socialism/Communism

Log-in to remove advertisement.
Post 04 Dec 2020, 18:16
Hey guys; I'm wondering about your opinion on the potential for artificial intelligence and communism.

I've seen a range of articles on this subject of late (some of them more crappy than others): ... f9146bc147 ... 6296f422fe ... un-society

Putin made a comment about AI taking his place as president recently, saying he hoped this would not happen, at least 'not yet', because AI has 'no heart, no soul, no feeling of compassion, no conscience.'

But I wonder, if the parameters of AI are set to strengthening the country and improving the lives of its people, whether it could say 'hey Russia WTF have you been doing these last 30 years and why did you abandon communism?' Of course, if its access to information is limited, and/or it is 'programmed by people' as Putin suggests, then it can be given alternative, more harmful goals by its masters, such as liquidating as many Russians as possible (if programmed by people like Chubais) or maximizing oligarchs' profits, etc.


In the mid-2000s, a Russian writer created "2032: Legend of a Lost Future" - a 'techno-opera' set in an alternate universe where Grigory Romanov became general secretary instead of Gorbachev in 1985, eventually approving the creation of a fully computerized planned economy. Later, AI was allowed to take over the planning process, and also making geopolitical decisions, leading humanity to the brink of nuclear apocalypse. The author's conclusion seemed to have been that AI-controlled socialism was possible, but not desirable. ... 1504397496
Post 06 Dec 2020, 12:51
I think artificial intelligence is more of an ideology (in the past known as cybernetics) than a new sort of being like in the movies Matrix or Terminator.

There are several things people mean my them, that I'll try to describe here:

1. Large-scale statistical analytical tools. This is basically something that allows people to analyze millions or billions of data points at once, where at most before they could only do hundreds by hand. This still requires a lot of manual labor, cleaning the data, usually using low-paid piecemeal work through online services (Amazon Mechanical Turk, Yandex Toloka) etc. After that, the analyst looks at formulas that fit the data. The simplest one is linear regression - as variable A on the x-axis gets bigger, variable B on the y-axis gets bigger/smaller in a linear fashion, therefore you can draw a predictive line of best fit. Of course it could be that there is the possibility that this relation is cubic or quadratic or a sine wave, and it is up to the analyst to subjectively decide which data points are irrelevant outliers and which formula fits the data best

Another thing important here is graph theory, for example if you are buying something on amazon and there is no other data on you, the suggestions will be simply a graph of the product with the closest similar ones based on meta tags (22 cm, tefal, skillet, aluminum) and the ones that rack up more points in similarity will be first.

2. Programmable logic based on this - basically taking the model the analyst creates and applying it to data in real time, hoping that this is applicable. For example if you buy upmarket items, your graph of suggested purchases will also be multiplied by your payment potential, so that you don't see cheap alternative products, while if you try to buy stuff on sale, products on sale will be given more priority. In addition, items that you have bought will be ranked down, but similar items (cookbooks, skillet lids, oven mitts) will be ranked up.

This is again not done just for you, but based on similar trends of people buying things, maybe not even skillets, maybe someone bought a rug and later bought a vacuum cleaner, and there was a similar trend in terms of distances between points on a graph that an analyst saw and applied to other points of data, then someone else corrected it for different categories of items. This is nothing new, mail order catalogues would also position items like that together, and if you go to a store, the floor salesman will help you pick an item and then suggest other items you may need. This simply automates that work.

3. Expert systems - this is an attempt to imitate a human through basically games of 20 questions, or something more complicated and theoretical. For example, much modern machine translation relies on algorithms developed in the Soviet Union in the 1950s. But generally compiled, again, by thousands of people filling out marketing/psychological surveys, as one member here did for a living. A much more famous example is computer vision - where people do nothing but look at pictures all day and tag them as what they are. Or Google Captcha where they give you a tag and you have to press objects.

4. Expert systems multiplied by statistical analytical tools, multiplied by real-time algorithms - where the knowledge that people enter is analyzed for predictive models and then used in real-time, for example for self-driving cars. A car has to know the rules of what signs mean and what car has priority at a stop sign (yes-no question expert system), and also what a stop sign is (image tagging), and that in this situation it is approaching a stop sign (expert system data corrected by statistics). This is what people consider AI, but it is a combination of different systems working together

None of these things create anything new, simply automate certain types of work, so instead of having drivers you can have many image taggers who can do work at any time anywhere in the world. This solves many problems - for example you could have trams and buses run 24 hours a day because you don't need to pay people to do it. Or instead of salesmen, you could have analysts look at patterns in what people buy and sell. Of course this does not eliminate jobs but creates new ones - instead of having a store you could have a drop-shipping business where you fill in the holes Amazon can't by starting your own store and using advertising (also run by AI) to attract customers using your own knowledge of how to sell stuff.

So in sum, it 90% depends on decisions made by humans, which can be wrong, and it helps eliminate jobs that are based simply on stuff like that which can be automated, which can hurt both educated people whose work is repetitive and procedural (lawyers) and uneducated people whose work can be done using statistical predictions (cashiers, security guards), but it's something you have to adjust to, just like industrialization eliminated the fabric weavers, but I don't think it will eliminate work.

In government I'm pretty skeptical of its use - while it's probably good for planning infrastructure and solving traffic problems (which already have advanced analytical tools), if it's used to spy on people, it's probably bad. I do doubt facial recognition though - while it's probably useful for catching criminals, it has to be very limited in its application so that it doesn't turn into a totalitarian social control. Also I had a situation where I had to go through gates with facial recognition so that you don't have to swipe your ID, and I signed up for it out of curiosity, but when I lost weight it stopped recognizing me, while my photo ID still works for buying alcohol, etc., so it's probably good that this technology is not everywhere or I would have to somehow re-register or lose access to everything.
Post 11 Dec 2020, 19:16
That's incredibly fascinating Kirov, thanks so much for the detailed response. I feel like I just had a lesson from a professor in an introductory class on the subject. Very interesting.

Kirov wrote:
For example, much modern machine translation relies on algorithms developed in the Soviet Union in the 1950s.

Yes, I had read about that. I even have this book from the 1960s about some of these algorithms, which apparently were developed with the help of some of our best literary translators (whose precision and creativity, as you probably know, was just mind-boggling at times).


'Translator, Historian, Poet?' by Dmitry Zhukov

By your definition of AI as an ideology, as cybernetics, I suppose we've been going down that road for many decades now then. I even found this other book from the 1980s about Soviet plans for self-driving cars! I opened the book and it's filled with formulas, so I understood very little. But the fact that this kind of stuff was being worked on while Mr. Mister's Take These Broken Wings was the number one smash hit on the radio is pretty awesome. Man how fascinating it would be to have a window into a parallel dimension where the USSR never collapsed/turned capitalist.


'Automatic Control Systems for Vehicular Movement', Vetlinsky V.N., Osipov A.V.

More Forums: The History Forum. The UK Politics Forum.
© 2000- Privacy.
[ Top ]