in a nutshell (I don't have the time to write a treatise on this, every word I write is 20w i could have read instead): I'm pretty confident (status: ~ .6) that if SVB had 90bi in gov bonds HTM instead of MBS and the like, it wouldn't have failed. You can say it would have had losses (especially in 2022), it would likely have been bought, but not failed. I know of no bank that has suffered a run because they had too much gov long term bonds (from the corresponding government, and unless the gov defaulted, ofc) in the last century (if you have an example, please enlighten me); not only there's a very liquid market for them, but, in the last century (perhaps since Badgehot) central banks will let banks convert them into money easily - because tax payers will suffer no losses. On the other hand, MBS (and other similar derivatives) may be linked to credit risks (even Agency MBS) and it's quite unsure how your liquidity line will work, and the market is not so liquid; and that's why they offer higher yields - which is why they dominated SVB's HTM. Thanks for the memory refreshing lecture on the crisis of 2008. But I still remember almost everything
Because I work for a regulator and am not allowed to do that? Also, many investors won't have enough incentives to read beyond what other investors are reading... except if, as you mentioned, u work w shortselling And shortsellers did make money in this case. So in this sense, the system works... but when it happens to a bank, that's not so cool
and the whole world will care a lot more about this term in bank reports, basically forever.
I'm usually astonished w how seldom investors and supervisors read the fine print in annual reports. I don't think "this time will be different". Unless GPT-like automated report-readers step in (or maybe precisely because humans will leave this boring details to machines), we'll see it happen again.
Btw, I just noticed that $9.3bi of these $15.2 are MBS - yeah, the same type of security associated w the great crisis of 2007-2008. And the HTM total more than U$90 bi, $72bi of which are MBS and CMBS - so dwarfing their investments in bonds, and their $17.6bi in AFS.
From the post above:
...something called Agency MBS (which you can just read as "government debt with extra steps")
I'm no expert in US markets, but I don't think that's true. For instance, if you try to get a repo w them, you'll probably need a larger hair-cut than w gov bonds.
if people had learned to read bank reports, I'd expect to read more comments on this, instead of the last three pieces I read that basically just said SVB had too much gov bonds.
EDIT: after googling "svb mbs htm," I found tons of surces commenting on this. So, my bad. And most of all, thanks for this post & for this comment, rossry. I believe you saved me at least 1h of googling - to have a better grasp of the situation.
Well, I'm kinda sure climbers wouldn't like to brain-upload to a utopia simulation, so maybe there's some connection between this and AI alignment.
(Curiously, just read today Will MacAskill's WWOTF mentioning he used to climb buildings in Glasgow in his teens, until he was almost cut open by glass...)
Thanks, I believe you are right. I really regret how much time and resources are wasted arguing over the extension / reference of a word.
I'd like to remark though that I was just trying to explain what I see as problematic in FiO. I wouldn't only say that its conclusion is suboptimal (and I believe it is bad, and many people would agree); I also think that, given what Celestia can do, they got lucky (though lucky is not the adequate word when it comes to narratives) it didn't end up in worse ways.
As I point out in a reply to shminux, I think it's hard to see how an AI can maximize B's preferences in an aligned way if B's preferences and beliefs are inconsistent (temporally or modally). If B actually regards sim-B as another self, then its sacrifice is required; I believe that people who bite this bullet will tend to agree that FiO ends in a good way, even though they dislike "the process".
Possibly. I said this AGI is “safer and more aligned”, implying that it is a matter of degree – while I think most people would regard these properties as discrete: either you are aligned or unaligned. But then I can just replace it with “more likely to be regarded as safe, friendly and aligned”, and the argument remains the same. Moreover, my standard of comparison was Celest-IA, who convinces people to do brain uploading by creating a “race to the bottom” scenario (i.e., as more and more people go to the simulation, human extinction becomes more likely – until there’s nobody to object turning the Solar System into computronium), and adapts their simulated minds so they enjoy being ponnies; my AGI is way "weaker" than that.
I still think it’s not inappropriate to call my AGI “Friendly”, since its goals are defined by a consistent social welfare function; and it’s at least tempting to call it “safe”, as it is law-abiding and does obey explicit commands. Plus, it is strictly maximizing the utility of the agents it is interacting with according to their own utility function, inferred from their brain simulation - i.e., it doesn’t even require general interpersonal comparison of utility. I admit I did add a tip of a perverse sense of humor (e.g., the story of the neighbors), but that's pretty much irrelevant for the overall argument.
But I guess arguing over semantics is beyond the point, right? I was targeting people who think one can “solve alignment” without “solving value”. Thus, I concede that, after reading the story, you and I can agree that the AGI is not aligned - and so could B, in hindsight; but it’s not clear to me how this AGI could have been aligned in the first place. I believe the interesting discussion to have here is why it ends up displaying unaligned behaviour.
I suspect the problem is that B has (temporally and modally) inconsistent preferences, such that, after the brain upload, the AI can consistently disregard the desire of original-B-in-the-present (even though it still obeys original-B’s explicit commands), because they conflict with simulated-B’s preferences (which weigh more, since sim-B can produce more utility with less resources) and past-B’s preferences (who freely opted for brain upload). As I mentioned above, one way to deflect my critique is to bite the bullet: like a friend of mine replied, one can just say that they would not want to survive in the real world after a brain upload – they can consistently say that it’d be a waste of resources. Another way to avoid the specific scenario in my story would be by avoiding brain simulation, or by not regarding simulations as equivalent to oneself, or, finally, by somehow becoming robust against evidential blackmail.
I don’t think that is an original point, and I now see I was sort of inspired by things I read from debates on coherent extrapolated volition long ago. But I think people still underestimate the idea that value is a hard problem: no one has a complete and consistent system of preferences and beliefs (except the Stoic Sages, who “are more rare than the Phoenix”), and it’s hard to see how we could extrapolate from the way we usually cope with that (e.g., through social norms and satisficing behavior) to AI alignment - as superintelligences can do way worse than Dutch books.
My concern, which I interpret as being TAG's point (but with different words), is that your example of water vs. XYZ is immediately traceable (at least for anyone who knows the philosophical discussion) to Putnam's Twin Earth experiment. The way you express your point suggests you disregard this thought experiment - which is surprising for someone acquainted with it, because Putnam (at least when he wrote the paper) would likely agree that a substance with the same basic chemical properties of water would be water. He actually aims to provide an argument for semantic externalism - i.e., the idea that the meaning of "water" (or of other natural species) is H20, its chemical nature, and not the apparent properties commonly used as criteria to discriminate it (that it's a tasteless liquid...). He's so pushing against a conventionalist view about semantics (and philosophy of language), thus it's not about physics or ontology.
LW is quoted (in a kind of flattering way) in Simon Dedeo's awesome piece on the last Nautil.us issue. For spoiler lovers:
[...] Rationality is my ticket out. The only reason I can trust you is that you seem rational enough to talk to. But now you’re telling me that rationality is just a layer on top of the System—it’s just as irrational as the people I’m trying to escape. I don’t know which is worse: being duped by someone else’s priors, or being a biological machine.
Teacher: Don’t go too far. You’re a smart kid—you can iterate faster than most. You can match patterns better. Evolution set you up well. You’ll get better at predicting the consequences of your actions, and better at adapting your environment to your will. Rationality is systematized winning.
Ian: It’s not winning I’m worried about. It’s my mind. Maybe it’s silly, maybe it’s a fetish, but I want to know the truth. It’s the principle of the thing. Wanting to know the truth got me this far, but now the only option you’ve given me is believing in something I can’t see. If I know it at all, it can’t be through rational, scientific calculation. There’s some kind of extra-rational process I have to engage in—but what’s beyond the edge of reason?
Teacher: Many things. Dreams, intuition, transcendence, love, ascending the ladder, repetition and the leap of faith, philosophy itself ...
Ian: ... delusion, fairy tales, fascism!
Teacher: Childhood’s end.
Strict liability in tort law seems to be a pretty obvious example, doesn't it? I mean, I guess a lot of corporate law can be seen as "vicariously holding a group accountable"
It turns that the truth is more bizarre. From Matt Levine's Money Stuff:
The point I stressed before on government bonds was right: SVB could have borrowed against them. But it seems like I was wrong: it could have borrowed against Agency MBS, too.