khafra

20 ms·

271.
by khafra 5y ago
"Refutability" is a more general concept than epistemic legibility, in my opinion. One can (at least attempt) to refute abject nonsense that doesn't actually mean anything by saying "that's abject nonsense, that doe
272.
by khafra 5y ago
The reason I buy on Amazon is that finding anything you wouldn't see in a typical department store from somewhere else online takes a bit of effort; and it's an additional effort to gain some confidence that the "somewhere el
273.
by khafra 5y ago
Best practice is to remember that intelligence means optimizing for some state of the world. If you have a "smart" product, it may not be optimizing for your preferred state of the world. Most commonly, it's not even optimizi
274.
by khafra 5y ago
Falsifiability is about grounding out in an experiment that can be performed. Epistemic legibility is a generalization of the concept, where the arguer highlights the cruxes of potential disagreement, and points to whatever evidence they fe
275.
by khafra 5y ago
The trouble is that disagreeing with someone always has at least two effects: Changing viewers' confidence in the original claims, and taking some of the status of the original claimant as a person who makes correct claims. It would be
276.
by khafra 5y ago
One place you'll see this in real life is modded flashlights, running Anduril. My friend has an absurdly powerful light with 1 button and 20 modes: https://budgetlightforum.com/node/38364
277.
by khafra 5y ago
If you can ever build a mosquito defense system like that, the world will beat a path to your door.
278.
by khafra 5y ago
"Fusion is always 30 years away" is a heuristic that almost always works. https://astralcodexten.substack.com/p/heuristics-that-almost...
279.
by khafra 5y ago
The “initial duration X free” I mentioned solves all these problems without producing very onerous complications. Even if it didn’t, a small creator could get immediate cash infusions simply by pricing his newly created characters at a lar
280.
by khafra 5y ago
Pure Harberger taxes aren't a perfect fit for copyright--7 initial years free* is what one fan of the original paper proposed--but I can't see how, compared to existing copyright law, they encourage dishonesty, get messy, or requi
281.
by khafra 5y ago
> If you treat it as a tax (scaling by revenue or similar), it gets messy, encourages dishonesty, and pretty much forces a new subindustry to spring up. Not if you treat it as a Harberger* tax. If you do that, the small-time writer gets
282.
by khafra 5y ago
The quantitative difference is big enough to be qualitative. You have to buy a lot of life insurance, more than people normally get, to make it work. The disinterested third party beneficiary makes this affordable: > There are, however,
283.
by khafra 5y ago
For specifically the doctor version of the principle-agent problem, Robin Hanson published an incentive-compatible solution over 25 years ago: http://mason.gmu.edu/~rhanson/buyhealth.html
284.
by khafra 5y ago
I'll bet spammers assume anyone cautious enough to do the absolute minimum to obscure their email is too hard a target for anything but a higher-investment phishing attempt. The same reason they purposefully include spelling errors in
285.
by khafra 5y ago
Is your theory that nuking the USSR would have completely prevented nuclear proliferation for at least two centuries? That no other nation would have developed nukes in secret, or stolen some from the USA? Or is your theory that nuking Chin
286.
by khafra 5y ago
The book Moral Mazes is the best guide I know of. The basic takeaway is that if you're in middle management, it's already too late, you poor sucker. https://hbr.org/1983/09/moral-mazes-bureaucracy-and-man
287.
by khafra 5y ago
The word "why" (and other references to causation) don't work in a normal way when you're talking about fundamental physics, because fundamental physics is the lowest-level model within which causation takes place. To as
288.
by khafra 5y ago
"Defense vs Offense" is underspecified for this disagreement. Considering "defense" as the developers writing an application, and "offense" as the reverse engineers attempting to exploit it, defense may still b
289.
by khafra 5y ago
When you hold your breath, it's not decreased oxygen that triggers the burning sensation in your lungs and the desperate need for air; it's elevated CO2. So it's pretty easy to perceive if you're at all aware of the sign
290.
by khafra 5y ago
There is a common issue with limiting the angles in trig, and only requiring division with an integer or simple fraction solution: It provides a lot of help to the test-taker, based on their test-taking skills rather than their math skills.
291.
by khafra 5y ago
The old system had too much Slack. The new system has too much Moloch. https://slatestarcodex.com/2020/05/12/studies-on-slack/
292.
by khafra 5y ago
Antifragility advocates might say the occasional ransomware attack on infrastructure could be a good thing, in the long run, if it promotes a more resilient, less just-in-time based economy. Like Amazon's chaos monkey, but for whole ec
293.
by khafra 5y ago
> If we can't ask the AI to explain how it came to its conclusion it is not really intelligent, is it? Will it matter whether the AI is "really" intelligent when companies that implement the black box to spec thrive, while
294.
by khafra 6y ago
Seems like an additional tax on flights where a train is available would be a more sensible measure; especially if the tax revenue were earmarked for climate change mitigation.
295.
by khafra 6y ago
Regarding your final paragraph, the Grabby Aliens model implies that we are within the Anthropic Shadow[1] of the event. E.g., the humanity that lives in the possible universe where the galaxy has already been eaten does not exist. [1] htt
296.
by khafra 6y ago
I really wanted to get one of those, around 2008. Sounded pretty cool; but I wasn't sure how long it would take them to get such an unusual design to market.
297.
by khafra 6y ago
Sounds good, but how do you choose KPIs that are more trouble to game and Goodhart than they are to fulfill honestly? Have a prediction market select them?
298.
by khafra 6y ago
Attribution in security is really hard. But attribution in health supplements is far harder. Nobody's ever gotten a PoC for a health improvement to work every time in living organisms, even on the same organism version and platform.
299.
by khafra 6y ago
No homo economicus would be willing to pay $500,030 for a house when they could pay $500,000.30. 99.99% of humans would be fine with that.
300.
by khafra 6y ago
> it just reinforces existing inequalities ("comparative advantage was reinforced") This is a misunderstanding of comparative advantage. Reinforcing Absolute Advantage would increase inequality. Reinforcing Comparative Advanta

More ›