LESSWRONG
LW

gilch
27301694021
Message
Dialogue
Subscribe

As of January 2025, I have signed no contracts or agreements whose existence I cannot mention.

Sequences

Posts

Sorted by New

Wikitag Contributions

Comments

Sorted by
Newest
An Apprentice Experiment in Python Programming
Inefficient Markets
3gilch's Shortform
6y
26
Open thread, Apr. 24 - Apr. 30, 2017
gilch4mo40

Eight years later, but we're finally approaching the technical capability to perform my proposed experiment to elicit tetrachromacy in a human trichromat through eye tracking and targeting a subset of retinal cells: https://www.science.org/doi/10.1126/sciadv.adu1052#sec-3

See the DISCUSSION section proposing further experiments.

Reply
The case for aftermarket blind spot mirrors
gilch2y195

I feel like this has come up before, but I'm not finding the post. You don't need the stick-on mirrors to eliminate the blind spot. I don't know why pointing side mirrors straight back is still so popular, but that's not the only way it's taught. I have since learned to set mine much wider.

This article explains the technique. (See the video.)

In a nutshell, while in the diver's seat, tilt your head to the left until it's almost touching your window, then from that perspective point it straight back so you can just see the side of your car. (You might need a similar adjustment for the passenger's side, but those are often already wide-angle.) Now from normal position, you can see your former "blind spot". When you need to see straight back in your side mirror (like when backing out), just tilt your head again. Remember that you also have a center mirror. You should be able to see passing cars in your center mirror, and then in your side mirror, then in your peripheral vision without ever turning your head or completely losing sight of them.

Reply1
What can we learn from Bayes about reasoning?
gilch2y*238
  • It's not enough for a hypothesis to be consistent with the evidence; to count in favor, it must be more consistent with the hypothesis than its converse. How much more is how strong. (Likelihood ratios.)
  • Knowledge is probabilistic/uncertain (priors) and is updated based on the strength of the evidence. A lot of weak evidence can add up (or multiply, actually, unless you're using logarithms).
  • Your level of knowledge is usually not literally zero, even when uncertainty is very high, and you can start from there. (Upper/Lower bounds, Fermi estimates.) Don't say, "I don't know." You know a little.
  • A hypothesis can be made more ad-hoc to fit the evidence better, but this must lower its prior. (Occam's razor.)
    • The reverse of this also holds. Cutting out burdensome details makes the prior higher. Disjunctive claims get a higher prior, conjunctive claims lower.
    • Solomonoff's Lightsaber is the right way to think about this.
  • More direct evidence can "screen off" indirect evidence. If it's along the same causal chain, you're not allowed to count it twice.
  • Many so-called "logical fallacies" are correct Bayesian inferences.
Reply
What confusions do people have about simulacrum levels?
gilch5y*270

French, but because my teacher tried to teach all of the days of the week at the same time, they still give me trouble.

They're named as the planets: Sun-day, Moon-day, Mars-day, Mercury-day, Jupiter-day, Venus-day, and Saturn-day.

It's easy to remember when you realize that the English names are just the equivalent Norse gods: Saturday, Sunday and Monday are obvious. Tyr's-day (god of combat, like Mars), Odin's-day (eloquent traveler god, like Mercury), Thor's-day (god of thunder and lightning, like Jupiter), and Freyja's-day (goddess of love, like Venus) are how we get the names Tuesday, Wednesday, Thursday, and Friday.

Reply
LessWrong 2.0
gilch10y210

Why is Google the biggest search engine even though it wasn't the first? It's because Google has a better signal-to-noise ratio than most search engines. PageRank cut through all the affiliate cruft when other search engines couldn't, and they've only continued to refine their algorithms.

But still, haven't you noticed that when Wikipedia comes up in a Google search, you click that first? Even when it's not the top result? I do. Sometimes it's not even the article I'm after, but its external links. And then I think to myself, "Why didn't I just search Wikipedia in the first place?". Why do we do that? Because we expect to find what we're looking for there. We've learned from experience that Wikipedia has a better signal-to-noise ratio than a Google search.

If LessWrong and Wikipedia came up in the first page of a Google search, I'd click LessWrong first. Wouldn't you? Not from any sense of community obligation (I'm a lurker), but because I expect a higher probability of good information here. LessWrong has a better signal-to-noise ratio than Wikipedia.

LessWrong doesn't specialize in recipes or maps. Likewise, there's a lot you can find through Google that's not on Wikipedia (and good luck finding it if Google can't!), but we still choose Wikipedia over Google's top hit when available. What is on LessWrong is insightful, especially in normally noisy areas of inquiry.

Reply
Please speak unpredictably
gilch8d20

Beware transparency's illusion! Links help?

Reply
Open thread, Apr. 24 - Apr. 30, 2017
gilch4mo20

See the paper I linked in my other comment.

Reply
So how well is Claude playing Pokémon?
gilch6mo153

Update: Claude made it to Cerulean City today, after wandering the Mt. Moon area for 69 hours.

Reply
Open Thread Winter 2024/2025
gilch7mo30

See https://pauseai.info. They think lobbying efforts have been more successful than expected, but politicians are reluctant to act on it before they hear about it from their constituents. Individuals sending emails also helps more than expected. The more we can create common knowledge of the situation, the more likely the government acts.

Reply
How do you deal w/ Super Stimuli?
gilch8mo20

It's also available on Android.

Reply
Load More
6Effective Altruists and Rationalists Views & The case for using marketing to highlight AI risks.
1y
1
14Linkpost for Accursed Farms Discussion / debate with AI expert Eliezer Yudkowsky
2y
2
14An Apprentice Experiment in Python Programming, Part 4
4y
2
14An Apprentice Experiment in Python Programming, Part 3
4y
10
30An Apprentice Experiment in Python Programming, Part 2
4y
18
67An Apprentice Experiment in Python Programming
4y
4
26Should I take glucosamine?
Q
5y
Q
13
37Charting Is Mostly Superstition
5y
18
27Market Misconceptions
5y
8
22The Wrong Side of Risk
5y
12
Load More
Information Hazards
2y
(+4/-4)
Black Marble
2y
(+70/-9)
Information Hazards
2y
(+26)
Black Marble
2y
(+25/-1)
Black Marble
2y
(+94/-16)
Black Marble
2y
(+989)
Egregores
2y
(+7/-7)
Egregores
2y
(+79/-33)
Egregores
2y
(+163)
Egregores
2y
(+534)
Load More