jmount

7 ms·

91.
by jmount 1y ago
Really great article. Internalizing that a good realizable Mastermind strategy is to always eliminate the same number of possible solutions is a great way to internalize the value of information theory thinking. Getting hung up if it is exa
92.
by jmount 1y ago
all of these prompts are just making the responses appear critical. just more subtle fawning.
93.
Escape the over Tuning Trap(win-vector.com)
2 points by jmount 1y ago|0 comments
94.
Save Your Optimization Project from Overestimation Bias(winvector.github.io)
2 points by jmount 1y ago|0 comments
95.
Reducing Optimization Risk Through Portfolio Methods(win-vector.com)
3 points by jmount 1y ago|0 comments
96.
Modeling Uncertainty for Better Outcomes(win-vector.com)
3 points by jmount 1y ago|1 comments
97.
by jmount 1y ago
Really makes me wonder if this is a reproduction of a pattern of interaction from the QA phase of LLM refinement. Either way it must be horrible to be QA for these things.
98.
Computer science: how to squeeze space into time [video](youtube.com)
1 points by jmount 1y ago|0 comments
99.
Unexpected proof of the Gaussian correlation conjecture [video](youtube.com)
1 points by jmount 1y ago|0 comments
100.
by jmount 1y ago
Just ran into your writeup a day ago. Truly amazing!
101.
Toying with the Lambda Calculus(github.com)
26 points by jmount 1y ago|1 comments
102.
They did use a calculator for this 1924 puzzle(win-vector.com)
2 points by jmount 1y ago|0 comments
103.
by jmount 1y ago
As a fan of combinatorics, I love people looking into the topic. Though most of what a programmer would find useful are approximations or trends which are handled well by the Master theorem ( https://en.wikipedia.org/wiki&#x
104.
How About Pi – 31/32?(win-vector.com)
2 points by jmount 1y ago|2 comments
105.
by jmount 1y ago
I have added a crude heuristic argument that the number of labels Whitehill was able to expose per submission (6) may in fact be what one would expect: https://github.com/WinVector/Examples/blob/main/puzz
106.
Don't Let a Data Leak Sink Your Project(win-vector.com)
3 points by jmount 1y ago|1 comments
107.
by jmount 1y ago
Ah man- the power supply in the IMSAI 8080 was scary enough, plus you had the monitor power supply open. Fun times- the "book" sequence on ours was "fat finger in the paper tape reading software, read the cassette IO software
108.
by jmount 1y ago
Not the same effect: but a good related writeup: https://www.stefanmesken.info/machine%20learning/how-to-beat...
109.
Working Through a Trivial Algorithm Whose Analysis Isn't(winvector.github.io)
2 points by jmount 1y ago|0 comments
110.
by jmount 1y ago
Really exciting to see so many of the early systems revived. Eliza, Logic Theorist, SHRDLU ( https://hci.stanford.edu/winograd/shrdlu/ , I had thought this one was lost). Any other favorites?
111.
by jmount 1y ago
I grew up on 8088 and then 6502 and 8086 assembly. The 8086 had more registers and more useful variations of registers. However, it emotionally felt like after eliminating the accumulator and stack pointer: the 6502 had general purpose regi
112.
by jmount 1y ago
The chart is mixing things that are very minor at scale (such as choosing Gaussian versus Student-T versus cross-entropy losses) and things that are major (such as representation). I'd say it is more like the notes one prepares for one
113.
by jmount 1y ago
I'd worry a bit about the summary of what is being reported. > Facebook and Instagram deactivation improved emotional state index by 0.060 standard deviations (p < 0.001) The link didn't click through to the appendix. This s
114.
by jmount 1y ago
Or Intel.
115.
Demonstrating Kelly Betting with Chips(win-vector.com)
2 points by jmount 1y ago|0 comments
116.
Zero variance Kelly betting on shuffled cards again [video](youtube.com)
2 points by jmount 1y ago|0 comments
117.
by jmount 2y ago
Some prior works that work as prequels include C.M. Kornbluth's "The Marching Morons" and "The Little Black Bag."
118.
by jmount 2y ago
If they don't want to use Scala, why do they thing others will want to use Scala?
119.
Ask HN: Source of LLM model use by date and model name data?
1 points by jmount 2y ago|0 comments
120.
by jmount 2y ago
It is complicated. Tom Waits won a case where a commercial imitated his style.

More ›