I’ve often wondered about die thinning. I mean just looking at sd cards and micro sd etc, it was clear that there had to be a thinning step (or many steps) and it seemed amazing to me that such a complex step could be economically viable. And yet it clearly is. All a matter of being able to operate at scale I guess.
It’s not a step I’ve seen discussed very much, so it was kind of neat to see it discussed front and center in a lay article like this.
Not dissimilar to lens or mirror making, the process of grinding to a precise flatness. The key step here is bonding it to a glass carrier, presumably because otherwise it gets too thin to support itself.
See also the "reduced kerf diamond wire saw": technology critical to cheap solar panels is the ability to saw the boule into thinner and thinner pieces with less waste at each cut, like a big ham slicer.
Yeah, the scale is kind of mind-bending. You're taking something that's already extremely fragile, grinding most of its thickness away, handling it through several more process steps, and somehow doing that cheaply enough for mass-market products.
The critical back end step for HBM is thermal compression bonding - where they actually compress 13 very thin die separated by solder balls. Cracking is a major issue.
We are getting so many technologies like this that are so insanely specialized and require such deep knowledge of physical processes that some event like a 1918 flu that kills 10% of the population could take out enough people that we'd be unable to make things like this for years as we fill in the missing knowledge workers.
It's just crazy how far down the scale we're tickling atoms these days.
Not really at all. Between those dates and now the world has become fully global and concentration of knowledge has massively occurred. The past world was much more highly distributed, far less technical, and far more redundant.
The problem I’m finding, at least with today’s models, is that it produces disposable software. It’s not good at making well-architected, durable software. Stuff that could be maintained and bug-fixed.
I mean you can get good stuff out of it if you know what you are doing and guide it, but if you just say “here’s a regression suite. Write an implementation that passes” you will get something that works for a while but ages quickly and will need to be thrown away.
With today’s technology I’d still want a GPU driver developer guiding the LLM rather than some rando who is out of their element. But cutting down the exploration cycle time and giving the developer massive parallelism (have 10x agents exploring different hypotheses or features) is the real win. We don’t need to skip all the way to slop just to squeak out a little more effort savings.
OTOH, are you in an official role with extended powers because you represent the State? A role where you can routinely infringe on the civil and human rights of the people around you?
Why do you think the comparison is relevant at all?
I wish all this shit would just stop trying to search as I’m fucking typing. It’s like that person we all know that can’t hold their tongue for five fucking seconds and let you finish a sentence before the blurt something out.
Ah, but hitting enter now takes you to the top result! Therefore it has to search as you type, so there's something to activate as soon as you hit enter! /s
...I've been seeing this search style in more and more places recently, and I hate it. Let me type my search, then hit enter, then you can go find what I asked for. Don't slow everything down by searching on every character I type.
It’s called incremental search and it’s useful, but only when there’s some debouncing involved and the search results is presented at once or the items are sorted by the order they are found.
We talked about them in my neighborhood mainly because of package theft. If I ever thought there was a real chance of enforcement action, I’d be in favor.
You would rather infringe on the privacy of your community and invite a company known for indiscriminate multi-pronged dragnet surveillance (plates/faces/bluetooth beacons/etc.), with a nation-wide blanket coverage network that has already been provably abused by police on multiple occasions, all because of a handful of petty package theft cases?
I drive too even though there is a risk of death. The juice is worth the squeeze.
Except even armed with positive identification, the police won’t do anything. Laws with enforcement are worth nothing. Stalking and whatever else has happened with the data is already illegal. But who gives a shit without enforcement.
If you acknowledge that police regularly fail to act on crimes even with positive ID, why on earth would you allow a private company to collect all that information just for them to ignore/misuse? How is dragnet surveillance a fix for inept/corrupt policing?
Police have been granted wildly more powerful capabilities and arms over the course of my lifetime, but from what I've seen, those haven't improved their ability to solve crimes or better serve their communities nearly enough to justify their erosion of civil liberties; in my estimation, it's been an opposite trajectory, with violent escalations by police increasing and community trust in police decreasing (nearly monotonically) across the country.
Read the linked article; does that seem like "absurdly small" amounts of abuse to you? Besides, when we're talking about a surveillance network so vast you can't avoid it anywhere in the country, the law of large numbers kicks in: there IS abuse, and there WILL be further abuse at some level. The extreme reach of such systems means they should absolutely be treated as 4th amendment violations; the authors of that amendment could never have conceived of such a panopticon and its impacts on individual privacy and autonomy.
When I lived in FLA and it was super easy to get a concealed weapon permit. I carried a gun not infrequently just to exercise that right. I was serious about training with it as well.
The other states I’ve lived in haven’t been so straight forward.
I would expect it happens more organically where people’s discussion of the benchmark and posted results find their way into the training data, like anything else on the internet.
It’s not a step I’ve seen discussed very much, so it was kind of neat to see it discussed front and center in a lay article like this.
reply