MondSemmel

Karma: 3,533

MondSemmel 21 Dec 2024 1:35 UTC
2 points
0
in reply to: Zach Stein-Perlman’s comment on: Anthropic leadership conversation
Tiny editing issue: “[] everyone in the company can walk around and tell you []” → The parentheses are empty. Maybe these should be for italicized formatting?

MondSemmel 21 Dec 2024 0:36 UTC
5 points
1
on: Anthropic leadership conversation
Thanks for posting this. Editing feedback: I think the post would look quite a bit better if you used headings and LW quotes. This would generate a timestamped and linkable table of contents, and also more clearly distinguish the quotes from your commentary. Example:
Tom Brown at 20:00:
the US treats the Constitution as like the holy document—which I think is just a big thing that strengthens the US, like we don’t expect the US to go off the rails in part because just like every single person in the US is like The Constitution is a big deal, and if you tread on that, like, I’m mad. I think that the RSP, like, it holds that thing. It’s like the holy document for Anthropic. So it’s worth doing a lot of iterations getting it right.
<your commentary>

MondSemmel 21 Dec 2024 0:31 UTC
4 points
0
in reply to: Zach Stein-Perlman’s comment on: Anthropic leadership conversation
I did something similar when I made this transcript: leaving in verbal hedging particularly in the context of contentious statements etc., where omitting such verbal ticks can give a quite misleading impression.

MondSemmel 13 Dec 2024 10:36 UTC
2 points
0
in reply to: Roko’s comment on: The Dissolution of AI Safety
I think it would need to be closer to “interacting with the LLM cannot result in exceptionally bad outcomes in expectation”, rather than a focus on compliance of text output.

MondSemmel 12 Dec 2024 22:29 UTC
2 points
0
in reply to: Roko’s comment on: The Dissolution of AI Safety
Any argument which features a “by definition” has probably gone astray at an earlier point.
In this case, your by-definition-aligned LLM can still cause harm, so what’s the use of your definition of alignment? As one example among many, the part where the LLM “output[s] text that consistently” does something (whether it be “reflects human value judgements” or otherwise), is not something RLHF is actually capable of guaranteeing with any level of certainty, which is one of many conditions a LLM-based superintelligence would need to fulfill to be remotely safe to use.

MondSemmel 12 Dec 2024 22:16 UTC
5 points
2
in reply to: Sherrinford’s comment on: David Gross’s Shortform
How about “idle musings” or “sense of wonder”, rather than “curiosity”? I remember a time before I had instant access to google whenever I had a question. Back then, a thought of “I wonder why X” was not immediately followed by googling “why X”, but sometimes instead followed by thinking about X (incl. via “shower thoughts”), daydreaming about X, looking up X in a book, etc. It’s not exactly bad that we have search engines and LLMs nowadays, but for me it does feel like something was lost, too.

MondSemmel 12 Dec 2024 0:19 UTC
5 points
−1
on: MondSemmel’s Shortform
Media is bizarre. Here is an article drawing tenuous connections between the recent assassin of a healthcare CEO with rationalism and effective altruism, and here is one who does the same with rationalism and Scott Alexander. Why, tho?

MondSemmel 7 Dec 2024 11:26 UTC
40 points
8
in reply to: sapphire’s comment on: deluks917′s Shortform
Related, here is something Yudkowsky wrote three years ago:
I’m about ready to propose a group norm against having any subgroups or leaders who tell other people they should take psychedelics. Maybe they have individually motivated uses—though I get the impression that this is, at best, a high-variance bet with significantly negative expectation. But the track record of “rationalist-adjacent” subgroups that push the practice internally and would-be leaders who suggest to other people that they do them seems just way too bad.
I’m also about ready to propose a similar no-such-group policy on ‘woo’, tarot-reading, supernaturalism only oh no it’s not really supernaturalism I’m just doing tarot readings as a way to help myself think, etc. I still think it’s not our community business to try to socially prohibit things like that on an individual level by exiling individuals like that from parties, I don’t think we have or should have that kind of power over individual behaviors that neither pick pockets nor break legs. But I think that when there’s anything like a subgroup or a leader with those properties we need to be ready to say, “Yeah, that’s not a group in good standing with the rest of us, don’t go there.” This proposal is not mainly based on the advance theories by which you might suspect or guess that subgroups like that would end badly; it is motivated mainly by my sense of what the actual outcomes have been.
Since implicit subtext can also sometimes be bad for us in social situations, I should be explicit that concern about outcomes of psychedelic advocacy includes Michael Vassar, and concern on woo includes the alleged/reported events at Leverage.

MondSemmel 5 Dec 2024 18:57 UTC
3 points
0
in reply to: Alexander Gietelink Oldenziel’s comment on: Alexander Gietelink Oldenziel’s Shortform
I mean, here are two comments I wrote three weeks ago, in a shortform about Musk being able to take action against Altman via his newfound influence in government:
That might very well help, yes. However, two thoughts, neither at all well thought out: … Musk’s own track record on AI x-risk is not great. I guess he did endorse California’s SB 1047, so that’s better than OpenAI’s current position. But he helped found OpenAI, and recently founded another AI company. There’s a scenario where we just trade extinction risk from Altman’s OpenAI for extinction risk from Musk’s xAI.
And:
I’m sympathetic to Musk being genuinely worried about AI safety. My problem is that one of his first actions after learning about AI safety was to found OpenAI, and that hasn’t worked out very well. Not just due to Altman; even the “Open” part was a highly questionable goal. Hopefully Musk’s future actions in this area would have positive EV, but still.

MondSemmel 5 Dec 2024 18:53 UTC
14 points
10
in reply to: Alexander Gietelink Oldenziel’s comment on: Alexander Gietelink Oldenziel’s Shortform
all the focus on the minutia of OpenAI & Anthropic may very well end up misplaced.
This doesn’t follow. The fact that OpenAI and Anthropic are racing contributes to other people like Musk deciding to race, too. This development just means that there’s one more company to criticize.

MondSemmel 4 Dec 2024 13:06 UTC
5 points
0
in reply to: habryka’s comment on: (The) Lightcone is nothing without its people: LW + Lighthaven’s big fundraiser
Re: the history of LW, there’s a bunch more detail at the beginning of this podcast Habryka did in early 2023.

MondSemmel 3 Dec 2024 19:25 UTC
4 points
2
in reply to: habryka’s comment on: papetoast’s low quality shortforms
I could barely see that despite always using a zoom level of 150%. So I’m sometimes baffled at the default zoom levels of sites like LessWrong, wondering if everyone just has way better eyes than me. I can barely read anything at 100% zoom, and certainly not that tiny difference in the formulas!

MondSemmel 2 Dec 2024 12:22 UTC
2 points
0
in reply to: habryka’s comment on: (The) Lightcone is nothing without its people: LW + Lighthaven’s big fundraiser
I can’t find any off the top of my had, but I’m pretty sure the LW/Lightcone salary question has been asked and answered before, so it might help to link to past discussions?

MondSemmel 29 Nov 2024 20:22 UTC
2 points
1
in reply to: Sinclair Chen’s comment on: Sinclair Chen’s Shortform
Apologies if I gave the impression that “a selfish person should love all humans equally”; while I’m sympathetic to arguments from e.g. Parfit’s book Reasons and Persons^[1], I don’t go anywhere that far. I was making a weaker and (I think) uncontroversial claim, something closer to Adam Smith’s invisible hand: that aggregating over every individual’s selfish focus on close family ties, overall results in moral concerns becoming relatively more spread out, because the close circles of your close circle aren’t exactly identical to your own.
1. ^
  Like that distances in time and space are similar. So if you imagine people in the distant past having the choice for a better life at their current time, in exchange for there being no people in the far future, then you wish they’d care about more than just their own present time. A similar logic argues against applying a very high discount rate to your moral concern for beings that are very distant to you in e.g. space, close ties, etc.

MondSemmel 29 Nov 2024 11:10 UTC
2 points
0
in reply to: Sinclair Chen’s comment on: Sinclair Chen’s Shortform
Well, if there were no minds to care about things, what would it even mean that something should be terminally cared about?
Re: value falloff: sure, but if you start with your close circle, and then aggregate the preferences of that close circle (who has close circles of their own), and rinse and repeat, then this falloff for any individual becomes comparatively much less significant for society as a whole.

MondSemmel 28 Nov 2024 12:06 UTC
2 points
0
in reply to: ChristianKl’s comment on: Repeal the Jones Act of 1920
Maybe our disagreement is that I’m more skeptical about the legislature proactively suggesting any good legislation? My default assumption is that without leadership, hardly anything of value gets done. Like, it’s an obviously good idea to repeal the Jones Act, and yet it’s persisted for a hundred years.

MondSemmel 27 Nov 2024 23:29 UTC
3 points
0
in reply to: Raemon’s comment on: Repeal the Jones Act of 1920
This discussion felt fine to me, though I’m not sure to which extent anyone got convinced of anything, so it might have been fine but not necessarily worthwhile, or something. Anyway, I’m also in favor of people being able to disengage from conversations without that becoming a meta discussion of its own, so… *shrugs*.
I do agree that it’s easy to have discussions about politics become bad, even on LW.

MondSemmel 27 Nov 2024 23:21 UTC
2 points
0
in reply to: ChristianKl’s comment on: Repeal the Jones Act of 1920
I know that Trump doesn’t have control of the legislature or anything, but I guess I’m still not quite understanding how all this is supposed to relate to the Jones Act question. Do you think if (big if) Trump wanted the Jones Act repealed, it would not be possible to find a (potentially bipartisan) majority of votes for this in the House and the Senate? (Let’s leave the filibuster aside for a moment.) This is not like e.g. cutting entitlement programs; the interest groups defending the Jones Act are just not that powerful.

MondSemmel 27 Nov 2024 21:48 UTC
5 points
7
in reply to: sunwillrise’s comment on: Repeal the Jones Act of 1920
I don’t know, I’ve been reading a lot of Slow Boring and Nate Silver, and to me this just really doesn’t seem to remotely describe how the Trump coalition works. Beginning with the idea that there are powerful party elites whose opinion Trump has to care about, rather than the other way around.
Like, the fact that Trump moderated the entire party on abortion and entitlement cuts seems like pretty strong evidence against that idea, as well. Or, Trump’s recent demand that the US Senate should confirm his appointees via recess appointments, similarly really does not strike me as Trump caring about what party elites think.
My model is more like, both Trump and party elites care about what their base thinks, and Trump can mobilize the base better (but not perfectly) than the party elites can, so Trump has a stronger position in that power dynamic. And isn’t that how he won the 2016 primary in the first place? He ran as a populist, so of course party elites did not want him to win, since the whole point of populist candidates is that they’re less beholden to elites. But he won, so now those elites mostly have to acquiesce.
All that said, to get back to the Jones Act thing: if Trump somehow wanted it repealed, that would have to happen via an act of Congress, so at that point he would obviously need votes in the US House and Senate. But that could in principle (though not necessarily in practice) happen on a bipartisan vote, too.
EDIT: And re: the importance of party elites, that’s also kind of counter to the thesis that today’s US political parties are very weak. Slow Boring has a couple articles on this topic, like this one (not paywalled), all based on the book “The Hollow Parties”.

MondSemmel 27 Nov 2024 20:38 UTC
7 points
0
in reply to: ChristianKl’s comment on: Repeal the Jones Act of 1920
so, convincing Republicans under his watch of just replacing the Jones Act is hardly possible
Given how the Trump coalition seems to have worked so far, I don’t find this rejoinder plausible. Yes, Trump is not immune from his constituents. For example, he walked back from what some consider to be the greatest achievement of his presidency (i.e. Operation Warp Speed), because his base was or became increasingly opposed to vaccination.
But in many other regards he’s shown a strong ability to make his constituents follow him (we might call it “leadership”, even if we don’t like where he leads to), rather than the other way around. Like, his Supreme Court appointments overturned Roe v. Wade, but in this year’s presidential election he campaigned against a national abortion ban, because he figured such a ban would harm his election prospects. And IIRC he’s moderated his entire party on making (or at least campaigning on) cuts to entitlement programs, too, again because it’s bad politics.
This is not to say that Trump will advocate for repealing the Jones Act. But rather that if he doesn’t do it, it will be because he doesn’t want to, not because his constituents don’t. The Jones Act is just not politically important enough for a rebellion by his base.
A much bigger problem here would be that Trump seems to have very dubious instincts on foreign policy and positive-sum trade (e.g. he’s IIRC been advocating for tarifs for a long time), and might well interpret repealing the Jones Act as showing weakness towards foreign nations, or some such.

MondSemmel

Tom Brown at 20:00: