ssivark

26 ms·

181.
by ssivark 2y ago
This shouldn’t have been the least bit surprising — since this is what happened to every mass communication technology over the past few centuries: print, then radio, then television. The OG computing folks at Xerox PARC indeed foresaw all
182.
by ssivark 2y ago
Why is it a jab at physics? It's honest and beautiful -- I imagine this is exactly what an experience on the cutting edge of experiment is like! :D Making this measurement (an ancient discovery) with latest equipment is easy, but imag
183.
by ssivark 2y ago
To elaborate — the task definition itself is vague enough that any evaluation will necessarily be vibes based. There is fundamentally no precise definition of correctness/reliability.
184.
by ssivark 2y ago
Why is it a good model to allow some software engineer "at a distance" to enforce whether some downstream user must drop every other priority and upgrade? I agree that attitudes towards security are generally very poor, but breaki
185.
by ssivark 2y ago
This is basically the debate between Alan Kay and Rich Hickey... on HN, almost a decade ago :-) https://news.ycombinator.com/item?id=11945722 The key idea is (bare) data -vs- objects (which encapsulate data and let you inte
186.
by ssivark 2y ago
Related, from the same author: https://maxwellforbes.com/posts/dont-make-things-actually-wo... As potent and productive as the combined perspective sketched out by the author might be, it makes me wonder whether this m
187.
by ssivark 2y ago
With the author's emphasis on Science 2 , I found quite provocative an earlier post of his: https://maxwellforbes.com/posts/your-paper-is-an-ad/ Submitted it for discussion at https://news.ycombi
188.
Your Paper Is an Ad(maxwellforbes.com)
2 points by ssivark 2y ago|1 comments
189.
by ssivark 2y ago
As Heidegger pointed out in _"The question concerning technology"_ the driving mindset behind industrial technology is to turn everything into a (fungible) standing resource -- instrumentalizing it and robbing it of any intrinsic
190.
by ssivark 2y ago
Very excited to see parsing for ill-defined states! I like your naming scheme of using animal sounds, but just wanted to bring to your attention that Emacs already has a popular package named meow (for modal editing) https://gi
191.
by ssivark 2y ago
Why do you assume everyone in the company (including folks working on infra) are authors on the paper? That’s possible, of course, but isn’t it unlikely?
192.
by ssivark 2y ago
AI is such an ill-defined word that it's very hard to say what it's definitely not. Marvin Minsky -- father of classical AI -- pointed out that intelligence is a "suitcase word" [1] which can be stuffed with many differe
193.
by ssivark 2y ago
As one example of such a more general framework -- (variants of) belief propagation might be a good answer if dataset constraints could be cleanly formulated as distributions to be reasoned with.
194.
by ssivark 2y ago
I guess my response was to how I read your comment fitting in with the higher level discussion. My main point is that many of these algorithms are intuitive, and kids learn these much earlier than when they learn formal programming (which m
195.
by ssivark 2y ago
> no _way_ is binary search the first algorithm people learn. For binary search to even be intelligible you already have to know, at a minumum linear search and the concept of sorting Almost every 6-10 year old kid who had to use physi
196.
by ssivark 2y ago
Think from an information theory perspective. It is rarely true that you cannot say anything more about the data than what is assumed by classical algorithms. We almost always have some more information depending on the specific domain unde
197.
by ssivark 2y ago
Have you tried using PythonCall.jl (which allows you to basically use any random Python library, in principle) with Pytorch?
198.
by ssivark 2y ago
While this is from the Cline blog (and therefore not quite unbiased), it was interesting to see the contrasting framing of approaches. Sharing the post here with the hope of encouraging discussion on this.
199.
Contrasting the Cline and Cursor approaches to the developer-AI relationship(cline.bot)
3 points by ssivark 2y ago|1 comments
200.
by ssivark 2y ago
> On the other, something about the particular gradient used makes me deeply uncomfortable. Just obvious enough to catch your eye, but indistinct enough to make you feel like something's wrong with your eyes. Same; I personally find
201.
by ssivark 2y ago
I was always very frustrated with git workflow for working on multiple features/bugfixes simultaneously (multiple branches [1]). Changing between them, or combining them for testing is tedious -- constant stashing, switching, cherry-pi
202.
by ssivark 2y ago
> But we went past that very rapidly, and for the past half a year or so, we've already seen models excelling at every single task listed above simultaneously. Same architecture, same basic training approach, few extra modalities, e
203.
by ssivark 2y ago
Kudos; feels very timely! I feel that one underappreciated nuance is why we cannot use human examinations to judge AI. I haven't seen this satisfactorily spelt out anywhere, so I recently wrote a Twitter thread [1], including an exam
204.
by ssivark 2y ago
> you don't want a random person become hair stylist for the day and then switch to something else Why though? The easier it is to switch in and out of careers, greater the likelihood that people will be well-matched to job opportun
205.
by ssivark 2y ago
It's likely that inefficiencies like this contribute O(40%) to every slice of the pie, so fixing any one slice seems unimportant, but cumulatively they would make a big impact on the situation. A second hypothesis is that the inefficie
206.
by ssivark 2y ago
What makes you so confident that the benefit is less than the effort? Given the increasing likelihood of supply chain attacks, isn’t this a very prudent precaution?
207.
by ssivark 2y ago
But how does one get the composability of multiple dispatch? A problem with the Python world IME is the need to reinvent everything in Jax ecosystem, in the Pytorch ecosystem, etc -- because of fundamental language limitations. This is tedi
208.
by ssivark 2y ago
Adaptation need not only mean genetic (slow), it could just as well be epigenetic. The latter is not well-understood enough to be taught in school textbooks yet, but has been of great interest in recent decades. https://en.wikip
209.
by ssivark 2y ago
> Commits are snapshots of a tree. They have a list of ancestors (usually, but not always, just one). Tags are named pointers to a commit that don't change. Branches are named pointers to a commit that do change. The index is a tiny
210.
by ssivark 2y ago
Sometimes when I'm browsing at night, and have the OS/browser set to dark mode, it's really annoying when one website suddenly defaults to light mode instead of respecting my settings. The only thing more annoying is to have

More ›