I think a lot of people struggle to imagine the kinds of dirty-deeds ("ratf***ing") that are both possible and effective, especially when the perpetrators don't (feel) constrained by an implicit baseline of plausible consistency or morality. Being unable to brainstorm them up is, perhaps, a kind of backhanded compliment.
Imagine trying to warn someone in 2010 that in a few years an outgoing President, stung at an election loss, could foment a violent mob that would break into the Capitol to hunt and chase legislators that were formalizing that loss, issue blanket pardons for everyone involved, and his party would still protect him from being impeached over it.
For that matter, some people are still surprised to learn about the "Brooks Brothers Riot" [0] of 2000, where a crowd of Republican campaign staffers threatened workers into stopping a recount of certain ballots.
> You can't tell me with a straight face that all of the thousands of developers who develop these products/services care deeply about the quality of the product.
What about caring and being depressed because quality comes from systems rather than (just) individuals?
> We had the first 4+ years to learn that "malice or incompetence" is not the right question. There's been more than enough pathological input to show it becomes a denial-of-service attack on observers.
> The correct answer is both, until and unless the perpetrators wish to come forward and defend themselves as just malicious or just incompetent.
One might also view it as a kind politically-flavored nerd-sniping. [0] Sometimes the only winning move is not to play.
I'm reminded of the Xerox JBIG2 bug back in ~2013, where certain scan settings could silently replace numbers inside documents, and bad construction-plans were one of the cases that led to it being discovered. [0]
It wasn't overt OCR per se, end-user users weren't intending to convert pixels to characters or vice-versa.
JBIG2 does glyph binning, as you say not exactly OCR, but similar. So chunks of the image that look sufficiently similar get replaced with a reference to a single instance.
Glyph binning looks for any chunks in the image that are similar to eachother, regardless of what they are. Letters, eyeballs, pennies, triangles, etc without caring what it is. OCR looks specifically to try and identify characters (i.e. it starts with a knowledge of an alphabet, then looks for things in the image that look like those.
If the image is actually text, both of them can end up finding things. Binning will identify "these things look almost the same", while OCR will identify "these look like the letter M"
It's not too hard, while they share some mechanics, the underlying use-cases and requirements are very different.
_______ Optical character recognition:
1. You have a set of predefined patterns of interest which are well-known.
2. You're trying your best to find all occurrences of those patterns. If a letter appears only once, you still need to detect it.
3. You don't care much about visual similarity within a category. The letter "B" written in extremely different fonts is the same letter.
4. You care strongly about the boundaries between categories. For example, "B+" must resolve to two known characters in sequence.
5. You want to keep details of exactly where something was found, or at the least in what order they were found. You're creating a layer of new details, which may be added to the artifact.
_______ "Glyph compression":
1. You don't have a predefined set of patterns, the algorithm is probably trying to dynamically guess at patterns which are sufficiently similar and frequent.
2. Your aren't trying to find all occurrences, only sufficiently similar and common ones, to maximize compression. If a letter appears only once, it can be ignored.
3. You do care strongly about visual similarity within a category, you don't want to mix-n-match fonts.
4. You don't care about clear category lines, if "B+" becomes its own glyph, that's no problem.
5. You're discarding detail from the artifact, to make it smaller.
Jbig2 dynamically pulls reference chunks out of the image, which makes it more likely to have insufficient separation between the target shapes.
It also gives a false sense of security when it displays dirty pixels that still clearly show a specific digit, since you think you're basically looking at the original.
You asked what the difference was, and I said the difference. Was it unclear that to fit the phrasing of your question, we add "OCR doesn't"? I would not personally call Jbig2 OCR.
Let me try rephrasing to make the response to your original comment as clear as possible.
Question: "How can we describe OCR that wouldn't match this definition exactly?"
Answer: This definition largely fits OCR, but "reference to a single instance" is a weird way to phrase it. A better definition of OCR would include how it uses builtin knowledge of glyphs and text structure, unlike JBIG2 which looks for examples dynamically. And that difference in technique gives you a significant difference in the end results.
Is that better?
The definition you quoted is not an "exact" fit to OCR, it's a mildly misleading fit to OCR, and clearing up the misleading part makes it no longer fit both.
Right, that's actually a fair framing. I get to enjoy a walking commute in my city, and or the most part, I feel very anonymous on my walk into the office.
Blending into rush hour foot traffic is easy, and I never feel like I stand out enough to attract attention... though in the back of my head, I know that most commercial and government properties have some form of video surveillance, probably backed by some kind of (hopefully coarse) AI subject tagging.
That genie's out of the bottle, couple cheap cameras with cheap AI recognition and tracking, and every network getting pwned by somebody the result is, public spaces become obscenely public. The worst is private spaces, like when a technician or tradesman comes to one's house to do work wearing smart glasses. If they're Rx, what do you do? Refuse service? live with the broken pipe/modem?
No opinion on the decision from OP's post, just noting that, privacy is getting 86'd six ways to Sunday. All that's left are truly owned homes and natural spaces, and Orwell laid the pattern for both.
> That genie's out of the bottle [...] public spaces become obscenely public
A bit "meta" here, but this was never about surrendering to advancing capability, humans have been capable of eavesdropping ever since eaves were invented, people could bring hidden microphones and cameras in "private" spaces for decades, etc.
The "reasonable expectation" doesn't come from our fears of the worst, or else it would be meaninglessly permissive. Rather, it's an attempt to judge individual situations based on some kind of collective agreement about normality. Just because we can't preemptively block something doesn't mean it cannot be a crime.
In other words, we (should) have a lot more power to decide what they become than that. Supposing--and I'm not endorsing this--anyone caught recording in public typically gets their stuff smashed by an angry mob, then the "reasonable expectation" is that you're not being recorded.
Thanks for the nuance. I guess my issue is that pervasive recording has already been normalized, db hacks are effectively normalized (if illegal) so the public spaces are super public, as of yesterday not tomorrow.
IANAL, but for some months now I've been pondering what could happen if a site--like a personal blog--had a legally-strong "click-wrap" Terms of Service, which unlike GPL means it rests mostly on contract law, rather than copyright.
For example, imagine a ToS that says something like:
1. You acknowledge I am providing you something of value (my content) and you agree to provide compensation/consideration if you use that in an AI model.
2. By doing that, you grant me an irrevocable worldwide license to use and sub-license all content that emerges from the model, notwithstanding any future agreement you may make in reselling access or outputs of that LLM to anyone else. If a conflict should occur, you agree to indemnify me against claims by that other entity.
3. If you believe my content was not a material factor in some output, the burden of proof is on you to identify the specific output and prove that my content did not influence the it.
In short, this doesn't stop someone from stealing my art/writing, but it does put a potential hole in their attempts to monetize it.
For example, if they scrape my music and then license a copy of the new Music Generator 3000 to Disney, and Disney makes a movie with that music...
Music, meter, and rhyme are all (among other things) algorithms for indexing and error-correction, tools very suitable to the squishy hardware.
reply