brilee

10 ms·

91.
by brilee 4y ago
Appendix H in the paper does a good job of explaining why this cyclic topology thing makes sense - "David Wu’s hypothesis is that information propagates through the neural network in a way analagous to going around the cycle, but it ca
92.
by brilee 4y ago
Congrats! I see by comparing the author list of v1 and V2 of this paper[1] that a Go player[2] was consulted before making claims about defeating a Go AI. This is indeed a nontrivial exploit that passes muster with a Go player who was skept
93.
A Research Codebase Manifesto(moderndescartes.com)
3 points by brilee 4y ago|0 comments
94.
by brilee 4y ago
So the naive MCTS implementation in Python is ridiculously inefficient. Of course, you could reimplement it in C++ but this then requires you to use the C wrappers of Tensorflow/JAX to do the MCTS/neural network interop. I came up
95.
by brilee 4y ago
> If you have the research paper, someone in the field could reimplement them in a few days. Hi I did this while I was at Google Brain and it took our team of three more like a year. The "reimplementation" part took 3 months or
96.
A Research Codebase Manifesto(moderndescartes.com)
1 points by brilee 4y ago|0 comments
97.
by brilee 4y ago
A lot of transuranic elements exist for microseconds and just barely long enough for spectroscopic measurements to be taken. Yet, they're granted a spot at the end of the periodic table. Humans like categorizing things, perhaps to a fa
98.
by brilee 4y ago
Only FTE massage therapists were laid off, and their positions were reopened as contractor positions. It's a shitty but not unexpected move. Google's first masseuse was FTE and got to be part of the IPO, so the symbolism is rich..
99.
by brilee 4y ago
As somebody from the U.S., I'm struck by how much people from the U.K. value a strong background in theory. At Google my experience has been "what matters is how smart you are, whether you understand the problem we're trying
100.
by brilee 4y ago
(disclaimer: am googler) Just like the thesis of the slime mold talk, the talk itself is a result of the slime mold and doesn't particularly represent any company-authoritative take on the situation. It's turtles all the way down!
101.
by brilee 4y ago
PWS is great but not scaleable. https://www.quidnetenergy.com/solution/ is one possible scaleable alternative that I haven't been able to crunch numbers on (the website details are scarce)
102.
by brilee 4y ago
The report did say 72M for just electrolysis, but I used 120M in the analysis, an estimated +66% markup for all the other factors that you correctly point out as missing.
103.
by brilee 4y ago
6% curtailment seems pretty trivial in the grand scheme of things. Around 20-25% curtailment, it becomes economical to run those hydrogen electrolyzers part-time, even with the capital expenditures required. I wouldn't use that hydroge
104.
The FactoBattery(moderndescartes.com)
32 points by brilee 4y ago|31 comments
105.
by brilee 4y ago
Just use the grid, they can move energy through time a lot more efficiently than you can ever hope to do. https://www.moderndescartes.com/essays/factobattery/
106.
The Factobattery(moderndescartes.com)
2 points by brilee 4y ago|0 comments
107.
I was fired from HubSpot in 2016(moderndescartes.com)
2 points by brilee 4y ago|0 comments
108.
by brilee 4y ago
Another issue is that at pressures below one atmosphere, the system will suck in ambient air, causing condensation to collect inside your refrigeration circuit, which creates ice crystals, etc… nobody does that.
109.
by brilee 4y ago
I wrote up a brief discussion here on the boiling points of refrigerants https://www.moderndescartes.com/essays/refrigerants/ The long and short of it is that if the heat pump works below -20F, then the boiling po
110.
by brilee 4y ago
This trick is called zobrist hashing in chess/go programming. It allows for incremental calculation of the board hash, which is particularly useful because all of the game variations you're exploring are all nearly identical to on
111.
by brilee 4y ago
It actually isn't saying that. What do you think Python is made of, under the hood? It's C modules. The argument is not that NumPy is written in C, but that it amortizes the cost of Python overhead over multiple data, rather than
112.
by brilee 4y ago
Doesn't this end up being effectively a binary heap, with a maximum tree depth of 23 (floating point mantissa precision)? I imagine there must be a rebalancing operation required every so often, possibly more frequently for pathologica
113.
For Want of a Join(moderndescartes.com)
1 points by brilee 4y ago|0 comments
114.
Data-Oriented Programming in Python(moderndescartes.com)
154 points by brilee 4y ago|44 comments
115.
by brilee 4y ago
Good question. I was also suspicious of the angle/framing of the paper as a smokestack capture solution. My thoughts are 1. smokestacks are a great place to start, it's thermodynamically easier, has low-hanging fruit, and can be a
116.
by brilee 4y ago
Chemist here. Formate does indeed act as a reducing agent under appropriate conditions, but its byproduct is typically CO2 gas, not formaldehyde (HCOO- -> CO2 + H+ + 2e-). Aldehydes are generally unstable with respect to disproproportion
117.
by brilee 4y ago
There are two families of rules used in Go - computer-unambiguous rulesets that enable computers to decide a winner, which commonly results in 1000-move games, and human rules, in which humans can decide that the game is over when they agre
118.
by brilee 4y ago
3 years is deceptive - in an exponentially growing company, the majority of employees are fresh, pushing the average tenure low. The relevant number is "length of tenure at time of severance", which I believe skews much higher. An
119.
by brilee 4y ago
It's frustrating that your link to an "explanation" is an intentionally buggy implementation of the algorithm :( I understand the point of the debugging game, but in this context, a link to the solutions page would be more he
120.
by brilee 4y ago
https://www.google.com/books/edition/Chemicals_Evaluated_as_... Here's your data. The "machine digitization" is referring to OCR of the PDFs and subsequent translation of archaic chemical names into

More ›