oofbey

18 ms·

391.
by oofbey 2y ago
This is really pretty cool. LLM's are so bad at images, it just makes sense to use reasoning to improve them. I'd love to see this applied to a bigger model than 3B, because this task is not difficult. But the attention visuali
392.
by oofbey 2y ago
If your stuff can fit on a single server at home, and are comfortable managing it, by all means do! It’s definitely way WAY cheaper, and if budget matters, that’s great. Nothing wrong with that IMHO. Obviously you can’t 100x overnight, b
393.
by oofbey 2y ago
Honest question - why would you want stateful workloads in a separate cluster from stateless? Why not just use a namespace?
394.
by oofbey 2y ago
Because galaxies spin. A spinning ball will compress into a disc after a while. The parts which are close to the axis just fall into the middle and some amount of friction slows them down until they stay close to the plane.
395.
by oofbey 2y ago
If you do find yourself wanting to create a cluster by hand, it's probably because you don't actually need lots of machines in the "cluster". In my experience it's super handy to run tests on a single-node "cl
396.
by oofbey 2y ago
I think connectivity would be a key challenge for this device. RPI zero2W only has WiFi4 == 802.11b/g/n. In a modern crowded building, 2.4GHz is often super crammed and busy, and I find wifi4 barely works. Lots of dropped packe
397.
by oofbey 2y ago
I abhor the style and disagree with most of their comments.
398.
by oofbey 2y ago
I'm guessing you're still writing every line of character of code yourself in emacs or notepad or something like that? Yes, copilots are really helpful, even if you do know what you're doing. Or just wait to get left behind,
399.
by oofbey 2y ago
"Not-so-distant" indeed. IMHO this sounds just like banning people from using spell checkers when writing documentation. Legal paranoia aside, it's completely untenable in even the short term.
400.
by oofbey 3y ago
I'd love to see a chart of lego part prices over the years. Instead, what I saw was a couple charts showing only the outliers of the most-expensive parts over the years. Data visualization fail.
401.
by oofbey 3y ago
Of course very few people are experts in visual display of data, but this one is so horribly bad. Log scale please? The title of the post should be a simple chart, and there are charts on the page, but it's completely impossible to g
402.
by oofbey 3y ago
These devices must now all be treated as if they’re rooted. That’s a gigantic security problem. Casual analysis here says “well it would be hard to install a new OS, so it’s not a big deal.” That is a terrible security assumption. We don’t
403.
by oofbey 3y ago
When a critical part of a security system is defeated, saying “that’s okay we have other layers of security so they probably can’t do much” would be IMHO pretty horrible security practice. Taking the laptops away is horrible, but from a sec
404.
by oofbey 3y ago
If it all fits in memory, isn't it simpler to just load it first? Sure, you're kinda painting yourself into a corner, but it can be a big corner. e.g. I know that until recently Amazon still ran a bunch of key recommendation alg
405.
by oofbey 3y ago
How difficult is it to port pandas code to use polars instead? I use pandas a bunch, but sometimes it can be quite slow, especially with file I/O. Anybody have benchmarks?
406.
by oofbey 3y ago
This is the whole reason for the 737 Max debacle. They tried to keep the controls the same so pilots wouldn’t need retraining, even though the plane was very different. They tried to paper over the differences in software. And hundreds died
407.
by oofbey 3y ago
Any decade now people are going to realize how awesome XMPP is and start using it for everything!
408.
by oofbey 3y ago
Nope. It's on its way down towards pre-covid levels, but still (by the article's estimate for 2023) 10% higher than pre-covid. Such a bad headline. Look at the raw data at https://cde.ucr.cjis.gov/LATEST/web
409.
by oofbey 3y ago
Complexity and understandability are very difficult to separate. If a coding style is unknown / bizarre to most coders on a team, but beautifully simple to a few who know the technique, which is it? There should be a balance between te
410.
by oofbey 3y ago
That argument would be more analogous if you picked say CoffeeScript. The point is it’s something that used to be reasonably popular but for reasons the vast majority of the world has moved on from.
411.
by oofbey 3y ago
The explanation all makes sense. But the key line of “we all know cvs” is effectively exclusionary to all the other developers in the world who don’t use cvs. At some point they will need new talent which will be harder to get.
412.
by oofbey 3y ago
The internet is of course much older than the WWW. And the internet had and still has many non-WWW ways to communicate and publish content. But applying the term "WWW" to any non-HTML non-HTTP technologies is revisionist, and I&
413.
by oofbey 3y ago
“The World Wide Web … is arguably roughly 40 years old.” 1993 was 30 years ago not 40. I see no valid argument that there was any WWW in the early 1980s.
414.
by oofbey 3y ago
Cute trick but the opposite of helpful. Is the goal of your paper to brag or educate?
415.
by oofbey 3y ago
It’s counter-intuitive, but the so-called hoop stress on a cylinder being compressed from the outside actually results in tensile stress not compressive. This is because the material as it resists shrinking to a smaller radius is being pul
416.
by oofbey 3y ago
Completely agree with the key point here. An example that blows my mind is the incredibly unresponsive UI built in to my $2,000 television. How does a very expensive device like that end up with a series of clickable boxes that lag so much
417.
by oofbey 3y ago
I’m a big fan of the functionality of witch and have had it installed for years. But honestly I’ve just about given up on it. It keeps breaking. Mac OS of course keeps updating itself and very often those updates cause Witch to stop working
418.
by oofbey 3y ago
GPT definitely makes mistakes. GPT4 less than GPT3.5, and way less than Bard. Having spent quite a lot of professional time working through real-world problems with GPT4, I'll say it very rarely does something I'd call hallucinat
419.
by oofbey 3y ago
Sure, my point is that only a googler would even consider this kind of breaking change as a sensible option. People in the real world with regular code tooling would reject the proposal before it got started. The analogy to Angular that ot
420.
by oofbey 3y ago
They are behind because what they have released sucks. Have you tried Bard? It's dumb. Like you're talking to some 20th century gimmick dumb. GPT4 is far from perfect, but when it makes mistakes and you point them out, it unde

More ›