Viewing profile — cocomutator
cocomutator
HN member- Joined
- Thu, Jun 16, 2016, 4:17 AM UTC
- HN karma
- 45
- Public activity
- 18 items
- HN profile
- View on Hacker News ↗
About cocomutator
No profile information was provided.
Recent public activity
-
comment
Comment #46469440
My life just changed forever too...
-
comment
Comment #45815811
today is the day i stopped reading opinion pieces from the technologyreview. not for presenting an opinion i don’t agree with, but for mistaking word soup for an argument.
-
comment
Comment #43267786
ah I see what you mean, sorry it took me a while. Yes, you're right, the two are dual: "fix complexity budget then optimize data error", and "fix data error budget then optimize cx…
-
comment
Comment #43266687
Right, but see how the complexity budget is prescribed ahead of time: we first set the regularization strength or whatever, and then optimize the model. The result is the best mode…
-
comment
Comment #43266633
I agree with your observation about the exact noise-free nature of the problem. It allows them to formulate the problem as "minimize complexity such that you memorize the X-y relat…
-
comment
Comment #43266216
I'm trying to distill the essence of their approach, which imho is concealed behind inessential and particular details such as the choice of this or that compression scheme or prio…
- story
-
comment
Comment #34803199
Bingotry (n. uncount.) /ˈbɪŋɡətɹi/ 1. The futile attempt by Microsoft to equip their search engine with artificial intelligence. See also: Bing (Microsoft search engine), Try (v.).…
-
comment
Comment #34574870
Not exponentially, just inverse-quadratically :)
-
comment
Comment #34503910
Here's GPT's own explanation what the purpose of that while loop is: --- This code uses JavaScript's `eval` function to obfuscate the code by looping over an array of strings and p…
-
comment
Comment #30477314
I shall learn from this post how to disagree respectfully on the internet. Hats off.
-
comment
Comment #29973484
I wonder how a modern language model would fare here -- use something like GPT-3 to evaluate the log likelihood gain of stitching together each of all N^2 possible pairs, then merg…
- story
- story
-
comment
Comment #15371407
as someone with a degree in math, this is the funniest thing I read in a while.
- story
- story
-
comment
Comment #11913953
I still don't understand why this ME feature has been created to begin with. Assuming that breaking it is a matter of time (someone clever enough thinking about it for long enough)…