3 ms·

Your postulates are wrong * Nobody ever had to memorize trig tables (unless you mean the values for 30, 45 and 60 degrees) * Understanding accuracy is not ass

Your postulates are wrong * Nobody ever had to memorize trig tables (unless you mean the values for 30, 45 and 60 degrees) * Understanding accuracy is not associated with just slide rules. It is ubiquitous whenever you make measurements. Moreso if you calculate with these measured values. * People have been learning algebra and calculus for hundreds of years. They still are doing so. * Symbolic math programs like maxima, maple, wolfram alpha, sage and others only solve some problems, but they have been used for 40 years or more for particular situations. Given this, forgive me if I think your conclusion is flawed. You can ask a LLM model that answer today. And you will get an answer. Understanding what that answer means and how it is a form of hooey still requires domain knowledge. And it will for a very long time.

gcanyon 4y ago
Trig tables: I remember reading that, but I can't find a cite and I'm happy to concede the point here, but having to have a book (you do acknowledge that such tables were a thing?) that you physically look things up in is hardly better. Accuracy: first, I can't believe I cross-wired accuracy and precision. Still, you seem to be responding as if I had gotten it right, so I'll respond in kind: sure, understanding precision is still an important concept, but it was much more so when working with a mechanical device that managed at most 2-4 digits of precision. I'm guessing that with 64-bit values, precision isn't a big topic of conversation outside of NASA. (That's hyperbole) You're saying that algebra and calculus are just as tedious/necessary now as they were 30 years ago? To be clear, I'm sort of fudging -- Mathematica is actually 34 years old. But my point is that tools like it have changed who needs to learn the techniques, and what learning the techniques look like. And of course it's easier to be certain that basically no one today needs to know how to use a slide rule than to say no one needs to be able to solve a quadratic -- the more recent the change, the less absolute it will be. Finally, "a very long time" seems optimistic. As an analogy, there were articles written in 2010 saying that it would be "a very long time" before computers beat anyone reasonably skilled at Go. It was less than ten years later that the very best humans couldn't beat computers. It's impossible to know where fast progress will be made vs. slow. I purposefully chose a hyperbolic example to illustrate the point, and it's entirely possible that architecture won't be a field handled well by AI/ML in ten years. But I think it's certain that over the next several years, multiple fields that humans regard as being as complex and domain-knowledge-dependent as architecture will be automated.

gcanyon 4y ago

Trig tables: I remember reading that, but I can't find a cite and I'm happy to concede the point here, but having to have a book (you do acknowledge that such tables were a thing?) that you physically look things up in is hardly better. Accuracy: first, I can't believe I cross-wired accuracy and precision. Still, you seem to be responding as if I had gotten it right, so I'll respond in kind: sure, understanding precision is still an important concept, but it was much more so when working with a mechanical device that managed at most 2-4 digits of precision. I'm guessing that with 64-bit values, precision isn't a big topic of conversation outside of NASA. (That's hyperbole) You're saying that algebra and calculus are just as tedious/necessary now as they were 30 years ago? To be clear, I'm sort of fudging -- Mathematica is actually 34 years old. But my point is that tools like it have changed who needs to learn the techniques, and what learning the techniques look like. And of course it's easier to be certain that basically no one today needs to know how to use a slide rule than to say no one needs to be able to solve a quadratic -- the more recent the change, the less absolute it will be. Finally, "a very long time" seems optimistic. As an analogy, there were articles written in 2010 saying that it would be "a very long time" before computers beat anyone reasonably skilled at Go. It was less than ten years later that the very best humans couldn't beat computers. It's impossible to know where fast progress will be made vs. slow. I purposefully chose a hyperbolic example to illustrate the point, and it's entirely possible that architecture won't be a field handled well by AI/ML in ten years. But I think it's certain that over the next several years, multiple fields that humans regard as being as complex and domain-knowledge-dependent as architecture will be automated.