phreeza

64 ms·

811.
by phreeza 6y ago
I don't think the rise of the AfD in Germany has much to do with a reduction in ethnic or cultural 'purity'. The general political consensus position in Germany has shifted considerably to the left over the last 30 or so year
812.
by phreeza 6y ago
> Twitter's been posting a profit since 2018. 12 years after it was founded.
813.
by phreeza 6y ago
Household names for a very specific set of households :)
814.
by phreeza 6y ago
With ICA, don't you have to have n independent mixes of the n speakers in order to demix them all?
815.
by phreeza 6y ago
https://pppreport.org/company/ayn+rand+institute+center+for+...
816.
The Unreasonable Effectiveness of Recurrent Neural Networks (2015)(karpathy.github.io)
2 points by phreeza 6y ago|0 comments
817.
by phreeza 6y ago
The problem with search in particular is that it gets better as more people use it, because click behavior feeds into the search ranking algorithms. So if you force Google to expose such an API, you effectively degrade the quality of search
818.
by phreeza 6y ago
Not an expert by any means, but from my experience with physics experiments I would guess the error goes down as 1/sqrt(n). So more satellites help, but with diminishing returns.
819.
by phreeza 6y ago
I have some music that I kind of consider slightly trashy but do enjoy on occasions. I have the feeling that Spotify knows that this is a kind of "guilty pleasure", and seems to explicitly not recommend it in my autogenerated play
820.
Thermoelectric Stoves: Ditch the Solar Panels?(solar.lowtechmagazine.com)
2 points by phreeza 6y ago|0 comments
821.
by phreeza 6y ago
That is the legal situation, but the question then arises if it is desirable for a society to develop in the direction where such ostracisms happen a lot. I think there is an argument to be made that speech which is overly constrained by so
822.
by phreeza 6y ago
Maybe the author should have gone with the faces of a Disdyakis triacontahedron
823.
by phreeza 6y ago
If you are talking about the movie that I think you are talking about, I don't think it is really so much about a superior way of communicating, just a different one.
824.
by phreeza 6y ago
I think the most detectable radio signals would come from stuff like early warning radar, not communications, anyway. Not sure if those can be compressed in a meaningful way.
825.
by phreeza 6y ago
If you enjoy this kind of thing, you may like the Geoguessr game ( https://www.geoguessr.com/ ), and this guy on youtube who is really good at it: https://www.youtube.com/watch?v=xApbqwtnSzs&list=PL_japiE6
826.
by phreeza 6y ago
This idea of FAANGs having a culture of ruthless political backstabbing is new to me, is it really the case? I have heard about things like misaligned incentives (eg launching a chat app to get people promoted and then discontinuing it a ye
827.
by phreeza 6y ago
I think the downvotes were not for asking people to look at it, but because the comment can come off as overly dismissive of the severity of the situation.
828.
by phreeza 6y ago
I'm not doubting that it looks like what you say for the area you looked at. I just wanted to make a more general point that it is not sufficient to look at f'/f and conclude that things will be fine if that is tending toward
829.
by phreeza 6y ago
Thanks for taking the time to write the detailed reponses. Definitely led me to think more closely about these vaguely held intuitions about bias and variance! I think you are exactly right that the crucial aspect is the variance when looki
830.
by phreeza 6y ago
I think it really depends on your underlying dynamics, but I just want to point out that f'/f going to zero does really not imply that f will converge. If for example you are in a still strongly growing (but no longer exponential)
831.
by phreeza 6y ago
If people are infectious for some finite number of days, f'/f will always tend to zero, even if f' stays constant.
832.
by phreeza 6y ago
My understanding is that generally, the error when extrapolating to areas not covered by the training data distribution would be considered to be part of the "bias" part of the bias-variance tradeoff. The way I see it, the varianc
833.
by phreeza 6y ago
The bias/variance trade-off is not really related to extrapolation. Think of a point cloud following a quadratic shape. A linear model will extrapolate terribly.
834.
by phreeza 6y ago
I'd argue that the main driver of volatility over the last few months was the Coronavirus, and not AI...
835.
by phreeza 6y ago
Wouldn't it make sense for them to expand the VC arm until the ROI becomes the same as other departments?
836.
by phreeza 6y ago
Could new physics be lurking at obtainable energies with dark matter and dark energy?
837.
by phreeza 6y ago
This comparison is not entirely fair because the human brain also benefits from priors baked in over the entire course of evolution.
838.
by phreeza 6y ago
Somewhat meta, but great to see that most of the accounts on the original thread are still active. I haven't got a successful startup to show for it, but I have been on HN for more than 10 years, and feel like it has played a significa
839.
by phreeza 6y ago
They also probably don't have a sufficient fraction of their net worth in public stocks to make insider trading worth the risk of the repercussions of being caught doing so.
840.
by phreeza 6y ago
Except this is not an orbit around any body. If you can do what the parent proposes, you can add the chemical energy of fusing the hydrogen to your kinetic energy and defeat the rocket equation

More ›