jchan
It’s a question of whether drawing a boundary on the “aligned vs. unaligned” continuum produces an empirically-valid category; and to this end, I think we need to restrict the scope to the issues actually being discussed by the parties, or else every case will land on the “unaligned” side. Here, both parties agree on where they stand vis-a-vis C and D, and so would be “Antagonistic” in any discussion of those options, but since nobody is proposing them, the conversation they actually have shouldn’t be characterized as such.
On the contrary, I’d say internet forum debating is a central example of what I’m talking about.
This “trying to convince” is where the discussion will inevitably lead, at least if Alice and Bob are somewhat self-aware. After the object-level issues have been tabled and the debate is now about whether Alice is really on Bob’s side, Bob will view this as just another sophisticated trick by Alice. In my experience, Bob-as-the-Mule can only be dislodged when someone other than Alice comes along, who already has a credible stance of sincere friendship towards him, and repeats the same object-level points that Alice made. Only then will Bob realize that his conversation with Alice had been Cassandra/Mule.
(Example I’ve heard: “At first I was indifferent about whether I should get the COVID vaccine, but then I heard [detestable left-wing personalities] saying I should get it, so I decided not to out of spite. Only when [heroic right-wing personality] told me it was safe did I get it.”)
#1 - I hadn’t thought of it in those terms, but that’s a great example.
#2 - I think this relates to the involvement of the third-party audience. Free speech will be “an effective arena of battle for your group” if you think the audience will side with you once they learn the truth about what [outgroup] is up to. Suppose Alice and Bob are the rival groups, and Carol is the audience, and:
Alice/Bob are SE/SE (Antagonist/Antagonist)
Alice/Carol are SF/IE (Guru/Rebel)
Bob/Carol are IF/SE (Siren/Sailor)
If this is really what’s going on, Alice will be in favor of the debate continuing because she thinks it’ll persuade Carol to join her, while Bob is opposed to the debate for the same reason. This is why I personally am pro-free-speech—because I think I’m often in the role of Carol, and supporting free speech is a “tell” for who’s really on my side.
I think this is not a great example because the virtues being extolled here are orthogonal to the outcome.
Would it still be possible to explain these virtues in a consequentialist way, or is it only some virtues that can be explained in this way?
And consequentialists can choose to value their own side more than the other side, or to be indifferent between sides, so I’m not sure what the conflict between virtue ethics and consequentialism would be here.
The special difficulty here is that the two sides are following the same virtue-ethics framework, and come into conflict precisely because of that. So, whatever this framework is, it cannot be cashed out into a single corresponding consequentialist framework that gives the same prescriptions.
It could be that people regard the likelihood of being resurrected into a bad situation (e.g. as a zoo exhibit, a tortured worker em, etc.) as outweighing that of a positive outcome.
Aren’t there situations (at least in some virtue-ethics systems) where it’s fundamentally impossible to reduce (or reconcile) virtue-ethics to consequentialism because actions tending towards the same consequence are called both virtuous and unvirtuous depending on who does them? (Or, conversely, where virtuous conduct calls for people to do things whose consequences are in direct opposition.)
For example, the Iliad portrays both Achilles (Greek) and Hector (Trojan) as embodying the virtues of bravery/loyalty/etc. for fighting for their respective sides, even though Achilles’s consequentialist goal is for Troy to fall, and Hector’s is for that not to happen. Is this an accurate characterization of how virtue-ethics works? Is it possible to explain this in a consequentialist frame?
Thanks everyone for coming! Feedback survey here: https://forms.gle/w32pisonKdwK1bHJ6
It’s also nice to be able to charge up in a place where directly plugging in your device would be inconvenient or would risk theft, e.g. at a busy cafe where the only outlet is across the room from your table.
I want to say something like: “The bigger N is, the bigger a computer needs to be in order to implement that prior; and given that your brain is the size that it is, it can’t possibly be setting N=3↑↑↑↑↑3.”
Now, this isn’t strictly correct, since the Solomonoff prior is uncomputable regardless of the computer’s size, etc. - but is there some kernel of truth there? Like, is there a way of approximating the Solomonoff prior efficiently, which becomes less efficient the larger N gets?
I’m unsure whether it’s a good thing that LLaMA exists in the first place, but given that it does, it’s probably better that it leak than that it remain private.
What are the possible bad consequences of inventing LLaMA-level LLMs? I can think of three. However, #1 and #2 are of a peculiar kind where the downsides are actually mitigated rather than worsened by greater proliferation. I don’t think #3 is a big concern at the moment, but this may change as LLM capabilities improve (and please correct me if I’m wrong in my impression of current capabilities).
Economic disruption: LLMs may lead to unemployment because it’s cheaper to use one than to hire a human to do the same work. However, given that they already exist, it’s only a question of whether the economic gains accrue to a few large corporations or to a wider mass of people. If you think economic inequality is bad (whether per se or due to its consequences), then you’ll think the LLaMA leak is a good thing.
Informational chaos: You can never know whether product reviews, political opinions, etc. are actually genuine expressions of what some human being thinks rather than AI-generated fluff created by actors with an interest in deceiving you. This was already a problem (i.e. paid shills), but with LLMs it’s much easier to generate disinformation at scale. However, this problem “solves itself” once LLMs are so easily accessible that everyone knows not to trust anything they read anyway. (By contrast, if LLMs are kept private, AI-generated content seems more trustworthy because it comes in a wider context where most content is still human-authored.)
Infohazard production: If e.g. there’s some way of building a devastating bioweapon using household materials, then it’d be really bad if LLaMA made this knowledge more accessible, or could discover it anew. However, I haven’t seen any evidence that LLaMA is capable of discovering new scientific knowledge that’s not in the training set, or that querying it to surface existing such knowledge is any more effective than using a regular search engine. But this may change with more advanced models.
One time, a bunch of particularly indecisive friends had started an email thread in order to arrange a get-together. Several of them proposed various times/locations but nobody expressed any preferences among them. With the date drawing near, I broke the deadlock by saying something like “I have consulted the omens and determined that X is the most auspicious time/place for us to meet.” (I hope they understood I was joking!) I have also used coin-flips or the hash of an upcoming Bitcoin block for similar purposes.
I think the sociological dynamic is something like: Nobody really cares what we coordinate on, but they do care about (a) not wanting to be seen as unjustifiably grabbing social status by imposing a single choice on everyone else, and (b) not wanting to accept lower status by going along with someone else’s preference. So, to coordinate, we defer the choice to some “objective” external process, so that nobody’s social status is altered by it.
An example where this didn’t work: The Gregorian calendar took centuries to be adopted throughout Europe, despite being justified by “objective” astronomical data, because non-Catholic countries thought of it as a “papal imposition” whose acceptance would imply acceptance of the Pope’s authority over the whole Christian church. (Much better to stick with Julius Caesar’s calendar instead!)
This may shed some light onto why people have fun playing the Schelling game. It’s always amusing when I discover how uncannily others’ thoughts match my own, e.g. when I think to myself “X! No, X is too obscure, I should probably say the more common answer Y instead”, and then it turns out X is the majority answer after all.
Thanks everyone for coming! Feedback survey here: https://forms.gle/Nx4vqmXZnJ8EuuKP9
What exactly did you do with the candles? I’ve seen pictures and read posts mentioning the fact that candles are used at solstice events, but I’m having trouble imagining how it works without being logistically awkward. E.g.:
Where are the candles stored before they’re passed out to the audience?
At what point are the candles passed out? Do people get up from their seats, go get a candle, and then return to their seats, or do you pass around a basket full of candles?
When are the candles initially lit? Before or after they’re distributed?
When are the candles extinguished during the “darkening” phase? How does each person know when to extinguish their own candle?
Is there a point later when people can ditch their candles? Otherwise, it must be annoying to have to hold a lit candle throughout the whole “brightening” phase.
What happens to the candles at the end?
I wrote up the following a few weeks ago in a document I shared with our solstice group, which seems to independently parallel G Gordon Worley III’s points:
To- | morrow can be brighter than [1]
to- | day, although the night is cold [2]
the | stars may seem so very far
a- | way… [3]
But | courage, hope and reason burn,
in | every mind, each lesson learned, [4]
[5] | shining light to guide to our way,
[6] | make tomorrow brighter than [7]
to- | day....It’s weird that the comma isn’t here, but rather 1 beat later.
The unnecessary syncopation on “night is cold” is all but guaranteed to throw people off.
If this is supposed to rhyme with “today” from before, it falls flat because “today” is not really at the end of the line, despite the way it’s written.
A rhyme is set up here with “burn”/”learned,” but there is no analogous rhyme in the first stanza.
It really feels like there should be an unstressed pickup syllable here, based on the expectation set by all the previous measures.
Same here.
The stanza should really end here, but it goes on for another measure. (A 9-measure phrase? Who does that?)
To clarify some of these points:
1 & 3: There’s a mismatch between the poetic grouping of words and the rhythmical grouping, which is probably why bgaesop stumbles at that spot. This mismatch is made obvious by writing out the words according to the rhythmical grouping, as above.
2: The “official” version has “night is cold” on a downbeat with the rhythm “16th, 8th, quarter”, which is a very unusual rhythm. Notice that in the live recording here, the group attempts the syncopated rhythm the first time, but stumbles into “the stars may seem...”, and then reverts to the much more natural rhythm “8th, 8th, dotted-8th” in all subsequent iterations.
7: Mozart’s Musical Joke makes fun of bad compositions by starting off with a 7-measure phrase. Phrases are usually in powers or 2 or “nice” composite numbers like 6 or 12; a large prime number like 7 is silly because it can’t be imagined as having any internal regularity. You could maybe get away with 9 if it can be thought of as 3 3-measure subphrases, but this song doesn’t do that.
In my opinion, a good singalong song must have very low or zero tolerance for any irregularities in rhyme or rhythm. In LW jargon, if you think of the song as a stream of data which people are trying to predict in real time, you want them to quickly form an accurate, low-Kolmogorov-complexity model of the whole song based on just a small amount of input at the beginning.
(I’ve always hated singing “the bombs” in the Star-Spangled Banner!)
I think most non-experts still have only a vague understanding of what cryptocurrency actually is, and just mentally lump together all related enterprises into one big category—which is reinforced by the fact that people involved in one kind of business will tend to get involved in others as well. FTX is an exchange, Alameda is a fund, and FTT is a currency, and each of these things could theoretically exist apart from the others, but a layperson will point at all of them and say “FTX” in the same way as one might refer to a PlayStation console as “the Nintendo.”
Legally speaking this is nonsense, but when we’re talking about “social context,” a lack of clarity in the common understanding of what exactly these businesses do might provide an opening for self-deception on the part of the people running them, regarding what illegal activities are “socially acceptable” in their field.
Meta question: What do you think of this style of presenting information? Is it useful?
The more resources people in a community have, the easier it is for them to run events that are free for the participants. The tech community has plenty of money and therefore many tech events are free.
This applies to “top-down funded” events, like a networking thing held at some tech startup’s office, or a bunch of people having their travel expenses paid to attend a conference. There are different considerations with regard to ideological messages conveyed through such events (which I might get into in another post), but this is different from the central example of a “tech/finance/science bubble event” that I’m thinking of, which is “a bunch of people meeting in a cafe/bar/park”.
Or alternatively, do it the way the church does and have no entrance fee and ask for donations during the event.
I would indeed have found this less off-putting, though I’m not sure exactly why.
I highly recommend Val Plumwood’s essay Tasteless: towards a food-based approach to death for a “green-according-to-green” perspective.
Plumwood would turn the “deep atheism” framing on its head, by saying in effect “No, you (the rationalist) are the real theist”. The idea is that even if you’ve rejected Cartesian/Platonic dualism in metaphysics, you might still cling for historical reasons to a metaethical-dualist view that a “real monist” would reject, i.e. the dualism between the evaluator and the evaluated, or between the subject and object of moral values. Plumwood (I think) would say that even the “yin” (acceptance of nature) framing is missing the mark, because it still assumes a distinction between the one doing the accepting and the nature being accepted, positing that they simply happen to be aligned through some fortunate circumstance, rather than being one and the same thing.