The cost driver for fabbing out wafers is the number of layers and the number of usable devices per wafer. Higher layer count increases cost and tends to decrease yield, and more robust designs with higher yields increase usable devices per wafer. If circles or other shapes could help with either of those, they would likely be used. Generally the end goal is to have the most usable devices per wafer, so they'll be packed as tightly as possible on the wafer so as to have the highest potential output.
> I assume they don't pattern the unused area, so the process should be quicker?
The primary driver of time and cost in the fabrication process is the number of layers for the wafers, not the surface area, since all wafers going through a given process are the same size. So you generally want to maximize the number of devices per wafer, because a large part of your costs will be calculated at the per-wafer level, not a per-device level.
Yes, but isn't a big driver of layer costs the cost of the machines to build those layers?
For patterning, a single iteration could be (example values, no actual values used, probably only ballpark accuracy) on a 300M$ EUV machine with 5-year write off cycle, patterns on average 180 full wafers /hour. Excluding energy usage and service time, each wafer that needs full patterning would cost ~38$. If each wafer only needed half the area patterned, the lithography machine might only spend half its usual time on such a wafer, and that could double the throughput of the EUV machine, halving the write-off based cost component of such a patterning step.
Given that each layer generally consists of multiple patterning steps, a 10-20% reduction in those steps could give a meaningful reduction in time spent in the machines whose time spend on the wafer depends on the used wafer area.
This of course doesn't help reduce time in polishing or etching (and other steps that happen with whole wafers at a time), so it won't be as straightforward as % reduction in wafer area usage == % reduction in cost, but I wouldn't be surprised if it was a meaningful percentage.
> Yes, but isn't a big driver of layer costs the cost of the machines to build those layers?
Let's say the time spent in lithography step is linear the way you're describing. Even with that, the deposition step beforehand is surface area independent and would be applied across the entire wafer, and takes just as long if not longer than the lithography.
Additionally, if you were going to build a fab ground up for some specific purpose, then you might optimize the fab for those specific devices as you lay out. But most of these companies are not doing that and are simply going through TSMC or a similar subcontractor. So you've got an additional question of how far TSMC will go to accommodate customers who only want to use half a wafer, and whether that's the kind of project they could profitably cater to.
Yes, but my understanding is that the wafer is exposed in multiple steps, so there would still be less exposure steps? Probably insignificant compared to all the rest though. (Etching, moving the wafer, etc.)
The number of exposure steps would be unrelated to the (surface area) size of die/device that you're making. In fact, in semiconductor manufacturing you're typically trying to maximize the number of devices per wafer because it costs the same to manufacture 1 device with 10 layers vs 100 devices with 10 layers on the same wafer. This goes so far as to have companies or business units share wafers for prototyping runs so as to minimize cost per device (by maximizing output per wafer).
Also, etching, moving, etc is all done on the entire wafer at the same time generally, via masks and baths. It's less of a pencil/stylus process, and more of a t-shirt silk-screening process.
> This goes so far as to have companies or business units share wafers for prototyping runs so as to minimize cost per device
Can this be done in production? Is there a chance that the portion of the wafer cerebras.ai can't fit their giant square in is being used for production of some other companies chips?
Utilitarian analysis inevitably is coopted by those in power to frame "good" as what the powerful want.
For example, some say it's more important to keep people's stock-invested retirement accounts solvent by increasing stock prices rather than take corporate profits and pay people who can't even afford groceries or rent, let alone invest in a retirement account.
Sometimes you'll have commercial/digital reps that handle high volume, low revenue customers. The company may also have a self-service/credit card swipe product option, or a free tier, or a support-only option.
I get what you're getting at, but LLMs aren't thinking machines. They literally just rearrange and regurgitate text that they've been trained on or have contextualized. How would you propose building a general purpose LLM that accomplishes what you're saying? How do we build a machine that is able to divine scientific truth from human outputs?
Well, probably by being much more selective about what we put in than just training on the most cheap and large corpus that is the internet.
This is not a technical limitation at all, this is purely about cost and time, and companies wanting to save on both.
There are also methods like RAG that try to give them access to fixed datasets rather than just the algorithmic representations of their training data.
Is it just me or was anybody else not sure what Obsidian was while reading this article?
The article doesn't explain it at all, so I'm left wondering if I'm incredibly out of the loop or whether this is a very niche thing that I haven't run into yet.
If I'd try to describe/categorize it, I'd call it a local, ?most often? single-user, scriptable and plugginable wiki-software.
I use it for taking notes, keeping a journal, TODO-list, and to bookmark/annotate stuff. Basically my own personal "knowledge base".
Technology-wise it saves notes as markdown, optionally with a yaml-style frontmatter, built on/with javascript, and exposes a bunch of APIs making it very extensible.
A strength/weakness with it is that it exposes the end-user to all this power, and does not enforce much in the way of ways of working, so you get to define for yourself how you go about using it.
I have mostly settled on workflows that works for me, but it also seems, if I am being completely honest, like there is always some little tweaking/refactoring going on.
It's kinda obvious from the context if you know that space, but yes, it's leaving out many explanations. It's not well written for random visitors.
Anyway, Obsidian is a knowledge-tool, using Markdown. It's for taking notes, collecting information and aims to support the user in building their network of knowledge. It's pretty powerful for a markdown-editor/manager, especially because it's supporting plugins, which is feeding a small community of developers bringing up all kind of ways to enhance Obsidian.
Obsidian is very popular and well known on this site, though attention for this space comes in waves, and I would say the big hype is over for the moment. So depending on what you read here, it's possible you just missed it.
this isn't a cool take (specifically words like "kinda obvious" and generally the sentiment which obligates the reader to have awareness of the subject) and we shouldn't assume readers here know everything or a single thing is "well known".
In short, it's a free note-taking app using the Markdown syntax, storing files locally and with wiki-like linking, tagging, a plugin system and tons of community plugins. Using it, you can organize knowledge, ideas, references, keep a journal, and more.
This seems like an overly simplistic take and reads like someone who hasn't worked closely with board members or been on the board of a company with a lot of money behind it.
First you had logs. Everyone uses logs because it's easy. Logs are great, but suddenly you're spending a crapton of time or money maintaining terabytes or petabytes of storage and ingest of logs. And even worse, in some cases for these logs, you don't actually care about 99% of the log line and simply want a single number, such as CPU utilization or the value of the shopping cart or latency.
So, someone says, "let's make something smaller and more portable than logs. We need to track numerical data over time more easily, so that we can see pretty charts of when these values are outside of where they should be." This ends up being metrics and a time-series database (TSDB), built to handle not arbitrary lines of text but instead meant to parse out metadata and append numerical data to existing time-series based on that metadata.
Between metrics and logs, you end up with a good idea of what's going on with your infrastructure, but logs are still too verbose to understand what's happening with your applications past a certain point. If you have an application crashing repeatedly, or if you've got applications running slowly, metrics and logs can't really help you there. So companies built out Application Performance Monitoring, meant to tap directly into the processes running on the box and spit out all sorts of interesting runtime metrics and events about not just the applications, but the specific methods and calls those applications are utilizing within their stack/code.
Initially, this works great if you're running these APM tools on a single box within monolithic stacks, but as the world moved toward Cloud Service Providers and containerized/ephemeral infrastructure, APM stopped being as effective. When a transaction starts to go through multiple machines and microservices, APM deployed on those boxes individually can't give you the context of how these disparate calls relate to a holistic transaction.
So someone says, "hey, what if we include transaction IDs in these service calls, so that we can post-hoc stitch together these individual transaction lines into a whole transaction, end-to-end?" Which is how you end up with the concept of spans and traces, taking what worked well with Application Performance Monitoring and generalizing that out into the modern microservices architectures that are more common today.
Then we might as well admit that "merit" is heavily influenced by starting conditions instead of pretending that everyone has "equal opportunity". How many times have I heard from people that the USA is about "equality of opportunity" and not "equality of outcome"? Legacy admissions is directly contrary to equality of opportunity, and undermines the idea of meritocracy in university admissions that people have been crowing about in anti-affirmative-action rhetoric.
A decade later, turns out I was wrong. Ironically, it’s because the racial identity politics of the last decade made me realize I’m not white, and that race is such a pernicious concept that even the good white people can’t be trusted with it.
That's fine , but you can't reasonably adopt a tone that there are obvious answers to this, or feign surprise that people disagree with you, when you yourself argued the exact opposite --- and when your reason for changing your mind is so idiosyncratic. I assure you that most people do not arrive at their opinion on racial preferences based on a realization that they aren't white. If that's the reason racial preferences are wrong, it's incumbent on you to disclose that up front.
You can think it's idiosyncratic, but I can assure you (both as an Asian and as someone who knows quite a few Asians) that many (maybe the majority of) Asians whose views on race-based treatment/preferences have changed were driven by the eventual recognition that they exist in the racial in-between. They get almost none of the "privilege" that white people do while getting none of the "protections" that other minorities do. And, at the same time, they are often penalized for being Asian.
It's not surprising or idiosyncratic that an Asian's viewpoint on race has evolved as they've spent more time being exposed to America's racial hypocrisy and experienced more and more of this in-betweenness.
I think we’re making a similar point, but I’d frame it slightly differently. For Asians, racialization is situational. Growing up, I rarely thought about race. Outside of dating and sports, physical differences never came up.
And now, Asians are racialized from both sides. And on the left, it’s this pernicious sort of racialization where we’re “people of color” when they’re trying to show how big their coalition is, but expected to “check our privilege” like white people and defer to the interests of black people and Latinos.
I agree we are making a similar point. My original post had a long rant about Asians only being minorities when it benefits someone's argument (on the left or right) but I deleted it before posting. Suffice to say that there is a long history of marginalizing racism against Asians in America (recall that the Chinese Exclusion Act or some form of it was in effect until the World War 2) and the fight that the left has put up to try to excuse or hide the clear facts that affirmative action policies have (had?) an adverse impact on Asians is the latest example of that.
My reason isn’t idiosyncratic at all: any system of racial preferences necessarily racializes people. Being racialized is unpleasant. I don’t like it for the same reason I imagine the majority of non-white people oppose racial preferences. They don’t want to be treated like that, don’t think it’s fair, etc.
Growing up as someone who wasn’t racialized and then becoming racialized as an adult is something most people don’t experience. That is idiosyncratic, sure, but that’s the reason my view on racial preferences changed, not the reason I oppose them now.
I'm not saying that you're making a risible argument, though I disagree with it. I'm simply saying that you're making a non-obvious argument, and the evidence that I offer for my claim is that you yourself didn't believe it just a few years ago, despite being one of the "token conservatives" in your law school class. It couldn't have too much to do with how you grew up, because you came by it lately.
Make whichever argument you like. I'll only object to your use of the rhetorical frame that implies surprise that someone else would disagree with you. You disagreed with you until recently, as I do now.
The argument I’m making about why a system of racial preferences is bad is the standard conservative argument against racial preferences. It reinforces the racialization society and perpetuates it to the next generation. “The only way to stop discriminating on the basis of race is to stop discriminating on the basis of race.”
But the question you asked me above was different: why my view changed. The reason for that may be idiosyncratic and non-obvious—I didn’t appreciate color blindness as a social norm until it wasn’t my reality anymore—but that’s a different issue than the first point.
It’s like if someone is a pro-taxes and regulation Democrat and then changes his view because he inherits his dad’s small business when he passes unexpectedly. The reason for the change in view might be idiosyncratic, but the reason for opposing taxes and regulation is standard.
Yes. Once again: I'm not saying that your argument is invalid because you've changed your mind. I'm saying that because you've deeply held both positions on the issue, it is disingenuous to pretend that one of those positions is obviously false. Perhaps you know something new, as the metaphorical inheritor of the family business. It's incumbent on you to share that, rather than posturing as if you knew it all along.
(Again, I agree with Rayiner Classic on this point, and not with New Rayiner, but that's neither here nor there.)
Aside from my thoughts on discrimination in general, which I'm willing to concede isn't a strong argument here, I think describing affirmative action as necessary is naive. Ensuring that disadvantaged groups grow up with quality schooling, healthcare, and whatnot should be the top priority. Creating/improving colleges (because Ivy Leagues aren't the end all be all, even if the social networking is better), reducing tuitions, perhaps providing money on the basis of poverty (which is, I daresay, objective more correct than the basis of race) in the communities and for college are secondary measures. Affirmative action is comparatively simple, convenient, and utterly propagandized policy. If money is the bottleneck, I daresay it's more due to the lack of political will to distribute money cough massive wealth disparity cough.
How would the current thoughts of someone commenting about the same topic a few years ago be particularly relevant to our conversation today?
As far as I see, the argumentation isn't particularly novel or unusual either. It's a pragmatist take, discounting methodological concerns in favour of expected outcomes.
I think that take probably summarizes a quarter of the current debate.
Remember, graduates of these institutions go on to run everything else, and take these ideas with them. Including bizarre conventions such as the people of color hierarchy, where Kenyans and Ghanaians are higher on the totem poll than equally poor Bangladeshis or Indians. All of us “people of color” have to live within this paradigm.
What pushed me over the edge was facing these policies from the receiving end. I now routinely have to explain my “diversity” to white decision makers in a business context. And my kids—I’ve had three since I wrote that post—are being encouraged to identify as “people of color.” So they can spend the rest of their lives writing diversity essays and trying to persuade white people to pick them.
College racial preferences aren’t the only problem here, but I’m convinced that—given how influential higher education is in American society—they’re the root of the problem.
> It has to do with correcting specific and massive historical wrongs in which our government was complicit, not lifting up disadvantaged people generally.
And I tuned out. I’m not paying for the sins of our fathers.
Most Americans came here after slavery ended, so no they’re not. If you can trace the “proceeds” of slavery anywhere, it’s not the pocket of some Polish guy whose parents got here in 1920, or some German American from the Midwest or some Appalachian. It’s to places like Harvard and the banks, insurance companies, etc., that Harvard graduates end up working at.
The Chicago suburb that I live in right now was debating a quota system to prevent Black families from moving here within my own lifetime, and was until the late 1990s spending tax dollars to subsidize apartments being held vacant for white families. The harm these programs attempt to correct wasn't in the distant past.
I can answer this, but first I want to understand better why it matters. It's deliberate, de jure harm inflicted on Black families regardless of the purported and actual benefits to white homeowners.
Because the original statement was that white people are gaining advantage because of the sins of our fathers. Most of us remember our lives and know this is not true, unless you think growing up poor with parents that couldn’t even afford school for their kids is an advantage.
This is nonsense, in any moral framework worth its salt.
Consider a simple situation:
1. Lying is wrong.
2. Someone's running from a mob that wants to kill them. They went right.
3. The mob stops, and asks you if the person in question went right.
4. Two wrongs don't make a right, so you tell the truth. Or don't say anything, and let the mob go off in the correct direction and chase that person down.
Your example is backwards. Lying is the right, and truth is the wrong in your example.
So in arguing for two wrongs are OK, you are suggesting you'd direct the mob to the person running away (maybe they were a person you didn't like, or were of privilege you resent).
> Lying is the right, and truth is the wrong in your example.
Lying is wrong! Except, according to you, when it leads to good outcomes!
It sounds like outcome-driven morality is what you're pushing for..? Then what's the problem with AA? Not using it to compensate for structural disadvantages is being in the wrong in its case...
"Lying is wrong" is a non-sequitur - like West is to the left. It is morality for 5 year olds or Sam Harris.
>Then what's the problem with AA? Not using it to compensate for structural disadvantages is being in the wrong in its case...
AA fails on that criteria as well. It isn't compensating the people who were wronged and the burden falls on people didn't do the wrong. Poor asian immigrant gets kicked out so Harvard can virtue signal and put a black face on their web page and course catalog. Never mind that kid is a wealthy immigrant from Kenya.
And to make it even worse, the AA admits do worse, drop out at higher rates, and drop down to lesser majors because many aren't academically competitive. They would have done better if they were matched on merit to schools.
On the contrary, when California enacted a ban on AA in university admissions in 1996, enrollment of minorities plummeted. Sure, maybe some of them would have dropped out, and some may have changed majors, but at least they would have had the opportunity, and certainly some would have been able to take advantage of it.
> Berkeley and UCLA, the most selective public universities in the state, saw a declines in Black and Hispanic enrollment of about 40% immediately the year after the Proposition was implemented. There was no net change in black and Hispanic enrollment at less selective California universities because while some black and Hispanic students lost access to those schools because affirmative action ended, they also gained Black and Hispanic students from more selective schools like Berkeley and UCLA that those students could no longer get into.
> seems like these very selective public universities in California just provided greater value to relatively disadvantaged Black and Hispanic students who came from lower-income neighborhoods, had poorer job networks, relatively less access to otherwise successful peers, and who were thus able to better take advantage of the resources provided by these super selective universities than the white and Asian students who took their places.
Certainly, specific demographics can vary by school. But the OP said that enrollment of MINORITIES plummeted in the California university SYSTEM. Asians are minorities. There are more schools than UCLA and Cal.
Moreover, we can examine the data directly at Berkeley and realize that the only demographic that has “plummeted” long term is whites. Immediate enrollment percentages may have dropped for black and Latino students but it has recovered and it seems like Berkeley has been able to somehow return to their previous highly consistent proportion of Latino and black students. UCLA is the same.
It’s almost uncanny how the racial percentages at top schools stay so consistent over the years.
Do you have data to back up the implicit claim you're making that dropping affirmative action will, all else equal, result in a larger number of lower-income people attending schools, and that affirmative action policies weren't aiding non-immigrant Blacks?
Fine. You think lying isn't always wrong. I don't care, it's not important, it's just an example.
Pick literally any other action that you think is always wrong. I'll give you a counterexample where failing to do it is wrong.
"Two wrongs don't make a right" is the 5-year old-level morality platitude that you should be turning your ire against. Because, as you're demonstrating, "Wrong" is entirely contextual. A 'wrong' that mitigates another wrong can make a right.
> How many times have I heard from people that the USA is about "equality of opportunity" and not "equality of outcome"? Legacy admissions is directly contrary to equality of opportunity, and undermines the idea of meritocracy in university admissions...
Understanding "equality of opportunity" to be literal and absolute is nonsense, because to do so would require hobbling people with natural talent (for instance), since not all people have the opportunities created by those talents (there's a famous sci-fi short story about that https://en.wikipedia.org/wiki/Harrison_Bergeron).
IIRC, rejecting "equality of outcome" in favor of "equality of opportunity", means rejecting explicit policies to pick winners and losers.
I think you have it backwards: targeting "equality of outcome" necessitates policies to pick winners and losers. Specifically, it requires policies that aim to minimize both.
Equality of opportunity doesn't mean hobbling talented people or enforcing that everyone be given identical opportunities. It simply means that people regardless of talent (in a given ability) ought not be restricted from an attempting to take an advantage of an opportunity should they happen to have it.
All admissions policies pick winners and losers, that’s the point. A collage which sifts the relative importance of SAT scores vs GPA is picking winners and losers, it’s hardly an argument in favor of any specific policy.
AA increases how much innate talent matters vs some peoples social position. It’s the same outcome as removing legacy admissions, just with a slightly different set of beneficiaries.
Collages have many leavers to get these kinds of outcomes. Favor elite prep schools and suddenly you have a surplus of high income families etc.
I don't think you do remember correctly, and equating the ideal of equality of opportunity to Harrison Bergeron is silly. Feel free to peruse any one of this to get an idea of the common usage the term.
I think the debate regarding affirmative action is very simple and not unexpected at all. Here's how I view it.
To start, in America I believe that most of us believe that the "default" behavior should be to avoid unfair discrimination, especially for protected classes. I think most people would agree to at least this, it's a pretty generic and obvious statement.
Therefore, when we deviate from this for some reason, generally, it REQUIRES a healthy amount of thought: the baseline should be at least a strong hypothesis to begin the conversation. The world is very complicated, so simply assuming something does what you expect it to because it intuitively sounds like it does is generally not a reasonable position.
And of course, the idea behind affirmative action, hopefully put into words that people feel is fair, is a sort of discrimination, but the intention is of course to try to adjust for past disgressions and injustice to try to "re-balance" opportunity. So unlike the four-letter-word that was discrimination in historical contexts, it is not based on racism[1], for example.
So does affirmative action work? It seems to do roughly what it is supposed to do, although honestly a huge problem is that it's sort of tautological. Of course it works, at doing what it's meant to do. Some have argued that it could potentially harm students by leading to a "mismatch", but the evidence is mixed and in any case it probably causes more good than harm in terms of outcomes. I am not an expert on this though, and I have not been into the studies for a while.
The real question that I think causes so much strife and pain is the one that hurts to try to answer: is it worth it? And that is not easy to answer, nor does it have an obvious objective answer. I truly believe that most of this argument boils down to proxies for this particular question. Some people who have a particular egalitarian bend to their views on life and society might blanket oppose such a policy on an ideological basis, whereas someone who is strongly anti-racist is highly likely to prefer such policies even at high cost.
Cost? By that, in this case I mean in terms of going against the basic belief of not discriminating. Ideology is important to people even when there isn't a discrete cost, but in this case the micro and macro views are very different. On the micro level, someone who is less qualified will be preferred over someone who is more qualified, on the basis of factors outside of their control. On the macro level, population demographics change, generally reducing biases.
There's a lot of finer points. Like clearly, on the micro level, when someone "less qualified" according to some criteria passes due to affirmative action, the idea is that it was beyond their control in the first place that they were less qualified, which may very well be true. And on the macro level, statistics may not tell the full story: demographics are a measurement of people, and people are not fungible. The numbers surely look better on paper, but one must wonder sometimes if it's actually doing what it looks like it's doing.
You might think that I am staunchly opposed to affirmative action based on my framing of this, and the truth is, I simply don't know. I think that it's potentially very powerful, but it also is damn scary to wield institutional discrimination even if it's supposed to be a force for good. This isn't exactly a slippery slope situation, of course, but the road to hell is paved with good intentions. I've personally flip-flopped probably a lot of times. All I can say is that I sort of hope people don't just assume this is the right way to solve all of the problem of injustices, or maybe even more importantly, that merely instituting policies like this doesn't "solve" America's history with racism and sexism; and I don't think most people believe that it does. For some of those things, I think only a lot of time will truly be able to heal most of that, and it's going to leave a pretty nasty scar.
Of course, beyond the fairly straightforward debate is the culture war bullshit surrounding it, but to me it's mostly noise. I look forward to a future with less influence from Twitter and news organizations so that people can go back to discussing things at least slightly more like human beings.
[1]: Using racism in this context to refer to the fairly strict definition of being related to beliefs about races rather than about discrimination.
What baffles me is that so many people argue solely about affirmative action, as if it's the only possible maybe-solution. I wrote more details here[0], but I think affirmative action can't be effective because it's so narrow, and the alternative should be lifting up poorer communities to an acceptable baseline. A strong foundation provides an opportunity for a strong house, after all. This would have an outsized impact on disadvantaged minorities but is more generally just correct: it doesn't violate equality of opportunity and it actually helps out minorities.
To go on a bit of a rant unrelated to your comment, but I think what I'm proposing will be very effective, assuming that the focus on education is held in poorer communities. Culture is a touchy subject. Anyways, education has been shown to be an empowering force, and what better way to eradicate racism to the fringes of society than to bring everyone to the same level? Intentional or not, the feasibility of racism would become less as the prevalence of undeniably qualified minorities grows. Affirmative action gets the goal right but not the method. Even if what I'm proposing is somewhat incomplete, it would be a lot better as a starting point of discussion. Surely there must be other people who have thought similarly.
>Using racism in this context to refer to the fairly strict definition of being related to beliefs about races rather than about discrimination.
A fairly meaningless distinction when beliefs about races drives the discrimination.
Affirmative action is just racist, it just is, it’s grounded on the idea that a racial group which outperforming another is in of itself proof of that group advantaging itself via racism and justifying racial discrimination to countervail this. People bought that for Whites, the same logic being applied to Asians broke the system because people find it appallingly racist and divorced from reality.
