krackers

14 ms·

301.
by krackers 6mo ago
This needs a story. What did you say to him?
302.
by krackers 6mo ago
You must be a shape rotator.
303.
by krackers 6mo ago
Beautiful explanation, thanks!
304.
by krackers 6mo ago
Yeah the ERD result was known before I think, see "The Remarkable Robustness of LLMs: Stages of Inference?" https://arxiv.org/abs/2406.19384 But the fact that the intermediary circuits are generic and robust
305.
by krackers 6mo ago
Platonic representation hypothesis? https://arxiv.org/abs/2405.07987 Maybe in the same way fourier/wavelet basis is just the most natural way to work with certain signals or images, there's a certain represen
306.
by krackers 6mo ago
On intel macs there used to be single user mode, but even then I don't think you ever had control over the framebuffer.
307.
by krackers 6mo ago
> this uses a harness This seems like an arbitrary restriction. Tool-use requires a harness, and their whitepaper never defines exactly what counts as valid.
308.
by krackers 6mo ago
Isn't this basically what javascript went through with Promise chaining "callback hell" that was cleaned up with async/await (and esbuild can still desugar the latter down to the former)
309.
by krackers 6mo ago
LLMs already do this and have a system role token. As I understand in the past this was mostly just used to set up the format of the conversation for instruction tuning, but now during SFT+RL they probably also try to enforce that the model
310.
by krackers 6mo ago
Most (all?) of this holds for quantizing convnets too, if you're looking for an easy exercise you can play around with quantizing resnet50 or something and plotting layer activations
311.
by krackers 7mo ago
Probably iStat menus or something similar
312.
by krackers 7mo ago
The LLM is just calling APIs though, if the LLM can do it then it should be exposed to the user. Why have the middleman.
313.
by krackers 7mo ago
Here's an evil business idea: Use the LLMs to identify the users most likely to be "vocal influencers" and then prioritize resources for them, ensuring they get the best experience. You can engineer a bubble this way. And the
314.
by krackers 7mo ago
>I recorded the call [^For training and monitoring purposes, of course] Heh I just realized that the "may be recorded" bit could also be interpreted as them giving _you_ permission to record the call
315.
by krackers 7mo ago
There is this famous quote from Bentley on asking programmers to write binary search >I’ve assigned this problem [binary search] in courses at Bell Labs and IBM. Professional programmers had a couple of hours to convert the above descr
316.
by krackers 7mo ago
Because the planetary alignments won't allow for another launch any time soon... which I guess has a natural correspondence in that the macroeconomic conditions (read: bubble) mean that now is a great time to play around with things wh
317.
by krackers 7mo ago
I mean people use IDA for decompilation instead of reading assembly directly, and that's counted as as valid reverse engineering. Why shouldn't throwing it an LLM count as well? I think reverse engineering is one of the fields whe
318.
by krackers 7mo ago
Unless it was someone else larping as him, as recent as feb this year he posted that he's still living there (apparently got an extension or something). He was arrested, but he got released within a few days.
319.
by krackers 7mo ago
Annoying that they removed coverflow and replaced it with an inferior gallery view. Not sure it was for patent reasons given that they eventually won that case, I suspect the 3D aesthetic just didn't fit in with the flat post-10.10 ui.
320.
by krackers 7mo ago
Oh yes I've seen lots of old images that have that sort of feel. I always assumed it was because of poor sensor quality or something
321.
by krackers 7mo ago
Can you post a screenshot comparison? I'm curious what these "crunchy jpegs" look like
322.
by krackers 7mo ago
>neat to use the binomial theorem to see this, Yeah as far as I understand this basically _is_ where the connection to calculus comes in. As the article points out, it's very similar to computing the derivative via infinitessimals,
323.
by krackers 7mo ago
>for any y >= n with f(y) = 0 (mod n), there's some x between 0 and n-1 There's a simpler way to see this, any such y can be represented as y = n k + x where i,j are divisor & remainder. Then f(y) = f(n k + x) = f(x) mod
324.
by krackers 7mo ago
It's actually a bit refreshing that they didn't brand this with the usual "LLM hype". And it's actually a good example of someone using LLMs to solve a problem by bringing in their domain knowledge. (The solution is
325.
by krackers 7mo ago
>Source code in a higher-level language is not really different anymore Source code is a formal language, in a way that natural language isn't.
326.
by krackers 7mo ago
In [1] I think a commenter actually speculated about a design just like this, where later layers can directly access outputs of previous layers instead of having to store it in the residual stream [1] https://news.ycombinator.com
327.
by krackers 7mo ago
The one thing I didn't quite understand (and wasn't mentioned in their paper unless I missed it), is why you can't keep stacking turtles. You probably get diminishing returns at some point, but why not have a meta-meta-verifi
328.
by krackers 7mo ago
Based on the comment from [1] it seems like the issue with nasdaq is that anyone tracking it is contractually obligated to include spacex? What about for other funds? VIFAX description says >The Global Equity Index Management team applie
329.
by krackers 7mo ago
There is https://magazine.sebastianraschka.com/p/technical-deepseek which shows an evolution in deepseek family
330.
by krackers 7mo ago
That makes sense, they haven't "built immunity".

More ›