Dagon

Just this guy, you know?

Wiki Contributions

Comments

Philosophical considerations of cessation of brain activity

I don't think there's a consensus on this, and I suspect it's because there's no truth to be had.  There's no territory that matches any map of these concepts.  

Personally, I model my identity as gradually changing, not constant.  Me and yesterday-me are very similar, but there are a lot of differences from 25-year-ago-me.  Legally, debts and assets follow, but psychologically that person is only partly-me.   Embedded in this model is that I'm using a distance model of brain-state, not a continuity measure of how gradual or smooth the changes have been.

Anesthesia is probably more change than a night's sleep, but less than a year's activity and sleeps.  

Wormy the Worm

temporarily dissolve the typical narrow conception of identity 

I think a lot of discussion around here is already different from the typical narrow conception of identity.  You'll often find people talking about "past me" as an indicator that there are aspects of identity which are best modeled as a different person.  One unsolved problem in probability is "to what extent is this reference class useful for this prediction", and that gets to the underlying point I think you're trying to make: identity is a cluster of different types of similarity,  which only make sense when specifying what you intend to infer with the concept.

and draw out a feeling of unity among all forms of life.

Wow, that's a stretch.  In me, at least, it draws out a feeling of isolation from previous and future versions of me.  In dissolving the notion of identity, why would that make me closer to things I now even more clearly recognize as "other"?

Framing Practicum: Dynamic Programming

I think these examples show some missing requirements in the set of problems suitable for dynamic programming.  Primarily, that the sub-problems must be very similar to the main problem, such that the same solution applies to them.  

DP is nothing more than recursion + memoization.  Recursion is solving a problem by breaking it into a trivial combination of smaller problems, and repeating that process until the smallest problem is itself trivial so you don't need to break it down further.   Memoization is just remembering the results of previous sub-problems, because you'll often need them multiple times.   

The OP misses out on the first step: defining the problem, and skips to the second, adding up the results.  DP is a way to notice that you'll have to solve it in reverse, but you still approach it forward.  You write the code that knows the best way from a node to Boston is the lowest sum of "cost to next node" and "cost of THAT node to boston".  So Miami's best choice is the lowest of 200+A->Boston or 170+B->Boston.  We don't know either of those sub-values yet, but we know how to calculate them - the exact same way, the lower of cost-of-path+cost-of-node-to-Boston.  Until we're one away from Boston, then it's just cost-of-path-to-Boston.

In setting up this problem this way, it becomes clear that we're performing calculations backwards from Boston, but we don't need to figure that out specifically, it's built into the nature of DP.

EniScien's Shortform

Be careful with thinking a phenomenon is meaningless or nonexistent just because it's an abstraction over an insanely complex underlying reality.  Even if you're unsure of the mechanism, and/or can't calculate how it works in detail, you're probably best off with a decision theory that includes some amount of volition and intent.  And moral systems IMO don't have an objective real cause, but they can still carry a whole lot of power as coordination points and shared expectations for groups of humans.

Formula for Dying Babies

The FDA, like most federal agencies, has immunity for most liability.  You cannot sue them for approving a drug that kills someone, you cannot sue them for failing to approve a necessary drug.  You cannot sue them for misrepresentation or incompetence, regardless of harm.

Federal prosecutors may be able to charge individuals at the FDA with fraud or crimes, if egregious enough.  But I don't know of it ever happening, and it would not happen without pretty significant evidence of malice rather than just incompetence or misaligned incentives.

lc's Shortform

This doesn't seem like an either-or question.  Freer trade and more individual interactions seem complementary to me.

Well, like all such probability questions, it depends on your priors (or, if you're a frequentist, the distribution of possible experiments).  And on the specifics of the lotteries and their auditors.

Assuming this happened (I haven't seen any reporting of it, so it's not big-money lotteries), it's good evidence for interference, and that should increase the probability of extraterrestrial, religious, and criminal hypotheses.  The first two increase from near-zero to still-near-zero, and the last one from low to pretty likely.

For me, I find shenanigans much more likely than some being outside of physics.  Perhaps it's rigged, or misreported, or the woman found a flaw in the mechanism.  All are far more believable than a conscious being who can meddle in outcomes on that level, and chooses ONLY to do it to benefit this woman massively, not to help the rest of humanity.

 

EniScien's Shortform

I'm not sure I get the analogy.  And in fact, I don't think that KNOWING the movements of atoms is sufficient to allow you to avoid heat transfer to your finger.  You'd need it to be true that there exists a place and duration sufficient to dip your finger that would not burn you.  I don't think this is true for any existing or likely-to-exist case.

If you can CONTROL the movements of atoms, you certainly can avoid burning.  This is common and boring - either a barrier/insulator or just cooling the water first works well.

How to invest in expectation of AGI?

"How to invest" does not (and CAN NOT) have simple passive long- or medium-term strategies for a general scenario like this.  What you should ask yourself are the following:

  • Does this timeline make my investing goals different from the median investor?  Am I looking for a large-win-or-total-loss risk profile because I think everything is going to change radically?  Are there even any investments that pay off in the dimensions I think will be important in an AGI world?  Maybe the best "investment" is in experiences that will become unavailable in the future.  Or in guns, training, and EMP devices for the coming wars.  OK, probably not.
  • What skills or knowledge do I have that "the market" doesn't, or isn't using very effectively.  How active will I be in my investments (both in working at the things you're investing in AKA sweat equity, and in rapidly moving your capital around as you get better information)?
  • How do my investments actually affect what happens?  Especially if you're mixing work and investment, is it in areas you want to win?

Note: I'm mostly investing monetarily using fairly standard "just get market returns" vehicles, with an outsized real-estate portion.  This implies that I assign a small-ish chance to truly radical change on a decade or two timeline.  I recommend this for almost everyone who's primarily passive in their investments (put money in, but don't actively manage it and don't have any particular passion, skill, or knowledge that amplifies the investment value).

Load More