4 ms·

Intelligent, yes. Knowledgeable, no. Consider programming. Compare programming in the 1970s and 1980s to now. We're way more sophisticated now than we were the

by bena 5y ago

Intelligent, yes. Knowledgeable, no. Consider programming. Compare programming in the 1970s and 1980s to now. We're way more sophisticated now than we were then. Knowledge builds on knowledge and is logarithmic in growth. When I was born, we had 2400 baud modems. When I was graduating high school, getting a 56k baud modem was hot shit. Now, if I can't stream high definition video in real time on my phone, I think the connection is shitty. We had to invent agriculture. We had to invent bread. We had to invent construction. The wheel. Pulleys. Ladders. Boats. Vacuum tubes. Transistors. Netflix. Etc. But we can't invent Netflix before we invent transistors. We can't even think of certain things before certain things are invented. But as more things are invented and recorded, the less leg work we have to do. I don't need to invent bread. Bread is solved six ways from Sunday. We got all sorts of bread. We're so beyond bread, we've invented and solved ways to cook bread after it's been baked. Pre-historic man can't even imagine bread pudding because he doesn't know bread exists. Progress, at the very least, has a logarithmic component. It could be an arctangent. Where there's a long build up, a rapid rise, then a long tail.

WalterBright 5y ago
Progress happens because the intermediate inventions are useful. Nobody jumps from wooden wheels to steel ball bearings. But there are some exceptions. The invention of the airplane had intermediate steps, but they were fairly useless on their own. You can see this in the series of prototypes the Wrights made, and then combined them to build a useful airplane.
xscott 5y ago
> Compare programming in the 1970s and 1980s to now. We're way more sophisticated now than we were then. I think this is revisionist history at best. Your comments about baud rates and CPUs getting faster is true of course, but look at something like programming languages: I'd put ML from 1973 (later Standard ML, 1983) against Go or Rust, and I'd put something like Icon from 1977 against JavaScript or Python. If anything, I think we've regressed because ML and Icon could actually run on 70s era hardware, but Rust can barely compile its own compiler on 2020s era hardware, and Python wastes CPU and memory as though it was a design goal. Seriously, look at Standard ML, Icon, or even Scheme and SmallTalk from the 70s, and ask how Python and Go improved on those. ----- https://en.wikipedia.org/wiki/ML_(programming_language) https://en.wikipedia.org/wiki/ML_(programming_language) https://en.wikipedia.org/wiki/Standard_ML https://en.wikipedia.org/wiki/Standard_ML https://en.wikipedia.org/wiki/Icon_(programming_language) https://en.wikipedia.org/wiki/Icon_(programming_language)
- kristianp 5y ago
  Wow, I didn't realise ML was developed so early. Is there a history of the early work in the 70s? What machines did it run on?
  - xscott 5y ago
    Other than the standard search engines or Wikipedia, I don't know where to find a good history of language or hardware. I'm only aware of these examples because they were part of my "comparative programming languages" course in the early 90s. I suspect this thread will fade without much more notice, but I'm half expecting someone to come along and set me straight that all this stuff was really invented in the 50s or 60s :-) (SNOBOL, Algol, Simula, or PL/I ???)
- bena 5y ago
  I didn't mention languages specifically because they're really only a small part of programming knowledge. A Turing-complete language is just that, Turing-complete. Sorting used to be hard. It used to be novel. It used to be something you could study. Now it's something you're expected to just know. B-Trees, A* pathfinding, mapreduce, etc. were all, at one time, cutting edge computer science. Now they're the basis of it. Knowledge breeds knowledge. And every piece of knowledge has the potential to open entirely new pathways of knowledge. Apple Pay depends on NFC depends on RFID depends on radio technology, etc.
  - xscott 5y ago
    > A Turing-complete language is just that, Turing-complete. It sounds like you're dismissing Turing complete languages as basically the same. I think that's kind of like saying all hardware and networking is just made out of electric circuits. It's true, but it's not illuminating. > Sorting used to be hard. It used to be novel. It used to be something you could study Quicksort was created in 1959 [0], Heapsort in 1964 [1]. Mergesort is from 1945 [2]! They were something undergrads were just "expected to know" shortly after that. There have been incremental improvements since then, but these things haven't been novel in a very long time. I'm sure you won't change your mind, but this doesn't really support your argument above, "Compare programming in the 1970s and 1980s to now. We're way more sophisticated now than we were then." > B-Trees, A* pathfinding, mapreduce, Again, 60s and 70s ideas applied on modern hardware. I'll spare you the Wikipedia links for these. [0] https://en.wikipedia.org/wiki/Quicksort https://en.wikipedia.org/wiki/Quicksort [1] https://en.wikipedia.org/wiki/Heapsort https://en.wikipedia.org/wiki/Heapsort [2] https://en.wikipedia.org/wiki/Merge_sort https://en.wikipedia.org/wiki/Merge_sort
danans 5y ago
> Progress, at the very least, has a logarithmic component. It could be an arctangent. Where there's a long build up, a rapid rise, then a long tail. Or as someone here once said, in the real world all hockey stick growth is eventually sigmoid.

by bena 5y ago

WalterBright 5y ago

Progress happens because the intermediate inventions are useful. Nobody jumps from wooden wheels to steel ball bearings. But there are some exceptions. The invention of the airplane had intermediate steps, but they were fairly useless on their own. You can see this in the series of prototypes the Wrights made, and then combined them to build a useful airplane.

xscott 5y ago

> Compare programming in the 1970s and 1980s to now. We're way more sophisticated now than we were then. I think this is revisionist history at best. Your comments about baud rates and CPUs getting faster is true of course, but look at something like programming languages: I'd put ML from 1973 (later Standard ML, 1983) against Go or Rust, and I'd put something like Icon from 1977 against JavaScript or Python. If anything, I think we've regressed because ML and Icon could actually run on 70s era hardware, but Rust can barely compile its own compiler on 2020s era hardware, and Python wastes CPU and memory as though it was a design goal. Seriously, look at Standard ML, Icon, or even Scheme and SmallTalk from the 70s, and ask how Python and Go improved on those. ----- https://en.wikipedia.org/wiki/ML_(programming_language) https://en.wikipedia.org/wiki/ML_(programming_language) https://en.wikipedia.org/wiki/Standard_ML https://en.wikipedia.org/wiki/Standard_ML https://en.wikipedia.org/wiki/Icon_(programming_language) https://en.wikipedia.org/wiki/Icon_(programming_language)

kristianp 5y ago
Wow, I didn't realise ML was developed so early. Is there a history of the early work in the 70s? What machines did it run on?
- xscott 5y ago
  Other than the standard search engines or Wikipedia, I don't know where to find a good history of language or hardware. I'm only aware of these examples because they were part of my "comparative programming languages" course in the early 90s. I suspect this thread will fade without much more notice, but I'm half expecting someone to come along and set me straight that all this stuff was really invented in the 50s or 60s :-) (SNOBOL, Algol, Simula, or PL/I ???)
bena 5y ago
I didn't mention languages specifically because they're really only a small part of programming knowledge. A Turing-complete language is just that, Turing-complete. Sorting used to be hard. It used to be novel. It used to be something you could study. Now it's something you're expected to just know. B-Trees, A* pathfinding, mapreduce, etc. were all, at one time, cutting edge computer science. Now they're the basis of it. Knowledge breeds knowledge. And every piece of knowledge has the potential to open entirely new pathways of knowledge. Apple Pay depends on NFC depends on RFID depends on radio technology, etc.
- xscott 5y ago
  > A Turing-complete language is just that, Turing-complete. It sounds like you're dismissing Turing complete languages as basically the same. I think that's kind of like saying all hardware and networking is just made out of electric circuits. It's true, but it's not illuminating. > Sorting used to be hard. It used to be novel. It used to be something you could study Quicksort was created in 1959 [0], Heapsort in 1964 [1]. Mergesort is from 1945 [2]! They were something undergrads were just "expected to know" shortly after that. There have been incremental improvements since then, but these things haven't been novel in a very long time. I'm sure you won't change your mind, but this doesn't really support your argument above, "Compare programming in the 1970s and 1980s to now. We're way more sophisticated now than we were then." > B-Trees, A* pathfinding, mapreduce, Again, 60s and 70s ideas applied on modern hardware. I'll spare you the Wikipedia links for these. [0] https://en.wikipedia.org/wiki/Quicksort https://en.wikipedia.org/wiki/Quicksort [1] https://en.wikipedia.org/wiki/Heapsort https://en.wikipedia.org/wiki/Heapsort [2] https://en.wikipedia.org/wiki/Merge_sort https://en.wikipedia.org/wiki/Merge_sort

danans 5y ago

> Progress, at the very least, has a logarithmic component. It could be an arctangent. Where there's a long build up, a rapid rise, then a long tail. Or as someone here once said, in the real world all hockey stick growth is eventually sigmoid.