Posts

Comments

Comment by Eric Askelson (ericask) on Ethan Caballero on Broken Neural Scaling Laws, Deception, and Recursive Self Improvement · 2022-11-04T22:30:54.004Z · LW · GW

Am I just inexperienced or confused, or is this paper using a lot of words to say effectively very little? Sure, this functional form works fine for a given set of regimes of scaling, but it effectively gives you no predictive ability to determine when the next break will occur. 

Sorry if this is overly confrontational, but I keep seeing this paper on Twitter and elsewhere and I'm not sure I understand why.

Comment by Eric Askelson (ericask) on Announcing $5,000 bounty for (responsibly) ending malaria · 2022-09-24T17:32:03.512Z · LW · GW

It's perhaps somewhat morally cowardly to worry about it, but what about legal concerns? In general offering bounties for illegal acts is illegal, and often prosecuted if someone takes you up on the bounty. I don't think eliminating a mosquito species is explicitly illegal in any jurisdictions, but then again, one could certainly imagine prosecution in some scenarios.

You could add a clause like "This bounty will only be paid out if you did not break any laws of the United States in the process." Uncertainty in payouts reduces the value of a bounty, but it may be worth it to encourage additional pledges.

That aside, I think this is an excellent idea. It may be a small amount of money, but you could easily imagine a junior researcher in an org working on gene drives being influenced on the margin by a bounty like this. I'll give some thought to contributing. If I'm convinced legal issues are immaterial I'm likely to match your bounty. 

Comment by Eric Askelson (ericask) on Turn your flashcards into Art · 2022-09-05T16:44:57.385Z · LW · GW

This certainly seems like a way to increase the likelihood that I associate a response with a flashcard in Anki, but will this actually help recall? 

For example, one of my Anki cards might ask how to recursively copy all files with a certain extension out of a directory tree; by adding an image I might strengthen the association between the front and back of the Anki card, but not increase the probability that I actually remember the correct invocation to move files when I really need to. 

This is a failure mode I've noticed before with spaced repetition, where I'll quickly have a response to one of my Anki cards because I only have one question in a given format, even though answering the true question would be harder. 

Is there a different style of flashcard where this wouldn't be a downside? The downside above applies if you associate Question + Image -> Answer, but maybe a custom card type that breaks down each pair into a triple would work? Question -> Image, Image -> Answer, and finally Question -> Answer. I'll give this some thought.