2,593 karma · joined September 29, 2015
Readability is dependent on screen brightness and difference in contrast between background and foreground colors. These decisions aren't based on what engineers assume matters, it's dependent mostly on A/B testing and perception psychology. It's the same kind of research that went into making the OPUS audio codec and deciding which parts people will notice missing and which parts don't really make a difference. Tiny to moderate changes in visibility/readability make a much smaller difference than you imagine.
https://www.colorblindnesstest.org/farnsworth-munsell-100-hu...
https://www.hdnumerique.com/dossiers/962-test-4k-ultra-hd-bl...
There is an endless list of poor Bluray and UHD Bluray transfers where the the film grain makes it unwatchable. The unfortunate thing is that the nicer the display is, the sharper the image and the more distracting the grain. In some cases, I would trust the AV1 endcoder decoder to do a better job at adding a more reasonable amount of grain than the 4k transfer did. It's not a slippery slope, bad transfers and poorly color graded films/shows already exist. Mistakes exist in both directions, the UHD remaster release of Terminator 2 is a exceptionally horrible example of too much DNR. In my opinion T2 is bad enough to bring real actors into the uncanny valley.
The developers aren't doing anything different from what films do when going through an analog to digital transfer.
[1] https://www.hdnumerique.com/dossiers/962-test-4k-ultra-hd-bl...
The meaning of art is up to the viewer, not the artist. There's nothing wrong with having a preference, like using things like temporal denoising or increasing/decreasing saturation. Even an IMAX movie theater isn't enough because that's not what the film was graded with. Pretending that art shouldn't be modified because it moves away from the artist's intent is a bad faith argument. There were certain House of Dragon color grading decision that made me feel very disappointed, and I was watching in Dolby Vision on a nice high dynamic range TV.
Artists' intent is useless if 99.999% of the population can't enjoy the actual art.
https://www.vulture.com/article/why-house-of-the-dragon-epis...
This reminds me of Steve Yedlin's, cinematographer on everything from Brick to Star Wars: The Last Jedi to Knives Out and its 2022 sequel, Glass Onion: A Knives Out Mystery, who showed through research that film can be completely emulated using digital techniques and mixed digital and film shots during Star Wars.
https://www.audiocheck.net/audiotests_dithering.php
https://www.polygon.com/2020/2/6/21125680/film-vs-digital-de...
https://www.yedlin.net/NerdyFilmTechStuff/MatchLensBlur.html
I dislike how much more political this debate has gotten. It's now impossible to defend one "side" without having a lot assumptions made about you, and often get anger aggressively targeted towards you. Very little online discourse seems to be happening in good faith.
There's a bunch of people exploiting the fine tuning to make models based on popular artists who explicitly don't want their art used in training sets. There are also a bunch of artists and non-researchers spreading misunderstandings about how art generation works and targeting researchers and AI artists very personally on places like Instagram and Twitter. I understand the anger people feel from both sides. But there's a very clear pattern of assuming people on each "side" as equally harmful as the most problematic on both "sides".
I think having a deeper understanding of the technology and it's limitations makes it clear that the best art is made by already trained artists. There's things that are generated wrong, like shadows in complex scenes, that everyone can recognize but only artists have the knowledge to recognize and fix properly. Like most things, the optimal solution is a hybrid approach, like using generated art as inspiration or mockups.
Midi instrument libraries exist but movie composers like Hans Zimmer still hire out philharmonic bands to record the score. I can't tell the difference but composers working at that level can. But they also have music banks of recorded music they rely on for certain instruments. But it's clear they will always prioritize live recording. Art is no different, in my opinion.
AI can generate an idea but it will not generate art in the exact way you want or imagine. Some situations allow you to get away with good enough if you don't look closely, but that's rarely the point of making art in the first place.
Over the last couple of years, YouTube managed to get top Twitch creators to switch with contracts of above 1 million dollars. Some leaked ad spots, like watching a trailer on stream or playing a game for 30-60 minutes, were between 2k and 10k USD for a streamer with around 8,000 average viewers per stream. There was a leak recently of how much top streamers on twitch made, and there were multiple 500K+ salaries. But the platform unfortunately follows the 1% of creators earn 90%+ of all income rule of thumb.
For local playback of media files, I use daum kakao Potplayer, I have never come across something more power user friendly. MPV is weaker in a lot of areas, in my opinion. Potplayer interfaces with everything I'd want including vaporsynth and madvr and it has options for subtitle resync based on a multiplier and also converting based on framerate 60 -> 30. There is even an option for live subtitle translations. I personally like using it with smooth video project, but I understand it's not everyone's taste.
Are there situations outside a power draw limits where choosing a Pi makes more sense than a used NUC or business mini PC at a comparable price?
The traditional TV industry is continuing to fail to adapt to the different way in which people watch content. Youtube demonstrates really well how people are no longer loyal to a network, they are loyal to a creator or topic. There have been certain groups of channels that sort of remind me of an old school TV channel, like MatPat and his youtube channels Game Theorists, Film Theorists, and Food Theorists could have been a TV channel in combination with some other video game or theory crafting channels but people don't search by a bigger topic. People find content based on recommendations, either algorithmic or popular things watched by friends. Netflix Originals and Amazon Originals seem better, but I can't understand the metrics Netflix uses when deciding what to cancel.
As someone who has shifted to online media as the main form of content I watch, I can't tell you what makes a cartoon network show a cartoon network show. I know because of the logos between shows and during ads, but I struggle to find a strong underlying identity.
The easier access to international media has also caused an oversaturation of available good media. I no longer need to worry about what part of the world a show is made in, if it's popular enough, you'll be able to find it sailing the seas. HBO Max, the supposed online equivalent to the cable channel HBO will drop Westworld from its online catalog. Some fantastic Disney+ Marvel shows, like the critically acclaimed Loki, have no announced plans of getting a physical release and Loki released back in Jun 2021. When a TV show enthusiast sees those two pieces of news back to back, a normal conclusion is that the only "sure" way to own shows is piracy.
Students will always feel some incentive to cheat, even if they wouldn't in isolation, they could feel pressured based on their peers. A strong curriculum isn't enough to encourage every single student not to take shortcuts. Despite all external circumstances, it's still an active choice to cheat. Cheating is framed one-way because it's generally discussed imagining the laziest and most adversarial student possible, similar to how you'd red team for security testing.
I think the concern is things like ChatGPT will lead to a lot more students cheating. I think it's human nature to want to take shortcuts. I agree with you that the way in which cheating is demonized in a school environment is counterproductive in the long run. I think admin and teachers have hid behind students' fear of academic integrity violations being reported on a college application to not have to admit that most are terrible at catching cheaters. It's from the perspective of imagining ignorant teachers and adversarial students. Most minor cases of cheating aren't caught, even at a college level.
I have to imagine Google HAS to dedup on the server side, otherwise even a 100,000 song cap would be too much data to host for free. Google doesn't play ads on uploaded tracks. Google stores not-transcoded music files and has since "Google Music". When GPM first came out, it was a storage locker and felt more like direct competition to Grooveshark than iTunes.
Record companies seem to have a very tight not just on the songs, but also the metadata surrounding them. No service really offers the "radio mode" or suggested tracks for local playback, from what I found. I'd happily give a company playback metadata if in return I got that feature back. I wonder if Pandora has patents for certain parts of music recommendation algorithms. Spotify and YTM can rely heavily on user-generated music connections, but maybe that is why it can't work for local/uploaded files.
People that care enough to keep a library of music after the popularity of Spotify are the same people to leave the ecosystem permanently when this kind of stuff happens.
Google Play Music was the last true third party hosted solution, in my opinion. YouTube Music allows you to upload files but segregates your uploads from the YouTube Library for recommendations and smart playlists. I haven't found a great "smart recommendation" service for self hosted files.
I have enjoyed self hosted solutions since YouTube Music broke the smart Playlist and integration for uploaded files with steaming files.
I know that the Nvidia Shield TV Pro is a flagship product and highly recommended in home media server circles. It's not hard to imagine Nvidia shifting their focus to that product and sunsetting a feature that doesn't overlap a lot with that crowd.
Nvidia Shield TV Pro is one of the only devices I know of that can play back Dolby Vision files properly, at least via Plex/Kodi. The specific DV profiles supported is a work in progress but the most common webrips are detected properly and enable Dolby Vision mode on the TV. BD Rips aren't completely supported by any device other than disk players, but non DV fallback usually works fine. Chinese TV boxes have unofficial DV support but certain surround codec support can be spotty and support could mean simply discarding the metadata. Kodi added support recently on android builds but DV needs hardware level support as well for true playback.
The distinction you are making doesn't make sense. Are you making an analogy to the idea that you can use the word fire but you can't yell fire in a crowded place? Facebook doesn't choose individually what posts are highlighted, the engagement of other users does. The complete weights of what is highlighted are also unique per user. You're arguing that Facebook has a duty of care, but as it stands section 230 already addresses that. The duty of care is fulfilled with good faith moderation. The only way Facebook can be found liable is if section 230 is weakened and good faith moderation is redefined.
"Tort liability indicates that someone is held accountable for wrong actions (other than under contract." Tort liability is the phrase you would search if you wanted to learn about previous legal rulings related to this. I thought the Columbia Law Review article I linked was pretty approachable. It's not possible to talk about this in a meaningful way without understanding tort. I haven't tried to obscure or distract from my position intentionally.
I think Force v. Facebook was the proper ruling. There's nothing new courts have to decide, because section 230 already addresses this. Fundamentally, people are hateful and will engage and interact with hateful content. It doesn't make sense to blame an algorithm for not realizing a post is hateful and highlighting it. I made my position clear without technical jargon in the top comment. The DW article says the lawsuit wants to legally force Facebook to change its algorithm. The change would directly be related to moderation. This would be a slippery slope. Failing to understand the limitation of what an algorithm does is the only way you could expect a separation of the concept of posting content vs posting it "like that". I hope I was able to clarify my position.
https://www.nytimes.com/2021/03/24/technology/section-230-he...
I don't think there's a way to allow recovery based on only information because then recovery become the primary method to take control of your account. Once you start adding limits to recovery attempts, you end up back where we already were before 2FA become popular. And what's stored in your head becomes a very easy target, especially for those less tech literate.
I really liked Facebook's idea of trusted contacts based recovery. It would allow a complete true loss of everything based on who and what you know and doesn't rely on a company having any information other than who your friends are. Which Facebook already knew since you were friends with them. Facebook's single identity model made that easy but for Gmail or Instagram it's common to have multiple that you don't check often.
I am not sure where friction exists for an audience of, presumably, tech literate people deploying TOTP 2FA. I understand the concern in a more diverse group, but I'd assume developers have a better understanding of security.
I can understand reservations about using a non open source service, but with apps like Aegis on Android and Ravio OTP on iOS. I have an encrypted backup of my secrets synced from my Android phone to Dropbox and Google severs. The encryption is based on a password I chose. I lost my device and restoring my 2FA was as smooth as I expected.
I am very annoyed by Google constant push to try to get me to start using my phone like a WebAuthn passkey and it's pushing me further and further from the technology. I don't believe security exists if I don't control my secrets. I hope they shift their approach to the same as what Github's currently is. Google nags users to login with the passkey every once in a while if it's registered, regardless of what primary 2FA you want to use.
"Authentication with a security key is secondary to authentication with a TOTP application or a text message. If you lose your security key, you'll still be able to use your phone's code to sign in."
https://docs.github.com/en/authentication/securing-your-acco...
Like and Dislikes are only as valuable as users allow them to be. A great counterexample to many of your arguments is the social media site we are discussing this topic on. The culture on Hacker News de-emphasizes Likes and Dislike. Comment karma is visible to the poster, but to a reader, a +100 karma comment is equal to a +4 karma comment with the only difference being what is seen first. I personally forget about that and end up reading comments to the end of threads I enjoy anyway.
I think letting such hostility build for the concept of social media is as damaging as any political extremism the censorship is supposed to combat. It feels like the extremism isn't gone, just redirected. Social media and the internet has changed who people interact with, not just how. In the past extreme ideas existed, but were tempered because of others around them. Social media has allowed people to choose to only listen to people with the same perspective, letting extreme ideas go uncontrolled. You can argue that it's the fault of the technology, but the main responsibility falls on the individuals.
Social media doesn't make people hateful, it makes it easier for hateful people to convince other to think like them. Bullies have existed long before social media and continue to exist in every context.
We fundamentally disagree on whether or not what Facebook "should" have expected or "should" have done is worth considering. I don't think it's legally relevant. Profits can demonstrate incentive that otherwise wouldn't exist.
The existing legal precedent for how algorithms are considered under tort liability [1] are still unclear. It's clear to me that allowing this lawsuit to move forward would be directly contradicting the ruling of Force v. Facebook (2019) and would force a ruling on how AI algorithms should be held liable with respect to section 230.
From the perspective of the user, making a platform liable would mean that the platform would fundamentally shift what users are allowed to post and strongly weakens basic rights to freedom of expression. It's hard to depict how different and more censored the internet would be if section 230 continues to become weakened.
"Critically, the finding that AI algorithms can be found to be harmful allows for the tort liability process to come into effect. While this kind of legal approach has not yet been taken, Patel indicates that corporate liability is a possibility. If an AI algorithm causes harm, it must be prosecuted through tort liability."
"However, in the same timeframe as Patel’s ruling in 2019, ... a group of U.S. citizens who were victims of a Hamas terrorist attack, sued Facebook, arguing that Facebook implicitly assisted Hamas by keeping their violence-promoting content online through their AI algorithm’s oversight. [12] However, in accordance with 42 U.S.C. 230(c)(1), the rule that protects social media platforms from liability for content on their platforms, the court found that Facebook could not be found to have criminal or civil liability for content. [13] This case complicates the implications of Patel, as it found that Facebook was shielded from liability because the content itself was not created by the platform. "
[1] https://www.culawreview.org/journal/extending-corporate-tort...