The mathematics of training a neural network. I understand how they work once trained, but that you can train them almost seems too good to be true.
Also, for me, how all these advances in ML/computing are alleged to be on the horizon, when I hear that A/D/C(NN)s are actually so slow in learning. How can something that has to be trained with 1M of be smart? What's the next thing?
Ask HN: What scientific phenomenon do you wish someone would explain better?
631–640 of 839 posts
Re: Ask HN: What scientific phenomenon do you wish someone would explain better?
#632Electricity is always explained by its effects, but never by its actual nature. I'd like better explanation :-)
Re: Ask HN: What scientific phenomenon do you wish someone would explain better?
#633Macroeconomics. Central banks are "creating" a trillion here, a trillion there, like nobody's business. But what are the consequences? What is the thought process that central bankers have gone through to make these decisions? Also why, exactly, are they buying the exact assets that they are buying (govt. debt, high-yield bonds, etc..) and why not others (e.g. stocks or put money into startups)? And then, what happen…
PhD student here. Not expert on all those questions but: > Macroeconomics. Central banks are "creating" a trillion here, a trillion there, like nobody's business. What is the thought process that central bankers have gone through to make these decisions? The general consensus is that central banks should stay passive and keep prices stable. However, in periods of crisis, like the one we live in, the central bank coul…
>Yep, pretty much...
I would add that if the system is fractional reserve then it increases the proportion of the bank's reserve allowing more money to be created. So while it's technically true that it's destroyed you could see the next loan as its reincarnation, no..?
I didn't go here in my response above because my vague understanding is that we're not strictly a fractional reserve system any more, though I don't understand how.
Re: Ask HN: What scientific phenomenon do you wish someone would explain better?
#634Abiogenesis. I understand that this is not considered understood but I'd be interested in hearing a qualified scientist talk about how we get from "dumb" matter to self-replicating, goal-driven matter. The closest I've ever heard anyone get (in a personal conversation) is that chemistry is about transformation, so "dumb" matter isn't really dumb in the sense that it's static. Still lots of hand waving to get from bak…
Edit - I just lost 20 mins reading the start of https://en.wikipedia.org/wiki/RNA_world which is interesting on that stuff
Re: Ask HN: What scientific phenomenon do you wish someone would explain better?
#635- How computers translate bits into electrical signals - What does it mean the universe is expanding - Bayesian statistics - How information is stored in magnetic tapes
You can build logic gates out of 2 or 3 transistors, and combine those logic gates into more complex gates until you've got a computer.
But how does a transistor work? Basically, you've got semiconducting materials of two types (Phosphorus or Boron doped silicon), one of which wants a few more electrons to be a conductor, and the other wants a few less electrons to be a conductor. If you stick the two types next to each other, the electron-wanting (N) one snatches up the electrons from the electron-offering (P) one, and you get an Electric field going from the P to N. Now, that alone makes a diode. Already cool, nonlinear electronics. But what if we go N-P-N? Now we've got two electric fields, going opposite directions. With three leads, you can adjust the strength of those electric fields with one, creating a variable resistor, a transistor.
Re: Ask HN: What scientific phenomenon do you wish someone would explain better?
#636Macroeconomics. Central banks are "creating" a trillion here, a trillion there, like nobody's business. But what are the consequences? What is the thought process that central bankers have gone through to make these decisions? Also why, exactly, are they buying the exact assets that they are buying (govt. debt, high-yield bonds, etc..) and why not others (e.g. stocks or put money into startups)? And then, what happen…
I'd love to see this properly explained, because it definitely has a counter intuitive ring to me.
Re: Ask HN: What scientific phenomenon do you wish someone would explain better?
#637Macroeconomics. Central banks are "creating" a trillion here, a trillion there, like nobody's business. But what are the consequences? What is the thought process that central bankers have gone through to make these decisions? Also why, exactly, are they buying the exact assets that they are buying (govt. debt, high-yield bonds, etc..) and why not others (e.g. stocks or put money into startups)? And then, what happen…
I've wondered the same thing lately whenever someone here posits that defaults cause the destruction of money. I'd love to see this properly explained, because it definitely has a counter intuitive ring to me.
Re: Ask HN: What scientific phenomenon do you wish someone would explain better?
#638Eddy currents. Why does a magnet move slowly down a copper tube?
Re: Ask HN: What scientific phenomenon do you wish someone would explain better?
#639Why the earth is a sphere.
Re: Ask HN: What scientific phenomenon do you wish someone would explain better?
#640Earlier quoted context omitted.
I've wondered the same thing lately whenever someone here posits that defaults cause the destruction of money. I'd love to see this properly explained, because it definitely has a counter intuitive ring to me.
Whoever defaulted spent the money on something else, so it's still in circulation somewhere, just not in a form the original creditor can get hold of.
I'm pretty ignorant in this field, and usually I've been a day or so behind the posts (missing the window to press for more information), but I feel like there's definitely some contention there.