anon9001

I'm shocked at how many people seem to be doing OK or have minor complaints.

For me, I've realized that if I have to do anything for 40+ hours a week on a schedule, after a couple of weeks, I'd prefer being dead.

I could be a professional ice cream taste tester or race car driver, and after about 3 weeks of showing up at a fixed time and putting in 8 hours, I'd still be ready to off myself again.

When I have week-long breaks from work (like holidays mixed with PTO), I regain the will to live and there's so much I want to do, and I'm so sad about how much time I've wasted. I start projects, I start forming relationships, I feel happiness. Then I go back to work, and the suicidal tendencies are way stronger than before I left.

The best way for me to survive seems to never allow myself time off of work, because I dare not remind myself of what a free life would be like, or I might not be able to endure my current circumstances for another day.

I just don't understand how people function as workers in society. I hope I can retire before this kills me.

It probably doesn't even matter for AI research. I don't know any professional AI researchers, but I bet reading arxiv and reproducing other's work, and publishing your own work is how you go about that. Why would you do coursework when there's papers to read?

> Yep, everything newer is automatically better.

I didn't say newer is better, I'm saying that your recollection of how a compiler worked in university will not be of any use to you if you need to write your own language for production today. I've also never heard anyone say "oh yeah, I did this once in college, here's how you do that".

> For those people who do need to know, their work wouldn't be possible without it.

Who? Seriously, who would ever be employed to author search algorithm using the knowledge they obtained in a 4 year computer science program? Extremely few people are involved in the work of implementing anything that low level, and thinking you understand what's going on under the hood is just tricking yourself. For example, your 4-year CS degree holder would surely understand the trade-offs between a linked list and an array, right? Then you find out none of that is true in practice: https://dzone.com/articles/performance-of-array-vs-linked-li...

> I personally switched from pre-med to CS as a junior in college with only a single QBasic high school class under my belt.

I'm not saying only the l33t can do CS, I'm just giving the example that if you actually do know what's going on and you have to take all those classes, it's very obvious that the professors have nothing of value to add that you can't learn faster by yourself online. I grew up cracking windows software for fun and I had to watch a professor stumble through the basics of x86 assembly. It was an obvious waste of time for all parties. Educational resources today are vast, and if people want to learn something, they can just go learn it.

Also, have you looked at professor salaries vs engineer salaries? I know there are some people who teach at night or do it for the passion of it, but the reality is that it's not going to attract the most ambitious minds in our society. Meanwhile, so many major pieces of software are available for free online, and you can actually talk to the teams doing the work, and they'll let you contribute and give you feedback... for free!

> Finally, as others have pointed out, (and as much as I hate admitting to this phrase that was so often spouted out by my college professors, perhaps I have drank the Kool-Aid), you're learning how to learn. That is, you're practicing juggling abstract concepts in your head, making connections and weighing solutions.

This is my primary objection, and perhaps you have drank the Kool-Aid. If you get a CS degree at any major institution, you are not learning how to learn. You're both preventing learning and picking up bad learning habits. Doing a pre-built lab in a CS course is so much worse than contributing to literally anything on github. Studying how algorithms used to work in the 70s is useless compared to diving into any modern piece of code and benchmarking and learning how to optimize.

Look at what is actually in a CS program from a respected school: https://cse.engin.umich.edu/wp-content/uploads/sites/3/2019/...

There's a 4th-year course just called "Algorithms", check it out: http://www.eecs.umich.edu/courses/eecs477/f02/syl.html

They might as well have called it "Inefficient implementations of already solved problems".

I can't imagine why there's a "Databases" class and "Web Databases" class, but I think you see my point. Nobody is implementing a database with anything they learned in uni, and nobody is doing a better job of learning how to use a database in school than they would with experimentation and online explanation.

Maybe the fact that you're graded on learning these things, coupled with the enormous price, drives a student to independently research each of the topics presented throughout your 4-years? Then through research and play, they would gain understanding of each topic. But if the school is simply serving as an extremely expensive prompt for self-learning, then what value is the school really adding?

> I found my CS classes were 10x better at making me a general problem solver than my pre-med classes (mostly my discrete math/logic classes)

Every time I hear that reasoning, I think it's a rationalization for spending huge amounts of time and money doing something pointless. I also like "the college experience" as a good reason it was valuable.

Maybe general problem solving is being taught and is valuable, but that's not what these CS degrees are being sold as. To pick on umich again, look at the "Student outcomes": https://cse.engin.umich.edu/academics/undergraduate/computer...

So if you graduate with this program, you'll be able to "Analyze a complex computing problem and to apply principles of computing and other relevant disciplines to identify solutions" and "Apply computer science theory and software development fundamentals to produce computing-based solutions"? Maybe if you study independently while also stressing about passing your exams and doing your homework, then you'll be able to do those things, but it'll happen at a slower rate than if you just started building software and researching as you went. I just looked at a few of the syllabuses for this program as an example, and there's no way it's going to deliver what they're promising.

What exactly is the purpose of a CS degree?

If you go into your CS degree as a young hacker, adept at *nix with a penchant for assembly language, with enough experience to appreciate the CS concepts, you're going to be disappointed. There's a strong chance that you'll know more than the professors.

Alternatively, if you go in knowing nothing and really work hard to sponge up everything the degree program wants to teach you, you're going to think you've learned something but you won't actually leave with any useful knowledge. You'll have dated and often half-baked understandings about how computer systems should theoretically be designed. You might have an idea how a basic OS kernel works from your OS class, or how to write a sorting algo from your data structures class, or maybe a parser from a language design class. None of that is useful at all, other than maybe to get you interacting with your machine, in the hopes that you'll learn how to use a computer (which is a completely different set of knowledge) by the time you graduate. And if you did ever want to write your own programming language, you can throw out everything you learned in class and start over by Googling it and following the best practices of today, like we do for everything else.

The only point I can see is to be able to say you have a CS degree, and hopefully that's losing value.

If you like this, you should also check out the work around Retro City Rampage. RCR is a GTA1 clone for multiple platforms from a few years back, but the developer also made a real NES ROM of it. Here's a great talk about that process: https://www.youtube.com/watch?v=Hvx4xXhZMrU

(I'm not related to the project, but it's one of the few games I've 100% completed because it was just so good)

I'm sure there are some quants reading this thread, so here's a real question: Can somebody explain the technical pipeline from tweet->execution?

I assume there's some low latency twitter API that you can pay big money to access? And then I guess that goes to some array of GPUs to run the NLP model as quickly as possible? And the NLP results get traded on by placing orders with an exchange?

Are there companies that roll up many low latency data feeds for purposes like this?

At minimum, those substances would be very useful in a therapy setting with a trusted therapist.

They might also be useful on their own merits, as in, if you have severe depression and eat a bunch of mushrooms maybe you won't be so depressed anymore. Given the current state of anti-depressants, it certainly seems worth a shot.

Obviously anything experiential can be dismissed as "meaningless woo", but as someone who's done a fair amount of LSD and is a lifelong skeptic, it's an unmistakable experience that the hallucinations are taking place in a higher ordered space.

To add a little clarity to what the hallucinations look like, checkout what people describe here: https://psychonautwiki.org/wiki/Geometry

The process of hallucinating on LSD feels like you're sliding a window through a higher dimensional structure projected onto a 2D plane. LSD hallucinations aren't "3D" in the same way that mushroom hallucinations are. LSD feels like infinite 2D planes stacked on top of each other, rather than a 3D, if that makes any sense.

If you pause to meditate and quiet your mind, you can get the hallucinations to briefly stall or at least retain their current character. If you focus on one area and let your mind explore the hallucinating you're looking at and just sort of "let it happen", or perhaps change music/environments/stimulus, you can push the hallucinations in a new direction.

The 2D nature of the hallucinations and the lens-like focusing mechanic does give the impression that you're viewing a projection of a higher dimensional space.

I understand it sounds like woo, but the experience is remarkably similar to the feeling of playing with something like http://4dtoys.com/ and I can very much understand what mirimir is describing.

I don't have any science-based understanding of how the brain works, but maybe I've added enough color that you can understand what the experience feels like.

It would be fun to check, but the attackers could just turn it off.

It's a user-facing option: https://support.apple.com/en-us/HT202100

I'm not 100% sure I understand your comment, but I don't think it's dishonest to frame drug use as a crutch.

One of the odd things about drugs is that if you do use them as a crutch (and I certainly do), then you are forfeiting recreational value.

I smoke to feel "normal" and to avoid nightmares (it turns off dreaming, which is quite useful). That means I can't get "high" like non-potheads can. When I haven't smoked in a while, I get giggly and have the munchies and every idea is mindblowing, and that's truthfully quite a lot of fun and is the reason most people start smoking weed.

Unfortunately, if I don't smoke after work, I can't really feel like my work day is over, and I'll keep thinking about work until I exhaust myself. If I don't smoke before bed, there's a good chance I'll have horrible nightmares and wake up in a cold sweat. These aren't symptoms of withdrawal that go away in time, that's just how I lived my life before discovering weed, and that's how I feel daily even if I take a months-long break from smoking.

I would love to not need weed as a crutch, both to be healthier and so I can occasionally get a giggly-munchies-omg-this-music-sounds-amazing type of high.

The problem is, if I quit smoking, I feel less mentally healthy, and all of the psychiatric drugs that could possibly help are worse in health and effect than weed.

Everyone knows (or should know) to be careful with edibles. I smoke a lot but they still scare me and I avoid them.

It's a subculture. Being a pothead is culturally very similar to being a gun nut. If you take a restricted or poorly understood item that provides enjoyment but is a bit taboo, and give people space to modify/grow/experiment/etc with it, it's spawns a subculture and people associate their identity with it.

It happens with hackers too, and porn, and weird nicotine vape culture, and I'm sure many other things.

I've had approximately the same experience. Here's hoping it doesn't cause any long term issues, but I think I would have had a heart attack by now if I didn't smoke to calm myself down and realize I shouldn't take things so seriously.

A truly enlightened individual doesn't need any drugs, but I'm not there yet. Maybe in time. For now I think the weed is a net positive.

3blue1brown is a special kind of awesome.

I'd love to filter the internet by "content that can give you insights right now that would otherwise take years of study in a specific discipline to even know exists".

Another that I saw recently on an HN comment (thanks tptacek) is https://www.youtube.com/watch?v=nfY0lrdXar8

A quick google for "3blue1brown awesome github" (my usual strategy for finding good similar content) landed me here: https://github.com/rossant/awesome-math

The awesome lists are pretty good starting points for finding good content, but there's a ton of them of variable quality, so you end up with stuff like this: https://github.com/jonatasbaldin/awesome-awesome-awesome (which has 68 forks...)

Here's a KMS server emulator, all open source: https://github.com/Wind4/vlmcsd

If you want to skip straight to the piracy, here's an open source package for you, complete with keys, hosted ironically on Microsoft servers, that uses vlmcsd: https://github.com/ekistece/vlmcsd-autokms

It's a pretty clever hack. It runs the KMS emulator locally and fakes the network connection with a TAP device from OpenVPN. Works perfectly.

That's a pretty bad example because any good programmer will have at least a rough idea how to profile code in their environment.

Can you name any software that's "engineered"? I can't. All software is built by craftsman of varying degrees. Nobody sits down and engineers anything up front.

The original point "agile" was that code is an iterative craft, so everybody should get together and work toward incremental improvements and respond to change, rather than try to engineer the whole project upfront. In other words, the way software is "supposed" to be built in most places these days is explicitly the opposite of engineering.

* Individuals and interactions over processes and tools

* Working software over comprehensive documentation

* Customer collaboration over contract negotiation

* Responding to change over following a plan

Doesn't sound like "engineering" to me, and that's the process with the most adoption (even though it's bastardized now, that's still the idea that's being sold).

> I believe that people deserve to be able to grocery shop without having to worry about letting someone down easy or dealing with someone who gets aggressive when rejected.

Delivery is so fast and ubiquitous that it makes me wonder if the only people actually going to the grocery store are people looking for social interaction.

Maybe we should have designated, opt-in, "social spaces" where human<->human interaction isn't considered a breach of norms? Or the other way, there could be designated areas where interacting with strangers is forbidden for your convenience.

When we finally get AR, I wouldn't be surprised if you could put yourself in "busy mode" IRL, so other AR-enabled people would know to leave you alone.

Webdev here that just started ML a couple of weeks ago. I tried reviewing all the math and concepts and it didn't help much. What really helped was signing up for Kaggle and working through example problems.

Here's what I think I understand so far, feel free to correct me if I'm wrong:

* Most ML solutions come from published papers, and even ones that are years old are still effective and relevant. You're going to be tweaking other people's designs and that's fine.

* Effectiveness of ML seems to come from two things: Network design improvements (some combination of brute forcing, guessing, understanding math, and implementing papers) and from better data (if you have access to more data sets, you can do better things).

* Before you start with the ML part, you need to first be able to do "Exploratory Data Analysis". Which really just means to learn matplotlib, which means you'll need to know how to use pandas, which means you'll need to learn how to use numpy. You want to be able to understand your data, find outliers, graph some examples, etc.

* There's a few different types of ML, but the ones that seem to solve problems are "reinforcement learning", "supervised learning", and "unsupervised learning". The supervised learning is the most common one to get started with, and it seems to work well if you have a big set of "X should produce Y" type of data. Unsupervised learning is similar but you don't have Y and you're looking to find groupings of your data instead of matching to specific labels. Reinforcement learning seems most useful for things like game AI, where you have some state and you have various moves you can do to try and increase your score (OpenAI Gym is really fun for this).

* It is actually unreasonably effective. There are "pre-trained models" that offer a starting point for networks that have already seen huge data sets. You can "transfer" that learning and then train on top of it. It's amazing how little effort is required to start classifying images.

* Keras backed by TensorFlow seems to have the best support all around, and the code is pretty easy to read. TensorFlow is a little like OpenGL where it has its own rules and state, and you should do some examples to see how it works, but it's very low level. Keras is like any other high level python library and it does almost all of the TF interaction for you.

* If you're doing it on your machine, use conda, because it manages all the python stuff for you. If you want the fastest way to get started online without your own GPU, use Kaggle. It has a maximum runtime of about 9 hours but that's plenty to get started.

Neil Young is an audiophile in the worst sense of the word. He's tried to make a player and music service https://en.wikipedia.org/wiki/Pono_(digital_music_service) and it didn't work because there isn't demand.

If there was demand, Spotify would just add FLAC. They already added 320kbps compressed streams, and I doubt anyone can hear the difference between that and FLAC.

The vast majority of the audio perception people are going to pick up on happens during production. And https://en.wikipedia.org/wiki/Loudness_war is why everything sounds worse than it used to. When records were being made for vinyl, the physical material wouldn't allow for the level of "loudness" that removes dynamic range from the track. That changed when all music started being produced digitally for digital distribution. Ironically, digital audio actually has a much larger dynamic range and should be better, but because of the way it's mastered it comes out louder but with less dynamic range.

Audio compression is the least of the problems for people interested in high quality recordings.

> Everything I do in social interactions, in order to fit in, is a learned behavior that I didn't pick up until my university years.

You might be improperly socialized or have Aspergers, or you might be a sociopath (I don't mean this in a derogatory way, some people are just not capable of emotion, often due to past trauma), or you might have anxiety, or it might be ADHD.

My advice would be to try more drugs to try and get a better perspective of why you feel the things that you feel. If it's ADHD, Ritalin or Adderall will cause a positive change. If it's anxiety, Xanax/Klonopin/Ativan/Valium might help. If it's an empathy thing, there aren't many legal pharmaceuticals for that, but you could try weed or MDMA or a traditional psychedelic. IMO the relatively low risk of trying a few chemicals a few times is well worth the gain in perspective, so you have a better sense of what you're trying to change and what levers can be pulled.

The average developer cannot pick up an entirely new language in a few days. Maybe I'm incompetent, but it takes me a few months before I start to feel comfortable with a new language and its ecosystem, and a year or two before I feel like I really understand the nuances and know just where to look when something goes wrong.

I can easily find someone with 5+ years of full time Java development (or JS, Python, and Ruby), and that means they're going to come fully equipped to dive into a problem, using familiar patterns and libraries. I don't want to hire people to make them learn Scala as quickly as possible and hope that they're able to figure it out very quickly.

> the average one chooses a major language and framework and hacks things together with stack overflow

Call me average, but I spend most of the day googling and reading results on SO. I think most developers do the same. Even if I already know how to do something, I usually google it to see if there's a better way to write it. If you're working with Scala (or some other less popular language), you're going to have less results doing these things, because less people are going to have hit your special problem. There is no amount of amazing language features that make this trade-off worth it.

And it's not just true for languages, it's true for frameworks too. Using common stuff (language choice, framework choice, library choice, compiler choice, etc) makes it so much easier to solve the problems you'll inevitably run into.

I personally believe that people who bring functional programming into commercial settings are wrong to do so. There's a real trend lately to go "big" and "functional", but know that it causes actual business problems.

If I have to inherit code, I'd take (JavaScript|Rails|PHP|Python)+(MySQL|PostgreSQL) over (Scala|Haskell|F#|Lisp)+(MongoDB|Cassandra|Kafka) any day.

I understand that all of these tools have their uses, but the most important thing is ease of finding decent developers. Lots of people can do JavaScript well, but few can do Scala, so you're going to have a smaller pool of talent. More popular languages are also going to have more developed patterns, more native libraries, and better tooling options.

If you need Scala, please just write Java. You'll always be able to hire a Java dev. Everyone hates writing Java, but everyone can write Java.

And related, if you're using any kind of NoSQL for your business data, you're wrong. I understand NoSQL for caching and document storage, but not everything is a document. Most things are actually best in SQL tables. Especially if you're a startup.

If you know how to use the tools but can't make anything that feels like music, you just have to keep copying songs until you understand how it works. It's just practice like anything else. There's a bunch of fantastic youtube channels that show the whole production process now too.

If you don't know how to use the tools, that's a whole world. There's a bunch of different camps, but if you're trying to do EDM, then Ableton Live or FL Studio are pretty popular. There's open source stuff too, but I don't know much about it. Some people prefer to use hardware instead. Then there's those weird modular synth nerds. It really depends what you want to do and how much you want to spend. A commercial package like FL Studio will come with a bunch of standard plugins that would get you pretty far before venturing into the world of virtual instruments.

There's a few books on the subject (Dance Music Manual comes to mind), but honestly I don't think they're very helpful. You can get extremely far without knowing any music theory, and it's hard to describe audio stuff in a book, so you'll probably get more out of youtube searches about whatever you're trying to do.

> For scent detection dog handlers, beliefs that scent is present might result in either sufficient inadvertent postural and facial cues so that dogs will respond regardless of the absence of scent

I don't know how good dogs are at pose detection, but it's definitely plausible that they can tell when an alert is expected.

Now do the yield curve!

Seriously though, that was a great explanation.

I'd rather ride with a stoned driver than a drunk driver any day. I'd also prefer a stoned driver over an elderly driver or a sleep deprived driver. And if I had the choice between a totally sober driver or a driver on stimulants, I'd take the stimulant driver.

If this was really about reducing risk on the road, we should somehow incentivize people like OP to work on the self-driving car problem, rather than come up with a device to catch stoned drivers. Think how much brain power and law enforcement time is going to be taken up dealing with stoned drivers. The rational thing to do would be to let people drive stoned, arrest those who are visibly impaired and can't perform a field sobriety test, and focus our efforts on getting AI to replace all human drivers.