gcr

45 ms·

901.
by gcr 4y ago
EXACTLY! there's so much research and citizen science being done, and even the field of CS is no good at identifying, preserving, and disseminating the good ideas that our lossy proxies miss. (proxies like "university prestige&quo
902.
by gcr 4y ago
Right, but transformers are artificial neural networks though, so it's not clear to me what you mean. The leap from RNNs to the Transformer architecture in 2017ish is similar to vision networks' leap from fully-connected layers
903.
by gcr 4y ago
i'm surprised that Apple doesn't do directory-level metadata directly in the filesystem itself, e.g. with xattrs on the directory inode.
904.
by gcr 4y ago
I think OP was talking more about upending the universal dogma of "statistical optimization to minimize feedforward loss functions via gradient descent." We now have plenty more tricks of the trade (ReLUs were a big one, self-atte
905.
by gcr 4y ago
i am genuinely curious, because I think CL could do some amazing things in the symbolic algebra/compiler space here. Tensorflow's AutoGrad and AutoGraph (and I think JAX maybe) had to do terrible things to the python bytecode inte
906.
by gcr 4y ago
Every week or so, my Windows 10 desktop pops up a dialog telling me to upgrade to Windows 11. I spend about a minute looking for the one tiny link to dismiss the dialog without upgrading. I swear it changes locations each time... Lately I&#
907.
by gcr 4y ago
Most users interpret "Export" as an operation that lossily finalizes a project from its native representation to a common third-party format. "Export" is one-way, "Save As" is bidirectional. - Audacity "Sa
908.
by gcr 4y ago
Practitioners make distinctions between machine learning (statistical analysis + optimization) and artificial intelligence (symbolic reasoning) for important reasons. For starters, machine learning systems based on matrix multiplies and acc
909.
by gcr 4y ago
on the minimal chance you're not being sarcastic, could you say more about this?
910.
by gcr 4y ago
interesting, is this a supported SQLite feature? I understand that when the WAL file gets too big (e.g. 4mb), SQLite will then checkpoint its contents back to the main sqlite database. How can one generate a WAL file big enough to contain a
911.
by gcr 4y ago
USA and European university rankings have a strong bias against universities in the Global South (especially India), in part because of the stigma against Indian programmers and in part because they're generally wanting to promote US a
912.
by gcr 4y ago
I do that too! It’s nice to have some “scratch space” at the end, like an artists’ palette to dab colors from. Ideas naturally flow from scratch to outline to fleshed-out paragraphs this way.
913.
by gcr 4y ago
From my understanding (which seems to differ from OP), the benefit of running models on ANE is for much lower power usage, not necessarily speed. Apple’s always-on Hey Siri wake word detection uses ANE for minimal battery life. I’m not sure
914.
by gcr 4y ago
Tell us more. What hurts the most? From your post, I'm hearing that you have a lot on your plate -- I'm seeing some disgust with your body, some frustration with your creativity and mental abilities, some fear about making a relat
915.
by gcr 4y ago
In NYC, some restaurants could reopen and conduct business as usual if they were classified as outdoor venues. This interacted oddly with another regulation that let a restaurant count as “outdoor” if part of its seating area had outdoor se
916.
by gcr 4y ago
like either you believe openAI could be a better steward of this technology than politicians, or you don't. justifying crappy corporatist behavior from oAI by reasoning that governments and society also behave badly is ceding ground im
917.
by gcr 4y ago
This is how linden labs used to represent user-created models in Second Life! you could just upload a .jpeg of your model geometry. Not sure if that happened in the vertex shader - probably not - but a similar idea
918.
by gcr 4y ago
Inigo Quilez does some incredible stuff with signed distance fields! Those don't use textures in the traditional sense, but see https://www.youtube.com/watch?v=8--5LwHRhjk
919.
by gcr 4y ago
sorry for nitpicking but that's actually a misconception - the GPU doesn't generate intermediate vertices as a result of normal maps! all of that happens in the fragment shader, which only changes how light behaves at the fragment
920.
by gcr 4y ago
Christians genuinely have the faith that gods and devils exist. It's a genuine belief, they're convinced by whatever evidence they see, even though outsiders don't necessarily understand. Maybe they're not 100% there, bu
921.
by gcr 4y ago
Let's flip this around a little. "The agenda of Christianity here is very simple: make sure the devil doesn't kill us. Christians believe that earnestly. That's the end goal. Everything is in service of that and not the
922.
by gcr 4y ago
There are strong parallels between a belief in AI supremacy (roko's basilisk and the like) and belief in divine punishment (pascal's wager and the like). I've heard people mockingly describe superintelligence as "catholi
923.
by gcr 4y ago
Hang on - if I understand correctly, all of the following is true for Pixel 6, Pro, and 6a users?? - There's an exploit out there that lets attackers own my phone if they know my number - A patch is not available for my phone yet - It&
924.
by gcr 4y ago
“Being brought in” could be the crucial difference, yeah? When a company starts with beans on staff from the start, maybe there’s more of a sense of security
925.
by gcr 4y ago
Ah so I am! Thanks for the correction
926.
by gcr 4y ago
I took this course when it was PL-focused in 2015ish! It starts with an in-depth exploration of the untyped lambda calculus (with the usual fixins like Church numerals, various combinators, etc), spending half a week or so on semantic expan
927.
by gcr 4y ago
Fella after my own heart. On all my video calls in 2020, I had a Zoom H1 with a fuzzy pop shield feeding into a simple USB interface on a rickety tripod. Even with this simple setup without bells+whistles, I got so many compliments on how g
928.
by gcr 4y ago
There are also good ways of using LLMs to write better papers that don't even require the LLM to contribute to the text! Consider feeding your outline through the LLM alongside prompts like: - "What's missing from this draft
929.
by gcr 4y ago
My understanding is that the original Transformer paper ("Attention is All You Need") and BERT papers came from Google. In particular, BERT's pre-training task is to predict how to fill-in-the-blank from a sentence that'
930.
by gcr 4y ago
Expanding on this history for readers here who are just catching up on this stuff, modern feedforward NNs descended from the Perceptron paper by Rosenblatt and friends from 1957. This was the first paper that really tried to position "

More ›