MontyCarloHall

8 ms·

91.
by MontyCarloHall 9mo ago
Counter-counterpoint: people pay exorbitant amounts of money for cable TV channels that still show ads. Even the premium channels (HBO et al.) implicitly show ads in the form of product placement, which incidentally is exactly how I think c
92.
by MontyCarloHall 9mo ago
I think ads will inevitably roll out across all tiers, even the expensive paid ones. Ad revenue isn't uniformly distributed across users, but rather heavily skewed towards the wealthiest users, exactly the users most able to purchase a
93.
by MontyCarloHall 9mo ago
Were many of the aforementioned advancements marketed as "replacing developers"? Absolutely. Did that end up happening? Quite the opposite; each higher-level abstraction only caused the market for software and demand for developer
94.
by MontyCarloHall 9mo ago
It's not so much about replacing developers, but rather increasing the level of abstraction developers can work at, to allow them to work on more complex problems. The first electronic computers were programmed by manually re-wiring th
95.
by MontyCarloHall 9mo ago
>What are you basing this off of? Do you actually have any experience making art? Or is this just learned helplessness talking? I've done ≥weekly life drawing classes for 20+ years, and have observed the distribution of progress peo
96.
by MontyCarloHall 9mo ago
I very much doubt the majority of adults with sufficient practice could do this level of work. I can say with 100% certainty that an infinitesimal minority of 12 year olds with sufficient practice could do this level of work.
97.
by MontyCarloHall 9mo ago
>Anyone can do this level of work - they just need to actually learn it. Sorry, that's like saying with enough math practice, any kid could perform at the level of young Terry Tao (e.g. teaching himself calculus at 8, winning a gold
98.
by MontyCarloHall 9mo ago
Of course it's much better, Schongauer was ~25 when he did the engraving. Michelangelo was 12 when he copied it. Likewise, it goes without saying that Haydn's symphonies circa 1765 were much better than Mozart's from the same
99.
by MontyCarloHall 9mo ago
Agreed! I knew about the second chance pool but didn't realize it reset all the timestamps.
100.
by MontyCarloHall 9mo ago
I'd call it "Stochastic Newton's Method" then. :-)
101.
by MontyCarloHall 9mo ago
>If the Hessian-vector product is Hv for some fixed vector v, we're interested in solving Hx=v for x. The hope is to soon use this as a preconditioner to speed up stochastic gradient descent. Silly question, but if you have some cle
102.
by MontyCarloHall 9mo ago
Bizarre; this was originally posted 5 days ago (as evidenced by clicking on the poster's submission history [0]), but timestamps on the submission and on comments throughout the post show 11 hours ago. @dang, is this a bug in HN softwa
103.
by MontyCarloHall 9mo ago
It is indeed a builtin, but `/bin/[` still exists for compatibility reasons! $ which [ /bin/[ $ type [ [ is a shell builtin The same is true for the `test` command: $ which test /bin/te
104.
by MontyCarloHall 9mo ago
If it {tastes,feels,looks} like eating shrimp, I don't see why people wouldn't go for it. After all, much weirder looking bugs of the sea (e.g. lobsters, crabs) are considered luxuries.
105.
by MontyCarloHall 9mo ago
All user data is E2E encrypted, so the government literally cannot force this. This has been the source of numerous disputes [0, 1] that either result in the device itself being cracked [0] (due to weak passwords or vulnerabilities in devic
106.
by MontyCarloHall 9mo ago
Because their focus on user privacy makes it difficult for them to train at scale on users' data in the way that their competitors can. Ironically, this focus on privacy initially stemmed from fumbling the ball on Siri: recall that App
107.
by MontyCarloHall 10mo ago
A neural-GP could probably be trained with the same parallelization efficiency via consistent discretization of the input space. I think their absence owes more to the fact that discrete data (namely, text) has dominated AI applications. I
108.
by MontyCarloHall 10mo ago
It's utterly baffling to me that there hasn't been more SOTA machine learning research on Gaussian processes with the kernels inferred via deep learning. It seems a lot more flexible than the primitive, rigid dot product attention
109.
by MontyCarloHall 10mo ago
100% agreed. Attention finally clicked for me when I realized "wait, it's just a transformed, weighted dot product and has nothing to do with key/value lookups." I would have gotten this a lot faster had they called the
110.
by MontyCarloHall 10mo ago
The confusing thing about attention in this article (and the famous "Attention is all you need" paper it's derived from) is the heavy focus on self-attention . In self-attention, Q/K/V are all derived from the same
111.
by MontyCarloHall 10mo ago
>They all are on foliage, which changes depending on season and wind. Geolocating that picture accurately _automatically_ using features is next to impossible. Seems plausible enough to me. The trees are evergreens in a place that doesn&
112.
by MontyCarloHall 10mo ago
They both fit Gaussians, just different ones! OLS fits a 1D Gaussian to the set of errors in the y coordinates only, whereas TLS (PCA) fits a 2D Gaussian to the set of all (x,y) pairs.
113.
by MontyCarloHall 10mo ago
>But at a more fundamental level, our material quality of life is that of literal kings. This cannot be overstated. To wit, a Honda Accord (or equivalent mid-range car of today) is objectively superior to a Rolls Royce from the 90s in te
114.
by MontyCarloHall 10mo ago
Exactly. If you're working on an application where these numbers matter, Python is far too high-level a language to actually be able to optimize them.
115.
by MontyCarloHall 10mo ago
I agree it's incredibly stupid. Just saying that our current society's valuation of scarcity makes access to an elite education very much a zero-sum game.
116.
by MontyCarloHall 10mo ago
Very interesting observation. Multicore systems have been fairly standard for the last 10+ years, and while you occasionally notice a misbehaving process hog an entire core, it never visibly impacts system performance because there are stil
117.
by MontyCarloHall 10mo ago
>Celeron CPUs were usually CPUs that shared the same core architecture as the current Pentium standard, but often had a lower core clock speed, lower core memory speed, and/or had smaller L2 caches. This had an interesting side effe
118.
by MontyCarloHall 10mo ago
>The fact of resource extraction from society and externalities like pollution not being counted by capitalist because they "can’t count them" and just bundle them as externalities demonstrably destroy any concept of non-zero s
119.
by MontyCarloHall 10mo ago
>The difference between these two sides is which value they prioritize. Yup. Though there is a third option: completely ignore the meritocracy vs. equity zero-sum game and simply argue that demographic-based weighting of applicants is an
120.
by MontyCarloHall 10mo ago
>Indeed, I'm sure the author would agree that part of the comprehensive solution is to increase the amount of university admission slots. A large part of the value of elite education is its scarcity, and adding more slots dilutes th

More ›