1. 1

    This post especially resonates with me because I know nothing about deep learning but I’d like to. I’ve been searching for a reference sheet that explains “how do I construct a dataset and model to answer my question and why do I make these specific decisions.” I have yet to find good explanations about even simple things like “how many layers do I need to classify an image and why.”

    I have a hypothesis this has to do in part with the fact that deep learning is still a new and rapidly changing field, and that many of the people with substantial knowledge know better than to waste time documenting rationales that could easily be obsolete in under a year.

    Maybe I’m not looking hard enough, but even many simple flappy bird YouTube tutorials link repositories that don’t compile/run anymore due to the rapid change in the ecosystem. It’s relieving to hear a similar grievance voiced, but I don’t know what approach would best mitigate this problem.

    1. 10

      I wrote some about ad blocking over here a few months ago: http://daemonforums.org/showthread.php?t=11265 (username=Carpetsmoker); some relevant parts:

      Advertisements have been an important revenue source for the publishing industry for over 100 years, and I’m actually in favour of allowing ads that don’t have the tracking and such that most do; at its most basic this could just be an <img src="/ad.png">.

      It’s easy to just say “all ads bad”, but things turn out to be a bit more nuanced once you start building products that are very hard to monetize otherwise. For example, one product I worked on a bit last year is a better recipe site with some novel ideas, but … how do you monetize that without ads? Not so easy. I ended up shelving it and working on something else that’s easier to monetize (hopefully anyway…)

      There should probably be more transparency from AdBlock, but having advertisers pay some amount of money to vet their ads by AdBlock is not wholly unreasonable.

      [ … some replies snipped …]

      Yeah, I agree it’s complex; I have no easy answers either.

      Right now my website has some ads from codefund.io – which explicitly advertises itself as ethical ads and is fully open source – which are blocked by default by uBlock origin (and I believe also AdBlock). These adblocking tools do much more than just block requests to third-party data-collectors, they frob with the HTML in an attempt to remove every single last ad, and while there is some merit to this for the really intrusive/annoying ones, I feel this is fundamentally a misguided approach which doesn’t consider the perspective of publishers/product makers.

      I use uBlock origin too, for all the same reasons as you do, but I don’t really like it because of this. I’m currently working full-time on open source stuff, and the income from these ads isn’t a bit of pocket-money on the side, it’s part of my income. Removing these kind of ads kind of rubs me the wrong way. Basically, I’m “collateral damage”.

      I maintained a “track blocker” (trackwall) for a while, which I think is much more reasonable than an “ad blocker”.

      The current situation sucks for both users and publishers :-(

      1. 13

        The current situation sucks for both users and publishers :-(

        I completely agree. Today’s situation is a perfect example of a defect-defect equilibrium.

        From the advertisers’ standpoint: if users don’t have ad blockers, then they’re better off if they act badly (you save the money that would be needed to vet ads), but if the users do have ad blockers, they’re still better off acting badly (milk whoever remains for all they’re worth). As for the end users, it hardly matters what the ad company does, because blocking ads is always better than not doing so.

        I still blame the ads industry for being greedy. Setting up ad blocking takes time, and while it’s hard to give it up once you have it, few would go through the time investment unless things were really bad.

        1. 12

          It’s easy to just say “all ads bad”, but things turn out to be a bit more nuanced once you start building products that are very hard to monetize otherwise.

          IMHO, “products that are very hard to monetize otherwise” sounds a lot like “products that people don’t think it’s worth paying for”. I’m very much into things that people aren’t usually willing to pay for myself, so I sympathize, but not everything that you or me like is also a good business idea. If the only way to monetize something is by having a third-party siphon other people’s data, that probably means it’s just not a great commercial venture. Lots of things aren’t, it’s just how it is.

          1. 10

            I think it’s a little bit more complicated than that. There are an untold number of cooking books, websites, TV shows, etc. Almost everyone cooks at least occasionally, and many do it regularly. It’s a big market, and these kind of products (including sites) bring a lot of value to people. Cooking is not even an especially big hobby of mine, it’s just something I noticed where improvements can be made, and I have a few ideas on it.

            I think it’s more a “race to the bottom” kind of thing. It’s just really hard to compete against “free”: given two websites which are equal where one is €5/month subscription and the other is “free”, then many (probably most) will use the “free” one (often with adblock), leaving me with no option to offer a “free” option as well. Even with a great product, you really need a clear edge to compete against that.

            Or to give an analogy: if I set up a stand with “free cooking books” next to the book store, then I’m fairly sure the book store’s sales of cooking books will drop.

            1. 6

              First off – I apologize if I was/am harsh or unfair. I understand this is probably important to you (and as the “happy owner” of several entirely unmonetizable – and, worse, probably expensive – passions, I know how it can feel). I’m not trying to discourage you or anything – the tl;dr of these posts is “here’s why x64k doesn’t wanna pay for recipes”, not “here’s why your idea sucks and you suck”. The fact that I’m unwilling to pay for recipes may well put me so much outside your target audience that if I were to think your idea sucks (which FWIW I don’t, since I don’t know too much about it in the first place), it might actually mean it’s great :-D.

              So, with that in mind:

              There are an untold number of cooking books, websites, TV shows, etc. Almost everyone cooks at least occasionally, and many do it regularly. It’s a big market, and these kind of products (including sites) bring a lot of value to people. Cooking is not even an especially big hobby of mine, it’s just something I noticed where improvements can be made, and I have a few ideas on it. I think it’s more a “race to the bottom” kind of thing. It’s just really hard to compete against “free”

              It is – but then again, “free” is what recipes have been since practically forever. It’s not like websites with free recipes have “disrupted the recipe market”. If all free recipe websites were to close tomorrow, and only subscription-based websites were to exist, I still wouldn’t pay a dime for these, since all I need to do in order to find out how to make pastitsio (random example ‘cause that’s what I had for lunch and it’s not from my part of the world) is ask someone who knows how to make pastitsio. I learned how to cook dozens, if not hundreds of things from family and friends. Some of them from far, far away, which is how I can cook Carribean or Japanese dishes in the middle of Europe (when I can get the ingredients, that is). I’ve never paid a dime for any of these things, I just asked nicely when I was abroad, or when someone from abroad was here, and did a few hilarious screw-ups until I got it right.

              Of course it’s a race to the bottom (the bottom being “free”) when learning how to cook from friends and family is how virtually everyone who’s not a professional in the field learned has done it since practically forever, and when this is still available.

              1. 3

                I didn’t take it as either harsh or unfair, so no worries :-)

                Recipes are not even eligible for copyright, but it’s not the recipes as such that you charge for but rather the making available of them. There are about 18 million cooking books sold in the US every year, so there’s definitely a market.

                I personally wouldn’t pay for a recipe website either, as I’m too casual of a user. Then again, I have bought cooking books, which are pretty much the same. Kinda funny how that works 🤔 I think this is a general problem with the internet: if every “casual user” would pay a low (<$1, depending on usage) amount of money then there’s no need for ads – that’s basically what ads are right? – but there’s no real way to do that AFAIK. I know Brave has gotten a lot of flak here, but I really like it for at least trying to solve this problem (although I’m not really sure about their whole cryptocurrency approach; but that’s a different discussion).

                I’m not angry, bitter, or resentful about any of this, by the way, they’re just observations. I wrote this as a hobby thing early last year after I quit my job after a bit of a burnout and it was the first time I had fun programming in a long time. In a way it’s probably the best thing I’ve ever worked on; it was very therapeutic. After I decided I wanted to work on my own product I looked at my options with what I already had, and came to the conclusion this probably wasn’t the best way to earn a living so shelved it for the time being. I still want to get back to it as a hobby thing once I have some time (and think of a name for it!)

                If you look at the existing big recipe websites then almost all of them suck: they’re shitty “recipe dumping grounds” with untold popups and all sorts of bullshit, and in a perfect world someone providing a better product shouldn’t have a hard time making an earning from it without crapping their product up with a zillion ads.

                I think the best thing we (as programmers/community) can do here is to make good alternatives available, which could be a “sane ad network” or perhaps something entirely different. I know that’s easier said than done but people do things to solve real problems they have (the same applies to many other things, and many people are working on this, with varying degrees of success).

                1. 2

                  Cookbooks are a pretty decent example of how recipe-centered sites can monetize, though. Binging with Babbish on YouTube is always pushing his cookbook, along with whatever product has sponsored his video. If people want to feel like they’re getting something that took effort to make (like a physical book) in return for their $20, then that’s how the recipe space will be monetized. Patreon is also gaining significant ground in monetizing quality content made available for free.

                  At the end of the day, people seem to recognize that the marginal cost of serving information on a website to an additional person is near zero and they balk at paying (subscription or not) for something digital unless it is perceived to be expensive to provide (like music, movies, internet itself, etc.).

                  1. 1

                    If we’re being honest, most recipe books suck, too – I guess it’s easier to pay for them because they’re a physical item and people are used to paying for books in general. But by and large, I discovered that even expensive books with fancy endorsements turn out to be… well, not wrong, these are recipes after all, but somewhat untrue to their topic. For example, I’m yet to see an English-language book that recommends anything even close to an useful amount of olive oil for recipes around the Mediterranean (or at least the ones I’m familiar with).

                    What I really wish we had, especially now that the world is much better connected and Internet access is universal, is something akin to rec.food.recipes, alt.gourmand and rec.food.historic :).

                    Edit: FWIW, I kindda like the idea of what Brave did. Back when rec.food.recipes was still a thing, I didn’t really have a problem with ads. I used some ad-supported software, including Opera (which I still miss so much), and I thought it was okay. I didn’t really mind the large banners, either, some of them were actually pretty neat..

                    1. 1

                      I suppose Sturgeon’s law (“90% of everything is crap”) applies 😅

                      I was a long-time Opera user as well, and used the ad-supported version back in the day. That was a different time though, before the ad-tech industry became what it is today. I’d be more hesitant to use a similar model today. And I also still miss that browser :-( Pretty much everything about it was fantastic.

              2. 5

                Yeah, “ideas” aren’t “businesses”.

              3. 4

                at its most basic this could just be an <img src="/ad.png">.

                Part of the problem is that even this type of innocuous-looking “non-tracking” ad can, without much effort, be turned into a tracker. It doesn’t take that many ads like this one to be able to generate and track a unique user identifier via HSTS, for example.

                1. 7

                  The amount of information it can track is quite limited though, compared to canvas/font tricks and whatnot. It’s also much more in transparent and in your control on what you send. It would still me a massive improvement over the current situation.

                  You can use src=data:base64 too.

                2. 3

                  Maybe if publishers had real businesses that deserved to make money because they actually provide a good service to the community then they wouldn’t have this problem! :)

                1. 11

                  I think I’ve read this post before…

                  1. 1

                    Yep, we all did. And the ending so good:

                    “… If you’re reading this and are interested in $HYPED_TECHNOLOGY like we are, we are hiring! Be sure to check out our jobs page, where there will be zero positions related to $FLASHY_LANGUAGE …”

                  1. 1

                    It’s interesting to see Google’s all-encompassing monolith slowly start to show cracks. I wonder what web players will do when Google is no longer the de-facto provider for fonts, captchas, analytics, online word processing, email, web browsers, etc. and instead the best services come from a variety of companies that don’t integrate as cleanly.

                    1. 1

                      I personally like google fonts, but I find they load faster when I host them myself. I use font squirrel to convert regular font files to a webfont format, and then I self host that.

                      Thanks to arp242 https://lobste.rs/u/arp242 for creating a lovely 3rd party web analytics solution, I use goat analytics instead of google analytics, which was my main gripe with google’s monolith.

                      Google’s recaptcha is horrendously arduous though, doing recaptcha after recaptcha makes me so angry. The method that recaptcha uses to detect bots is asinine, any kind of minigame could be so much more fun for users.

                    1. 20

                      That was an amazing read and I am in awe of byuu’s skill and dedication.

                      1. 35

                        Thank you so much ^-^

                        1. 12

                          Man this community is so cool because authors regularly pop up in the comments about their own software.

                          People like you are hard to come by though, you’re awesome.

                          1. 8

                            Thank you so much for your work, I love your take on it as your life’s work. These games will still be fun to play 100 years from now.

                            I remember reading a year ago about some improvements you (or others?) made in upscaling bsnes to HD displays and still need to update my RasPi to pull in a recent version with those changes. My brain hurts every time I see my kid playing Mario Kart with the terrible shimmering!

                            https://arstechnica.com/gaming/2019/04/hd-emulation-mod-makes-mode-7-snes-games-look-like-new/

                            1. 1

                              It doesn’t look like RetroPie has pulled in the latest bsnes. In fact, development of RetroPie seems to have halted a year ago… :-( https://github.com/libretro/bsnes-libretro

                              1. 2

                                https://github.com/libretro/bsnes-libretro

                                That fork is not intended to be current, and even changed its library name to bsnes2014. Upstream bsnes supports libretro by itself now, which is probably the reason that fork is semi-abandoned.

                                development of RetroPie seems to have halted a year ago

                                RetroPie/EmulationStation last commit was yesterday, RetroPie-Setup two days ago??

                                1. 2

                                  Thanks for the pointers but commits don’t matter if they aren’t shipping to users. I’m referencing https://retropie.org.uk/news/ where the latest release announcement was 9 months ago and the forum also doesn’t seem to have threads from 2020.

                                  I’m going through the steps to install Lakka instead, which appears to have modern bsnes support baked in.

                                  1. 1

                                    Retropie has an update menu in the UI, so these commits are making it to users who run updates.