2 ms·

No, in books and journals. We're still printing them by the billions every year. Software of various kinds might be lost, since we usually don't print them out

by kijin 7y ago

No, in books and journals. We're still printing them by the billions every year. Software of various kinds might be lost, since we usually don't print them out. But the theory required to re-implement them are all available on paper, sitting in a library somewhere. A book on algorithms and data structures will be far more useful to a future civilization than raw source code for a hardware architecture that won't even exist anymore.

YeGoblynQueenne 7y ago
I often have to read articles published in the MLJ, or the proceedings of IJCAI and AAAI in the 1990's. That was the time when machine learning was predominantly preoccupied with modelling the world with logic languages, particularly propositional logic (and, less often, first-order logic). Although I have a background in logic programming and I study logic-based machine learning for my PhD, I find that I often have trouble to understand important terminology - and I mean "important" in the sense that entire papers are based around it, but take it for granted that the reader is very familiar with the terms, what they mean, and how they're used. For example, I recently struggled with the following terms: "finite axiomatization", "explanation-based learning", "constructive induction", "knowledge-level learning", "weakest preconditions", and others. I eventually managed to get on top of those terms, sufficiently to read and understand the papers that used them. I did this by searching online for an explanation ("finite axiomatization" is used in mathematical logic) and by digging in the literature for terms that were defined in previous works. I found all of that material online, of course- and where I couldn't find a copy online, I basically had to "wing it" and hope I grokked it right. Searching for anything published even in a major AI conference or journal before the 1990's is a big pain and even university libraries seem to lack physical or electronic copies. To conclude- yes, there are going to be various works left behind if civilisation collapses. But a lot of them are going to be like maps without the key. The people surviving us will be able to read them, but not able to understand them, because they will miss key knowledge, that is currently spread around hundreds of thousands of sources that only experts in niche sub-fields are even aware of.
TheOtherHobbes 7y ago
A book on software and algorithms would be useless without hardware to experiment on. There's only so much you can do with cogs and rods, so software only really starts to make sense with electronics of some kind - and only becomes useful when you have enough speed and memory to do real work. Which in turn assumes that you have real work to do. One of the earliest civilian computing systems was a payroll calculator - which is only useful if payroll is a problem you need to have solved, which in turn is only necessary if your economy is complex enough to need payroll calculations at significant scale, and human "calculators" aren't available. The point is that the Greeks and the Romans had the IQ to invent modern technology. The Greeks especially also had the philosophical background. But neither had the politics or the economics. Modern computing is a political phenomenon, not just a scientific one. It exists because it solves certain economic and political problems, and there's no reason to assume that a different culture would have identical problems - even if it could understand the theory.
- kijin 7y ago
  All the more reason to appreciate the general knowledge we've written into books, rather than the cutting-edge implementations that only live on computer disks. The former is much more likely to be useful to a civilization with radically different needs and available tools.

by kijin 7y ago

YeGoblynQueenne 7y ago

I often have to read articles published in the MLJ, or the proceedings of IJCAI and AAAI in the 1990's. That was the time when machine learning was predominantly preoccupied with modelling the world with logic languages, particularly propositional logic (and, less often, first-order logic). Although I have a background in logic programming and I study logic-based machine learning for my PhD, I find that I often have trouble to understand important terminology - and I mean "important" in the sense that entire papers are based around it, but take it for granted that the reader is very familiar with the terms, what they mean, and how they're used. For example, I recently struggled with the following terms: "finite axiomatization", "explanation-based learning", "constructive induction", "knowledge-level learning", "weakest preconditions", and others. I eventually managed to get on top of those terms, sufficiently to read and understand the papers that used them. I did this by searching online for an explanation ("finite axiomatization" is used in mathematical logic) and by digging in the literature for terms that were defined in previous works. I found all of that material online, of course- and where I couldn't find a copy online, I basically had to "wing it" and hope I grokked it right. Searching for anything published even in a major AI conference or journal before the 1990's is a big pain and even university libraries seem to lack physical or electronic copies. To conclude- yes, there are going to be various works left behind if civilisation collapses. But a lot of them are going to be like maps without the key. The people surviving us will be able to read them, but not able to understand them, because they will miss key knowledge, that is currently spread around hundreds of thousands of sources that only experts in niche sub-fields are even aware of.

TheOtherHobbes 7y ago

A book on software and algorithms would be useless without hardware to experiment on. There's only so much you can do with cogs and rods, so software only really starts to make sense with electronics of some kind - and only becomes useful when you have enough speed and memory to do real work. Which in turn assumes that you have real work to do. One of the earliest civilian computing systems was a payroll calculator - which is only useful if payroll is a problem you need to have solved, which in turn is only necessary if your economy is complex enough to need payroll calculations at significant scale, and human "calculators" aren't available. The point is that the Greeks and the Romans had the IQ to invent modern technology. The Greeks especially also had the philosophical background. But neither had the politics or the economics. Modern computing is a political phenomenon, not just a scientific one. It exists because it solves certain economic and political problems, and there's no reason to assume that a different culture would have identical problems - even if it could understand the theory.

kijin 7y ago
All the more reason to appreciate the general knowledge we've written into books, rather than the cutting-edge implementations that only live on computer disks. The former is much more likely to be useful to a civilization with radically different needs and available tools.