19h

16 ms·

121.
by 19h 4y ago
This article reeks of weak sauce. Delivering truth in a "sensitive" manner is a sad testament to the frail sensibilities of modern society. If the unvarnished truth bothers you, that says more about your inadequacies than the pers
122.
by 19h 4y ago
Unified memory allows both CPU and GPU to use the same memory, effectively giving a MacBook with 96GB of memory 96GB of VRAM (minus OS overhead obv).
123.
by 19h 4y ago
This is weirdly relevant -- I made some funny soviet-style propaganda posters the last few days and it happily let me generate them for Gaddafi, Putin, Kim Jong-un, etc. but when I wrote Xi Jinping it rejected it. I tried "general secr
124.
by 19h 4y ago
Using electricity uses electricity..
125.
by 19h 4y ago
If you have enough memory, I the entire file eventually is in memory because it doesn't need to be paged out. If it takes up less memory (because your memory is limited), it'll still need to page the file off disk when it hits t
126.
by 19h 4y ago
Pretty sure there’s barely any meaningful Training data for this? Would be interesting to see these models return a kind of coefficient indicating how “often” a semantic context has appeared in the training data..
127.
by 19h 4y ago
Not sure if I understood that right -- is that something like Whisper + an LLM? Like [0]? If OpenAI adds speech input to ChatGPT -- and considering the upcoming plugins -- isn't a possible enterprise specialisation of VoCode the only v
128.
by 19h 4y ago
It boggles my mind how they're not absolutely checking the user & conversation id for EVERY message in the queue given the possible sensitivity of the requests. How is this even remotely acceptable? In the one reddit post first sur
129.
by 19h 4y ago
I used bazel for a few internal repositories written in Rust and it's amazing .... until you hit an issue. Edge cases in Bazel are about as documented as any random fish you might find in a crack at the bottom of the ocean. And the mos
130.
by 19h 4y ago
I wonder though --- will GCC have to comply with a specific API of cargo? Wondering if this will end up being an awkward kind-of RUSTC="gcc rust" cargo build thing, or if this will be some lower level API.
131.
by 19h 4y ago
Really want to know about the supply chain of Qualia (from Neurohacker Collective). How would I go about finding this out?
132.
by 19h 4y ago
Unfortunately we're not just searching for things but extracting word frequencies of every user for stylometric analysis, so we need to do custom crunching. Spreading this task into many sub-slices of the files is annoying because the
133.
by 19h 4y ago
To analyze and process the pushshift Reddit comment & submission archives we used Rust with simd-json and currently get to around 1 - 2GB/s (that’s including the decompression of the zstd stream). Still takes a load of time when th
134.
by 19h 4y ago
You're absolutely right, but I felt like the axiom-approach by x-complexity was fascinating for breaking down the discussion, and I was positively surprised by how GPT4 approached it.
135.
by 19h 4y ago
Sorry about that wording. I'm a bit biased. In the previous firm I worked as VPE I was forced to let an entire team go due to financial constraints, and the following reviews were somewhat shocking. I'll never forget this one revi
136.
by 19h 4y ago
I agree this probably shouldn't be here, but I took all comments from this thread and asked GPT4 to 1) break down all arguments into axioms and 2) try to resolve them. Since I feel like this is pretty fascinating and a good read, I dec
137.
by 19h 4y ago
Literally all this shows is that women are more likely to say that the workplace is toxic nothing else; they literally just analyzed a lot of glassdoor reviews and that’s their entire base for this claim. Also for some reason I’m pretty sur
138.
by 19h 4y ago
Georgi rewrote the code on top of his own tensor library (ggml[0]). [0] https://github.com/ggerganov/ggml
139.
by 19h 4y ago
I took the insertion_sort impl from the bottom of the post and asked gpt4 to rewrite it into idiomatic Rust: pub fn insertion_sort(n: i32, p: &mut [i32]) { for i in 1..n as usize { let tmp = p[i]; l
140.
by 19h 4y ago
Pretty sure if Tencent wasn’t involved the FTC wouldn’t be this invested into making Epic bleed ..
141.
by 19h 4y ago
What is the license of ATC audio? This could easily be made into an amazing repository for lo-fi music creators.
142.
by 19h 4y ago
Germany -- 5.529,00 € excl. Apple Care which is 149,99 €/Year
143.
by 19h 4y ago
I couldn't get the mps python version to run, it's insane how much setup it requires... I don't think the Gerganov C++ version uses CoreML or Neural Engine. I previously tried to play with ANE based off prior reverse engineer
144.
by 19h 4y ago
Currently running 65B on my 96GB M2 Max.. it's pretty good.
145.
by 19h 4y ago
420...00 as in weed.. Musk doesn't actually want people to have API access.
146.
by 19h 4y ago
Reminds me of the time in 2016 when I added a single line to one of our internal node modules to demonstrate how you can take over our entire Mesos cluster with a single compromised dependency. Spawned a program on every single node that wr
147.
by 19h 4y ago
https://github.com/ggerganov/whisper.cpp has diarization
148.
by 19h 4y ago
Actually came here to mention rewind, too. It’s simply put a groundbreaking game changer for me. Be it finding text in conversations, using it to recall the face of an applicant when their name doesn’t ring a bell, find code, find text in w
149.
by 19h 4y ago
I agree with the sentiment that dating apps have changed the dating landscape and have made it easier for people to connect with potential partners. However, this has also led to an overabundance of options and a culture of instant gratific
150.
by 19h 4y ago
I might have been compromised by AI recently but I really expected this page to tell me where I can a font and where it looks, feels and works best.

More ›