brilee

16 ms·

151.
by brilee 5y ago
I'm a python readability approver at Google and I don't understand how the import system works
152.
by brilee 5y ago
I estimate the risk of human error (chose the wrong modeling assumptions, bug in data processing code, etc.) at least ~1%, so there really isn't any point in claiming any statistic that is smaller than that.
153.
by brilee 5y ago
It's a space-saving mechanism; since forks are extremely common, sharing commit trees across all of GitHub makes the forking action very cheap.
154.
by brilee 5y ago
Healthcare providers tend to overcharge in proportion to how unnegotiable the service is. For example, ambulance charges are a great way to make money, because the person on the hook for the ambulance typically isn't in a position to d
155.
An Interview with Tenstorrent: CEO Ljubisa Bajic and CTO Jim Keller(anandtech.com)
4 points by brilee 5y ago|0 comments
156.
by brilee 5y ago
google "hydraulic jump" Separately, when you model surface tension, you get an equation for when a stream breaks up into droplets - https://en.wikipedia.org/wiki/Plateau%E2%80%93Rayleigh_insta...
157.
by brilee 6y ago
I did some thinking about this idea on my blog - https://www.moderndescartes.com/essays/homeostasis_volatilit... the short: think of it like the Kelly Criterion. There's an optimal level of risk to take here, whi
158.
Homeostasis and Volatility(moderndescartes.com)
1 points by brilee 6y ago|0 comments
159.
by brilee 6y ago
Ironically, a lot of the tricks Stockfish is using here are reminiscent of tricks that were used in the original AlphaGo and later discarded in AlphaGoZero. In particular, the AlphaGo paper mentioned four neural networks of significance: -
160.
Algorithmic Bias(moderndescartes.com)
1 points by brilee 6y ago|0 comments
161.
by brilee 6y ago
I'd agree with you in theory, but in practice, those API endpoints end up being exploit vectors for data leakage in various ways - either through actual security vulns or through the security vulnerability that exists between the monit
162.
Google Researcher Says She Was Fired over Paper Highlighting Bias in A.I(nytimes.com)
5 points by brilee 6y ago|1 comments
163.
by brilee 6y ago
A few answers: - the infected population tends to be younger this time around (the nursing homes already got cleaned out). - test availability in March was low, so only the severe cases got tested - better treatment awareness (e.g. proning)
164.
by brilee 6y ago
My Apple first-party USBC hub has bizarre ordering issues with my Macbook (have to plug it in, in the order HDMI, pass-through USBc charging, then the USB-A peripheral), and the hub doens't work with my Chromebook. A third-party hub do
165.
by brilee 6y ago
I realized the same thing when trying to compute F1-optimized thresholds; scanning through all possible threshold scores is O(n^2) but computing f1 score inline while scanning through the sorted array is O(n). https://www.modern
166.
A Virtually Unnoticed Migration(moderndescartes.com)
2 points by brilee 6y ago|0 comments
167.
by brilee 6y ago
It doesn't do what you think it does. From https://www.python.org/dev/peps/pep-0616/#rationale "There have been repeated issues on Python-Ideas [2] [3], Python-Dev [4] [5] [6] [7], the Bug Tracker,
168.
by brilee 6y ago
(googler) Not quite sure why you picked those particular examples... JAX also requires usage of lax.cond, lax.while_loop, and ops.segment_sum. Only gather has been improved with slice notation support. IMO, TF has landed on a pretty nice so
169.
Data Oriented Programming in Python(moderndescartes.com)
3 points by brilee 6y ago|1 comments
170.
by brilee 6y ago
It's because a large segment of classical music watchers are using it as a background music playlist, and listen to the same recordings repeatedly.
171.
Data Oriented Programming in Python(moderndescartes.com)
2 points by brilee 6y ago|0 comments
172.
by brilee 6y ago
Disclaimer: I worked on AutoGraph @ google. Your first two paragraphs describe TF 1 perfectly, and you also correctly point out that TF 2 has eager mode, which executes op by op. The part you missed is that @tf.function lets users tell Tens
173.
by brilee 6y ago
The "register account" flow also allows for user enumeration, but you don't hear people complaining about that security hole.
174.
by brilee 6y ago
It isn't idempotent, though - if you execute the query twice, you're kind of screwed, since your inventory numbers are now bad. A better way to do it would be to create a new table that consists of daily snapshots of inventory, se
175.
by brilee 6y ago
Making one move requires 1600 MCTS playouts to explore the game tree, so it's a 1600-1 correspondence of "forward pass" and "move played".
176.
by brilee 6y ago
This blog post doesn't seem to mention how long the process took - that's what I'm most interested in!
177.
by brilee 6y ago
This technique figures out how to find a subset of neural network edge weights that can replicate most of the full model's performance, and is indeed quite simple. The catch is that this sparse subset of NN edge weights has no structur
178.
by brilee 7y ago
Some updates on Computer/Human interaction from a Go AI author. Opinions are my own. In the Go world, the relationship between computer and human has turned into one of analysis. Every pro in Asia has a powerful GPU-equipped system wit
179.
by brilee 7y ago
I'm enjoying all the comments here! Thank you OP for demonstrating the effective use of Cunningham's law: "the best way to get the right answer on the internet is not to ask a question; it's to post the wrong answer.&quo
180.
by brilee 7y ago
The Polish test distribution is an interesting counterexample: it demonstrates that there is _not_ a discontinuity. Instead of failing immediately at a score of 29, the graph suggests that you start failing probabilistically, with greater p

More ›