j2kun

18 ms·

331.
by j2kun 2y ago
I don't think Apple is doing this. They compute the embedding on device in the clear, and then just do the nearest-neighbor part in HE (which does lots of dot products but no neural network). There are people doing NNs in HE, but most
332.
by j2kun 2y ago
Zama has already hit competitors with (French) patent charges. Apple's HE implementation is in Swift and uses BFV, which is very different HE from anything Zama does and doesn't use their source.
333.
by j2kun 2y ago
The majority of this article is scratch notes. How is this getting upvotes?
334.
by j2kun 2y ago
No, it's just textual input.
335.
by j2kun 2y ago
> Seeking and maintaining meaningful equations becomes difficult as we grow older. Meaningful equations, huh? Sounds like the kind of plagiarism where someone replaces words with things found in a thesaurus.
336.
by j2kun 2y ago
Yeah OP is out of their mind on this point. Even spotify doesn't have many true crime podcasts in their top 50.
337.
by j2kun 2y ago
> so I kept giving the model the output, and it provided updated formulas back and forth for 4.5hrs I read this as: "I have already ceded my expertise to an LLM, so I am happy that it is getting faster because now I can pay more mon
338.
by j2kun 2y ago
I've been down that rabbit hole, and while some few select people do it, the vast majority of people doing this rely on PL papers that long preceded the attempt to put category theory in the mix.
339.
by j2kun 2y ago
The main problem is that category theorists keep writing stuff claiming that, if only you get through the math, you'll eventually be able to apply the category theory to typical everyday programming problems. And by the time they get t
340.
by j2kun 2y ago
Presumably this doesn't apply to the standard being released here, nor any of its implementations made available. Each of these appears to have its own permissible license.
341.
by j2kun 2y ago
What a weird comment. Are you trying to argue by analogy that a decent fraction of the population are not capable of literacy? It seems self-evident that low literacy rates have nothing to do with innate ability. I see no evidence to sugges
342.
by j2kun 2y ago
The "information overload" problem always seems like a problem invented by the creators of these platforms to project on their users and justify coercive behavior.
343.
by j2kun 2y ago
Pay to email has been tried multiple times, and failed.
344.
by j2kun 2y ago
I mainly use my Remarkable 1 (am I ancient?) for scratch math, with the paper or code I'm reading open on a computer or laptop. So I concur that having both a writing surface and a reading surface is necessary.
345.
by j2kun 2y ago
Yes, but I think that it can stand as an answer to the OP regardless of the techniques status as useful for particular domains
346.
by j2kun 2y ago
The closest I know of is the work on UMAP. I interviewed Leland McInnes who explained to me in detail how category theory was a big part of helping him connect the dots, even though the final result does not strictly need it in the actual c
347.
by j2kun 2y ago
Of all the human rights you could come up with, "maximizing shareholder value (at the expense of society)" is the silliest.
348.
by j2kun 2y ago
I believe the term for this is scalar evolution.
349.
by j2kun 2y ago
They're not punishing success, they're punishing illegal behavior. Breaking up monopolies has a long, successful track record.
350.
by j2kun 2y ago
The author mentions that WordPress charges for everything, but after migrating off WordPress to a static site, I've found most of the features WordPress bundles together are, in total, more expensive when paid for separately. WordPress
351.
by j2kun 2y ago
Is there any good reference (in this book or another) that gives a sense of what coloring methods work well for various practical problems? Do they still use graph coloring for register allocation--and if so, which method is used? I have he
352.
by j2kun 2y ago
It's possible, but counter to the point of a PhD: apprenticeship for research.
353.
by j2kun 2y ago
I have a Pomera and I like it. Just wish I made more time to write.
354.
by j2kun 2y ago
If you're asking for a legitimate explanation for why mathematicians came up with groups, it's because they wanted to find roots of polynomials (or rather, prove one cannot find a general formula for solving large-degree polynomia
355.
by j2kun 2y ago
Aware, just feels like it may point to a bug, since I don't do this very often when reading blogs.
356.
by j2kun 2y ago
If you didn't use an LLM, how did you see this? I thought I had sufficient tooling in place to hide the html element from browsers and screen readers, e.g.
357.
by j2kun 2y ago
> some people's brains just don't like math Sounds like perverse socialization to me. What actually justifies this claim? Do some people's brains "just not like reading?" Do some people's brains "just n
358.
by j2kun 2y ago
It's useful for situations that would otherwise be illegal, so that tradeoffs are less relevant.
359.
by j2kun 2y ago
TIL about the project he's going to focus on: https://go.googlesource.com/oscar/+/refs/heads/master/README... An LLM-based architecture for helping maintain OSS projects. Seems cool.
360.
by j2kun 2y ago
I don't see how a bad actor can "rug pull" when everyone has a different encryption key. A cryptographic scheme weakness is a threat to all computing systems, it's not worse for FHE. It's just that FHE relies on new

More ›