3 ms·

You mean today's computers can assign prices to existing products and decide which new products to develop better than the market? The iPhone coming out of a ne

by _yosefk 10y ago

You mean today's computers can assign prices to existing products and decide which new products to develop better than the market? The iPhone coming out of a neural network which isn't close to passing even a pale shadow of the Turing test (which a lot of humans pass who certainly cannot develop new products or set prices better than the market)? I wouldn't put my money on it. (BTW Botvinnik, the first Soviet chess champion, was trying to push his (weaker than competition) chess-playing program as a basis for running the entire Soviet economy (he explained the weakness of his program for chess by it being more general than the competition; it apparently did not occur to him however that the fact that it lost to other machines and of course to people - this was 1990 - could perhaps mean it might run the economy somewhat worse than say a person. He was an engineer.)) As to electing engineers - let's put it this way: all else being equal, I'd rather be ruled by an EXPERIENCED ex-engineer (not just a guy with a degree) with a track record of achieving something, for which people who worked with them can vouch, than an ex-lawyer or an ex-journalist. But "engineer" by itself is not even a certificate of basic sanity, let alone the ability to lead someone somewhere other than the abyss, and for this HN's comment section provides ample evidence on a daily basis.

_delirium 10y ago
I'm not betting my labor vouchers either for or against it, just an interesting thought experiment. Some of the technical problems with 1960s cybernetics are solved now, especially for more defined tasks. The USSR had huge problems getting to work things that would now be called "supply-chain optimization" and similar and are now routinely done, in a pretty centrally planned style, at large vertically integrated companies with complex logistics. This kind of stuff just didn't work at the time at the scale Soviet planners hoped to use it (scales like "optimize agricultural production in the whole USSR"), and when people tried to use it it produced bad results. I do agree that how to switch production into new sectors and products and how not to have a terrible bureaucracy capture all these decisions is a different, probably more difficult question though, certainly having some logistics optimization tools that work isn't solving the full picture (it's just better than having those problems and also having logistics optimization tools that don't work).
- _yosefk 10y ago
  There's a huge difference between a vertically integrated company competing for customers and vendors and the entire economy of a country, IMO. One evidence is that large companies keep ossifying and failing to adapt all the time which means the economy needs more than one giant company to move forward. A new Google needs a way to tell the state/the giant computer/whatever, "you're wrong, search works badly and we can do better." It's very hard to do that if the giant computer is running everything and effectively decides what you do (and it cannot run the economy if it cannot decide who does what, or at least what projects are funded and which aren't.) Also the USSR didn't just have problems optimizing supply chains. It had a problem deciding what to make. You couldn't buy woolen sweaters so you bought woolen socks and made sweaters from the woolen threads making up the sock (yes, I fucking participated in this fuckery). How is a giant neural network going to figure out how many socks and sweaters people want better than demand affecting prices affecting production, even if I grant you the assumption that the problem of managing the socks supply chain is nailed today?
  - _delirium 10y ago
    What, A/B testing not good enough for you? Please rate how you liked these two socks, this will be taken into account in order to serve you better.
    - _yosefk 10y ago
      I hear you, I just think you're a lot like that Soviet cybernetician [1], meaning that you think you're very close to the technology of that ever-elusive tomorrow, but you aren't :-) Of course it's not really fair to say that since you qualified your statements, it just sounded good and I can't miss an opportunity to fire a cheap shot at the AI/big data bubble :-) [1] Nobody was called that as cybernetics was declared "a servant of Imperialism", I guess nominally because automating away jobs hurts workers or something, which is one reason the USSR never produced particularly good computing machinery. But I get whom you refer to.
      - deleted 10y ago

by _yosefk 10y ago

_delirium 10y ago

I'm not betting my labor vouchers either for or against it, just an interesting thought experiment. Some of the technical problems with 1960s cybernetics are solved now, especially for more defined tasks. The USSR had huge problems getting to work things that would now be called "supply-chain optimization" and similar and are now routinely done, in a pretty centrally planned style, at large vertically integrated companies with complex logistics. This kind of stuff just didn't work at the time at the scale Soviet planners hoped to use it (scales like "optimize agricultural production in the whole USSR"), and when people tried to use it it produced bad results. I do agree that how to switch production into new sectors and products and how not to have a terrible bureaucracy capture all these decisions is a different, probably more difficult question though, certainly having some logistics optimization tools that work isn't solving the full picture (it's just better than having those problems and also having logistics optimization tools that don't work).

_yosefk 10y ago
There's a huge difference between a vertically integrated company competing for customers and vendors and the entire economy of a country, IMO. One evidence is that large companies keep ossifying and failing to adapt all the time which means the economy needs more than one giant company to move forward. A new Google needs a way to tell the state/the giant computer/whatever, "you're wrong, search works badly and we can do better." It's very hard to do that if the giant computer is running everything and effectively decides what you do (and it cannot run the economy if it cannot decide who does what, or at least what projects are funded and which aren't.) Also the USSR didn't just have problems optimizing supply chains. It had a problem deciding what to make. You couldn't buy woolen sweaters so you bought woolen socks and made sweaters from the woolen threads making up the sock (yes, I fucking participated in this fuckery). How is a giant neural network going to figure out how many socks and sweaters people want better than demand affecting prices affecting production, even if I grant you the assumption that the problem of managing the socks supply chain is nailed today?
- _delirium 10y ago
  What, A/B testing not good enough for you? Please rate how you liked these two socks, this will be taken into account in order to serve you better.
  - _yosefk 10y ago
    I hear you, I just think you're a lot like that Soviet cybernetician [1], meaning that you think you're very close to the technology of that ever-elusive tomorrow, but you aren't :-) Of course it's not really fair to say that since you qualified your statements, it just sounded good and I can't miss an opportunity to fire a cheap shot at the AI/big data bubble :-) [1] Nobody was called that as cybernetics was declared "a servant of Imperialism", I guess nominally because automating away jobs hurts workers or something, which is one reason the USSR never produced particularly good computing machinery. But I get whom you refer to.
    - deleted 10y ago