finexplained

I work on very interesting technical problems with very smart colleagues for excellent pay. All my career progression is in compensation, so I can remain an IC forever and no one thinks that's a negative. I'm subject to no politics whatsoever, and there's very little politics in the company as a whole. The work I do every day has a direct material impact on the company and I'm rewarded proportionally to my impact. My WLB is so-so, but it's better than it was in grad school so I'll take it.

Regarding social impact, the world does have some demand for liquidity and price discovery. Providing those services is both essential and extremely difficult. It's definitely not the most social good I could be doing with my talents, but I think it's weakly positive.

I work at an HFT firm. Most fun I've ever had.

I thought the H100s were ~30-40k each. But they're not widely available and you usually buy multiple boxes from vendors that also come with expensive CPUs/RAM etc.

I'm sorry, I know we're not suppose to engage in ad hominem attacks. I'll refrain in the future.

The deltas between the military/technology of the worlds top economies might not even be noteworthy to a species capable of interstellar travel. From there perspective we might just all look the same.

Empirical studies can be surprising.

The five paragraph format will forever be suspicious.

tldr: LoRA fine-tuning of an open-source alpaca model.

See https://arxiv.org/abs/2305.15717 for why calling this a private ChatGPT is probably misleading.

> And are most of the people she‘s well-connected to not annoyed that she lost their money?

Ya admittedly one would have to assess that and also make some estimate about how much her notoriety will positively or negatively affect her offsprings' outcomes. I tend to lean toward it being a net positive, since fame seems valuable almost regardless of how it's acquired these days.

> And isn‘t attractive just stating the same as good genetics?

It also includes height, medical history, longevity, intelligence, ambition/drive, work ethic. I admittedly don't know much about medical history, but I might be willing to gamble.

> And is she really that smart,

No she's not really that smart, I'd estimate 1-2 stddevs from the mean. In my opinion that's smart enough for a potential spouse with other positive attributes, particularly if you think that you're smarter still.

I mean, she's attractive, smart, pretty good genetics, clearly well connected. I don't find it that surprising.

Elon Musk is a prominent member of the tech industry and of Silicon Valley history, dating back to the days of Zip2 and Paypal. It's just a fact.

The distribution is different. The median programmer provides much more value than the median artist. Additionally math/cs skills apply much more broadly than art skills. The entertainment industry would be much less leveraged without the tools we built for them.

Sure, but we actually provide business value and our toolset is applicable to a broad set of problems.

You decide you want to optimize for number of lives saved. You decide that future lives, those of people yet to be born, are worth as much as those currently alive. You place small, but importantly not zero, probability on existential risks to humanity, so that when you do the expected value calculation, even an infinitesimally small risk of humanity's complete extinction results in negative infinite utility. You're also very smart and realize that smart people can do damage if their objectives are misaligned, and you start to worry about something much much smarter than you with objectives misaligned to humanity's more broadly. In their defense reward specification is indeed a hard problem, RL agents find unexpected policies that maximize reward in even toy settings. At this point you're down the rabbit hole and no other problem seems to compare. Climate change will leave some people alive, pandemics leave some people alive, AIs have no such kindness.

Um, because I think words have meaning, and I think accusing people of actions that have legal implications when they are not in fact doing those things is bad.

This allegation of frontrunning references another trading participant, not the miners. Also, arguably, and certainly not legally.

I respectfully don't think it is too harsh. Every legal definition I know of frontrunning requires a client relationship. Even in the case of "non-public" information, frontrunning refers to trading ahead of your client based on non-public information related to the clients securities. Trading on non-public information when a client is not involved is referred to as insider trading.

Also in this context, the reason that these sandwich attacks can happen is because all the information is public. The adversary can see your transaction waiting to be validated, and pay the miner a higher gas fee to be executed first.

Yep, that one is my bad. That's why I added the edit.

Couple things.

(1) The word "frontrunning" applies only in a very specific context, where you trade in front of an entity with which you have a client relationship. That is different from capturing a trading opportunity because you are faster than another party. Frontrunning is illegal and law-abiding HFT firms don't do it. They often don't even have clients to which this applies! Being faster is not illegal. Can we please stop using this term to describe behavior we don't like.

(2) What do they mean they introduce the term MEV (miner extractable value)? Everyone calls it that, and it's part of the incentive that miners have to validate transactions.

(3) Might want to add (2019) to the title, this paper is pretty old. Perhaps this answers (2).

In the first case, they complain, and there's ~750 of them (in the course I TAed) so even a small number can take up a lot of time. The right way to think about it is for a small additional bit of time spent clarifying instructions you save yourself a larger amount of time later.

In the second case, it does depend upon how much the instructor feels it's their duty to uphold the integrity of the grades in their class. I'm not sure if I would have made the same choice in my advisor's shoes, but that is the decision he made.

This perfectly describes my experience as a TA in graduate school. At first I didn't understand why my advisor insisted on being so precise in assignment instructions. Then when TAing with him I saw how students could creatively misinterpret instructions, even when I could not imagine how to make them more precise. An exception for the new case would be added to the next iteration of the assignments. I only understood why we went to such lengths to prevent cheating because in my first year I watched my advisor spend two weeks of his time sitting down individually with each student and present evidence that they had cheated. Only about 10% of the students had cheated, but in a class of 1400, that's 140 students! I can't even imagine how much work that must of been on the head TA.

It's a controlled substance, you need a prescription.

I've never understood why the measure that universities seem to care about is the number of students they reject instead of the number of students above their bar that they are able to educate.

Harvard could easily educate 1-2 orders of magnitude more students than they currently do with their resources. They explicitly choose not to. There are far more students capable of handling and benefiting from a Harvard-level education than currently have access to it.

And for those who are going to respond something akin to that's not the objective Harvard is optimizing, I'm well aware. I'm just lamenting the way things currently are.

I really don't understand why this wasn't done years ago.

> is concerned that the emergency declaration is being used to support her plan for more policing in the neighborhood, which could criminalize people with substance use disorders before resources such as the supervised drug-use site or a planned drug sobering center open.

Honest question. Is a supervised drug-use site an effective approach for reducing drug use long term? Does it help get people into services that help them deal with addiction? Or is it just meant to supervise potential overdoes and intervene?

It's just an instagram for a young (attractive) woman who likes taking nature travel photos. If you've seen one influencer instagram you get the gist.

> If that was the case then FAANG wouldn't be filled with Asians and engineers from ex-Soviet countries. The makeup of high paying tech companies does not represent in any way the makeup of the American population.

I don't think you can draw a cultural difference from this. I get what you're trying to say, particularly by narrowly defining "American"; that cultural influences put pressure/status on academics both by immigrant parents and non-American parents and that pressure causes students to excel academically and achieve these high paying positions. But there is so much selection bias here; both in the types of people who immigrate to a new country and those who leave their home country to attend school abroad. It's not clear to me Asian culture results in FAANG being disproportionally Asian, or simply that there are 2 billion Asian people in the world, and while the same percentage of people are sufficiently qualified as top American students (let's say 2%), 2% of 2 billion people is 40M people! Do you see the problem with jumping to this conclusion? You don't have sufficient evidence to reject the null hypothesis, that both cultures produce tops students at similar percentages.

A personal anecdote: I spend a lot of my interviewing candidates. Students for all the top schools in the world, international students and domestic students, undergrads and PhDs. Conditioned on having gone to a top university, I've never noticed a difference in performances between domestic students and international ones. (Or undergrad vs PhD to be honest.)

I also take a bit of issue with your definition of "American" on a personal level.

> To be clear, when I say American I mean someone whose ancestors have lived in America for multiple generations, since at least the 1950's.

My parents are immigrants, I'm a first generation American. But I am absolutely culturally American. There is no other country were I would feel at home culturally, and I've been heavily influenced by the country I grew up in. I'm also the product of US public schools, most notably UC Berkeley, which you cite. And I'm not Asian or Russian. And I do work at a FAANG level firm.

I don't really think the connection you're trying to draw follows immediately. There are very strong domestic students, probably enough to fill all the top PhD programs. The calculus for pursuing a PhD is different if you're a domestic vs international student. If you're international and want to move to and eventually live in the US, where some of the highest paying jobs in the world are available, then first getting into a well regarded graduate program, which will attract potential employers, coming in on a student visa and then getting an employer to sponsor an H1B is a pretty good strategy. If you're already a top student in the US at a top university, you already have access to those employers and intend to stay in the same country long term, so the opportunity cost of going to graduate school is so much higher.

You can't; the sentence is an intentional simplification for the purposes of a joke. A torus has two non-trivial elements in its 1-dimensional homology group: there are two types of loops that I can't contract to a point on a torus, around the handle and orthogonal to the handle. A lion with 1 entrance and 2 exits has at least 3 such loops one around each entrance, and 3 more (I think) passing through each of the 3 choose 2 aforementioned holes.

I'm going to go against the grain of the comments section here. As someone who likes both AI and 3D graphics, this is awful. It's graphically pleasing, sure, though a bit sluggish, but why? It's such an incredibly inefficient way to convey your qualifications/experience to another individual. Do you know how long it took me to learn that you work on GANS? PDFs aren't broken, don't fix them.

And how did your collection find its way into those discords... You outsourced the marketing to your followers, that's what social capital is. If you were a no-name individual with 20 followers, your NFTs probably wouldn't have sold despite being exactly the same, because the probability of them ending up in front of the right people would have been much lower, unless you went to considerably higher marketing effort. Thus the point of the parent comment stands.