I really wish sites would stop with the clickbait style headlines. “Mysterious” is unnecessary here since there isn’t much mystery about this. If anything, this kind of wording gives non-experts and non-scientists the wrong message since mysterious often gets read as “scary” or “unexplainable”, which is often misleading.
Cool site. One thing I wish we had were rich, open archives of news that could be fed into something like this so we could see not just the live updating perspective, but a retrospective time series of the fascism-o-meter over a long time. Especially seeing how the different contributing categories change over time, and how you likely would see some of them start ticking up earlier than people might expect.
This problem spans the sciences, not just ML. In the ML space I can see the IP issue people might claim given the commercial potential these days, but given that the same reproducibility issue arises in domains with no obvious IP protection motivation, that seems like a weak cause. In my experience the incentive just isn’t there to do the work to make stuff reproducible. Very few people get in much trouble if their work isn’t reproducible since most papers get very few reads and even fewer readers who would try to replicate it. I think publication venues (conferences, journals) should make artifacts a mandatory requirement for publishing.
A surprisingly large number of people here are sensitive when articles or opinions don’t agree with them. The flag and downvote are their usual weapons of choice. It’s odd to blow off “politics” that impact trade and business on a site crawling with people who, you know, want to get rich in trade and business.
Get into a different part of the industry: I’ve never touched any of that stuff and have gotten on quite fine for 30 years so far. From the outside, the software engineering world of the web-centric tech industry is one that sounds absolutely dreadful (and has for a while). There is a ton of software being written that isn’t nearly as miserable.
Of course, if ones metric is “FAANG or nothing” or maximizing compensation, you get what you seek - which may be miserable work in exchange for status and cash. I realized a long time ago that neither status nor cash outweighs enjoying what I do.
This guy keeps spamming this package here. Polynomial time approximation algorithms for NP complete problems are not a big deal and are pretty well known for different problems. Yes, they’re useful, but I don’t think anyone in theoretical CS views the existence of approximation algorithms as saying anything about P vs NP.
Yup! I recently took a pretty substantial project I wrote in 2006 in Java that was sitting in a tar file, unpacked it, and it ran just fine with a current JVM. Even better, I was able to compile it again with no problems: again, with a modern JDK. I wish I could say the same for years of projects written in Rust and Haskell: the Haskell ones are almost all bit rotten since I haven’t kept them up to date with GHC, and Rust is similar. Those two are even worse since with Java I had a rich standard library, so I didn’t have any 3rd party dependencies in my 19 year old project. The Haskell and Rust ones rely on Hackage and crates, and keeping things working for that long would require me to do bonus work beyond my actual project code. I’ve had similar but less painful problems in Python land: the library APIs are mostly stable: the biggest headache there are projects that predate python3 so there’s some work to bring them up to date. That’s at least a one time investment.
Java: no bonus work induced by third parties to keep projects alive, keeps working for a long time. Would recommend.
Tutorial link too hard to click? Seriously: these comments inevitably come up for most language related pages, and I don’t see how they don’t fall under the site guidelines of “Please don't post shallow dismissals”.
It’s happening in other parts of the research world too: a couple colleagues of mine were talking recently about a paper we found at a conference last year that had a web page to go along with it with a domain and fancy graphics and such. For a boring programming languages paper. We concluded this is the modern way to try to jack your citations by getting noticed for everything but the technical content of the work, which is a bit off putting.
Publication metrics. This has been a problem for at least the last few decades, especially when it comes to papers from some parts of the world. Before the current breed of low quality journals, we saw this happening with conferences that were often branded with professional societies like IEEE, but if you looked carefully it would be “IEEE Section of (random place in Asia)” and not a conference sanctioned by IEEE itself. Those used to be full of garbage plagiarized papers and junk. I think the idea is that nobody usually reads the papers when they make a decision based on publication metrics, so pumping a CV with garbage often goes unnoticed. It’s not a wise practice: in many places getting caught doing this would be a career ender.
If you read the original post carefully, side loading wouldn’t solve the issue. The issue is app developers on iOS either focus on streaming or fill their apps with ad garbage. There is nothing stopping developers from building an app that focuses on the local library and doesn’t do ads, but nobody is offering any quality ones.
Nothing is shocking about the insurance industry anymore since they’ve been pulling this kind of thing for years. To me the most shocking thing is the people who defend them. I understand the drive to make money and build huge businesses: what confuses me is that we treat human health the same as things that don’t directly impact the health of people. It seems to me that there are endless opportunities to get insanely rich: I’ve never understood why we don’t set aside certain areas and just agree that they are not the place to do that. (I know, leaving potential profit on the table is sacrilegious to those who worship at the altar of the almighty dollar…)
The rise? Those kinds of languages have always been here and widespread. If anything you’re seeing the tapering off of the rise in other languages (JS predominantly) that took place over the last 15 years or so.
I’m not seeing what the issue is with ollama. Can you elaborate? There are tons of open source projects that other stuff gets built upon: that’s part of the point of open source.
I’ve never understood the attitude of “dreamed for years to get into FAANG”. I’d guess you’ll regret it if you have an attachment to being able to tell other people you work for one of those companies, or are attached to the compensation packages and can’t match it elsewhere. If it’s the technology/work that brought you there, I doubt you’ll miss it since lots of other places have exciting interesting work.
I think you’re the only one who can answer that question, and it really boils down to being very honest with yourself about why you dreamed for years to get into those places and what you’ll lose by leaving.
This is basically a problem of automatic vectorization and parallelization. This is not a new area: if you’re curious about it there is literature about this in the compilers and high performance computing world going back to the 70s. Short answer: it’s quite hard, especially for generic code. It’s achievable to some degree for array-oriented, loop heavy code, but even then with severe limitations often related to the regularity of the logic in the loop bodies, the iteration space, and the use of pointers. There has been some work for Python to do this (eg, numba), but it is largely for array heavy code instead of general Python.
About 4 years ago I was at a VC pitch day and one of the teams rehearsing their pitch before hand was basically proposing this for augmented reality. I was mildly horrified that people saw that sort of thing as something people would want, but hey - $$$ is all that matters right? I really hope those folks never got funded.
The “trends in software” is the problem here: we’ve gotten greedy as developers and basically build assuming resources don’t matter. I am quite willing to bet that if we considered resources, we’d still be able to build all the stuff we want without assuming resources are free. Hardware would be viable longer. Efficiency unfortunately seems like the last thing anyone thinks about anymore, instead favoring time to market and layers upon layers that prioritize developer happiness and productivity over all else.
Read up on smt-lib: learning how to encode problems in that is a good way to start. The Python z3 bindings are a good starting point to play with it too.
Interesting post, but I’m not sure this really speaks to what goes into actually writing what would be considered a “fast” SAT solver. It seems more like a post about how SAT pops up in a lot of places if you look at them right. For the state of the art in what constitutes fast solvers, the annual SAT competition papers are quite interesting to read if you’re interested in the techniques people come up with to make them fast. A few years ago I was working through Knuth’s satisfiability book and writing my own solvers, and was always amazed how stunningly fast the SAT competition winners were compared to the ones I’d code up.
One thing I miss from that era of machines was just the way they looked: at the time, most machines were grey or black boxes, but the SGIs had some degree of personality to them. The O2's were fun little curvy boxes. One of my favorites were the large rack systems - one of my jobs had us working with the Origin 2000 and PowerChallenge machines. Compared to some of the generic clusters of rack mounted Alpha systems that we had around the same time, the SGIs just had a cool look to them.
Given how often people seriously say things like the top level comment being responded to around here, an explicit /s is almost necessary since it can be hard to distinguish from the usual cynical dismissive comments.
Don't confuse "giving software away for free because people have been conditioned to expect software that costs nothing" with "open source". And I have no idea why ARM is on that list: sure, they broke the Intel monoculture, but they certainly aren't free or open in any sense of the word.
I bought one of those for this reason. I really like starlabs and want to support them, but the keyboard cover for that device is .. not great. If it had a keyboard similar to the apple ipad keyboard (the expensive one with backlighting - I forget their name for it), the starlite would be a much more compelling device. At this point mine just sits in a desk drawer and may find its way to eBay.
I use it regularly. Mostly for visualization and prototyping ideas. I’ve tried over the years to replace it with Python/Sage, but the giant library of builtin functions always pulls me back. Plus, the lisp-ish language I prefer over Python. The visualization capabilities are quite nice too - not just data or function plots, but the ability to construct relatively complex graphics programmatically is quite useful. On a recent project where I needed to build some data structures for spatial processing in 3d, I was able to build a nice 3d visualization to help me debug my code by visualizing the calculations I was doing. I also really like the backwards compatibility of notebooks: I have mathematica notebooks dating back to the 90s, many of which still work fine (modulo some visual ugliness due to changes they made to layout and presentation over the decades). Few other tools have that longevity.
Sometimes it seems that people who write these kinds of pieces forget that not everyone in the world does web or even web-adjacent work, and node.js is something we don’t even consider to be part of our ecosystem. Rewriting useful things in non-JS has the benefit of letting folks like me who avoid JS like the plague use useful tools. Stop assuming everyone wants to get anywhere near the JS ecosystem: I’ve gone 30 years without touching it, and plan to continue that streak. Rewriting stuff is great from my perspective.
Distributed memory parallel computing library. Most common in supercomputers with a high performance interconnect where you want to run one program that coordinates separate instances of itself via message passing (“SPMD” - single program multiple data parallelism). Slightly different than distributed computing that generally doesn’t assume a tightly coupled system with a low latency, high bandwidth interconnect. Allows you to scale up beyond what you can do within shared memory spaces where you are limited to low core counts (hundreds) and low memory (<1TB) - MPI is what you use when you’re in the many thousands of cores, many TB of RAM regime. In practice, supercomputers require a mixture: shared memory for many core parallelism, some hybrid host programming model for bringing accelerators in, and then MPI to coordinate between these shared memory/accelerator enabled hosts.
I picked up a Hidizs player on Amazon as my modern iPod equivalent. Biggest use case for me: I don’t want a phone or device that does stuff other than music. Some moments I want something that doesn’t tempt me with distractions, apps, or the internet in general. Plus, I’m not always online: flying or out in the middle of nowhere, I can’t rely on streaming. Lack of connectivity happens just frequently enough for me to be annoying. The UI on the Hidizs device is clunky, but it does the job.