j2kun

27 ms·

451.
by j2kun 3y ago
> in couple of weeks you can have deeper understanding on this topic than most of the practitioners out there This book is just over two months old... Are you saying _you_ did this since Oct 31?
452.
by j2kun 3y ago
The use of "eigen" here seems to have stretched too far beyond its original meaning in mathematics, and in German more generally. In German, "eigen" means "own" like "my own idea." In mathematics, &qu
453.
by j2kun 3y ago
Or don't return any search results for searches for random hashes
454.
by j2kun 3y ago
You misunderstand that this technique was executed by using YouTube _search_ to find videos, not by querying the exact URL. They can doctor search results however they like.
455.
by j2kun 3y ago
You should see my math students who beg me to give them sheets of arithmetic problems.
456.
by j2kun 3y ago
Isn't there sufficient evidence to conclude that OpenAI is training on things they know they probably shouldn't be training on? Like the full text of novels that are not publicly available on the internet? I don't think it&#x
457.
by j2kun 3y ago
They are upstreamed in LLVM and work today, though they are feature incomplete. From what I understand, the main reason you shouldn't use them is that development is stalled because the people who were working on them went to work at M
458.
by j2kun 3y ago
It was like 30s compared to Justine's 45. Apparently I have the right drivers installed, but apparently you need to compile the executable with the right toolkit (?) to make it work in WSL: https://docs.nvidia.com/cuda&
459.
by j2kun 3y ago
Thanks for the tip! I have 5.8.1 but I see I'm one major version behind.
460.
by j2kun 3y ago
I just tried this and ran into a few hiccups before I got it working (on a Windows desktop with a NVIDIA GeForce RTX 3080 Ti) WSL outputs this error (hidden by the one-liner's map to dev/null) > error: APE is running on WIN32 i
461.
by j2kun 3y ago
I found TableGen frustrating when I first started working with MLIR, in part because I expected it to be a layer of abstraction, but it very much is not. It is (in MLIR) mainly convenience tool for autogenerating boilerplate code. Once I re
462.
by j2kun 3y ago
It's not entirely doubled down on. There is work on IRDL, a domain-specific language for dialect definitions, and PDL, a language for pattern definitions. In theory, those could replace tablegen one day. I haven't looked into eith
463.
by j2kun 3y ago
> It might not be the actual truth or what's going on inside the model. This sums up pretty nicely why prompt hacking is not science. A scientific theory is related in a concrete way to the mechanism by which the phenomenon being st
464.
by j2kun 3y ago
Except there's also, you know, building coherent theories and using those theories to predict the system behavior.
465.
by j2kun 3y ago
To what extent have the problems in the book been addressed since 1994?
466.
What should I do if I suspect one of the journal reviews I got is AI-generated?(academia.stackexchange.com)
140 points by j2kun 3y ago|59 comments
467.
by j2kun 3y ago
np! If you're interested, you might also want to check out the compiler toolchain I'm working on ( https://heir.dev ) which we're working on building into an industry standard.
468.
by j2kun 3y ago
What makes you think it's a scam and snake oil?
469.
by j2kun 3y ago
It's unclear when this was written, but there are a few stale aspects of the library list. For example, the TFHE library has been long deprecated, replaced by tfhe-rs and Zama's new concrete framework. And PALISADE was renamed a w
470.
by j2kun 3y ago
IIUC, the main thing that Google's internal codesearch does that makes it superior to external systems (outside of an IDE, like GitHub code search) is that Google actually builds everything, and so it can incorporate that information i
471.
by j2kun 3y ago
That's what an abstract is for: > Our experimental results support the conclusion that neither version of GPT-4 has developed robust abstraction abilities at humanlike levels.
472.
by j2kun 3y ago
South Korea used it for their COVID contact tracing system.
473.
by j2kun 3y ago
It's also not complete: only a handful of functions are known to have functional encryption schemes.
474.
by j2kun 3y ago
It's an exciting time to be working in homomorphic encryption!
475.
by j2kun 3y ago
What prior ML/AI experience does Sam have?
476.
by j2kun 3y ago
Smells like an instance of https://xkcd.com/882/
477.
by j2kun 3y ago
Oh hey, I enjoyed your hashing multisets article and wrote my own take on the topic here: https://jeremykun.com/2021/10/14/group-actions-and-hashing-u... Thanks for writing it and clueing me into the idea!
478.
by j2kun 3y ago
What age was that in your case?
479.
by j2kun 3y ago
The code has an internal analogue, and the tooling lets you choose whether to export the entire git history or squash it. They may have chosen the former, in which case it could just be 2 years to migrate and rework the code to be ready for
480.
by j2kun 3y ago
I'm aware the Haskell designers like category theory, and so yes, the question boils down to the degree of influence of category theory on particular language features. Still, to your last point, the existence of an equivalence doesn&#

More ›