Jump to content
Search In
  • More options...
Find results that contain...
Find results in...

the.crooked

VIP Member
  • Posts

    4,770
  • Joined

  • Last visited

  • Days Won

    1

Everything posted by the.crooked

  1. bump cus i miss this thread.
  2. This would actually be pretty interesting. The workflow would have to go something like this: scraper finds files with *.jpg or other image file extension places all files in some bucket train machine vision algo to assess whether "graffiti" or "not graffiti" have that algo run against collected images from step 1,2 compare images assessed as "graffiti" against all hosted images in 12oz forum that are still active links or uploads have bot upload into thread The problem here is the idea of constraining the images that are even reviewed in step 1, 2. The actual machine vision part is not too problematic, but it's precision might be pretty low due to the inconsistency of form across different styles. One might imagine running some sort of classification algo first across all images here on the site as it is. Rather than allowing for our perceptions of what define styles, let the algo segment them in an unsupervised fashion and see what it spits out. All in all this seems like a lot for very little return. But, a fun thought exercise no less. I think the segmentation in the last paragraph is the most interesting possibility. See what we missed but a machine sees across styles.
  3. I think you're absolutely right, @misteraven. Truth be told I had just spent roughly 8 hours staring at receipt images for work, and was reasonably drunk and feeling down about the general state of the world last night. I do think there are specific generational changes in how we now process information itself, but I don't think it's as bleak as I was positioning it last night.
  4. Slightly younger, 31. Will report back when 40+
  5. A few points from one of the younger but long term members: 1. I think most of us who grew up with this forum are faster at interpreting relevant information than we used to be. It's easier to skim internet writing now for relevant points, content, and points of conflict. 2. There's a specific strain of nostalgia that persists through the writing in the forum as it currently stands (specific to Channel 0 and the more off topic arenas) that almost feels redundant. This isn't a knock, just an idiosyncrasy of us attempting to realize/recreate the social structure pre social media that was emergent of the general "energy" that was pervasive within the forum. 3. I fear that it will never be as it was. I know it won't. I both am ambivalent towards that reality, and very sad about it. I recognize it as movement, as the only outcome of what was until now, but I also love and miss the intellect here. For all our bullshit, this was a concentrated space of very bright, brilliant, talented, and opinionated people. It was, in the truest sense of the word, a forum of ideas. I have to accept it won't come back to that. Passing thoughts. Passing greetings and updates. Passing experiences and missed connections. All that to say, wonk saggin, night owls, boogiehands, whodies, etc. Love this place and these people. ETA: i may not be sober rn.
  6. wuts a dawgy gotta do to get back into that VIP. i'm sittin here listenin to mad intense and depressing classical music on youtube (adagio for strings, if you're wondering) just whistfully remembering days of yore. ...some russian lady just got a standing ovation in another song for playing the oboe.
  7. I was having a conversation with my wife about this idea of whether or not forums would fall off. I think in the current climate, people are wanting to go back to something that feels more concrete in the experience. Even if this is still a digital interaction, and only a few of us know each other IRL, we all know each other here. Along with that sense of actual community is something that's a bit less transient than the news feed, or story feed, etc. Each one of those is a persistent stream of information that one ingests and moves on. Here, we represent our ideas in a static format, or at least, a movable piece that can be traversed at one's own speed. I'd like to think this place will pick up again. I suppose us having this conversation at all is indicative of such. Obvs, Raven can see the visit counts so he'll know that in the explicit sense. But, like the man said, it really is our space to do with as we please, so it's lack of movement or lack of interesting content is really on us.
  8. Oh for sure. That I'm not doubting.RE: all the data being used. I guess, again, for me it comes down to just not really caring.
  9. On the question of credit card data: Almost all of that data is being used for fraud detection. You mentioned geo-position data, most of that is being run through a logistic regression to to classify whether a given transaction is valid or fraudulent. Obviously there's secondary spending data that allows for spending profiles to be created, etc. but the main analytic function is fraud prevention. Risk and Fraud analysts/data scientists are some bad mother fuckers when it comes to statistics. If you wanna ever understand some strong real world modeling, talk to an actuarial scientist for a large insurance firm. Those folk are no joke in the math world.
  10. Straight up watched my cat do this from the inside of our apartment towards neighborhood bully cat. Bully cat had no idea wtf was going on.
  11. 50 brought IOU and Soup to hang out at my place in west oakland once if I remember correctly. I think it was actually at a party my neighbors were having. Some how marcofromhouston and I have never met in person despite him working closely with a lot of my good friends who work at FreePressHouston.
  12. And here's a video of real-time re-enactment of generated rgb video modelling: We are very much there with both voice and video. It's a slippery slope for sure. I think these things are more about the building blocks of strong autonomous AI in the format we have always thought of from sci-fi. Sure, there are potential nefarious uses, but that's always been the case with any technological advancement. Shit, more oft then not, it's the de-weaponization of technologies that have led to advancements in the public sphere of things that were crated under the guise of military or security direction. That said, Hua's initial questions regarding data security and use is a much different one than about the technologies above. It's the idea of combining all of this that leads to the very very grey area of questioning the social good of such technologies.
  13. and an example of it in use from 2 years ago when they debuted it in beta:
  14. I can only embed one video per post so here's 3 posts. Video released by Adobe and Princeton on their work together on VoCo a generative model for recreating natural speech in any voice:
  15. Ahh, Those are very real problems, to say the least. When I was in grad school, I had machinations of writing a program that scraped publicly available facebook posts to create not just political profiles of regions and people, but rather rhetorical ones. What is the language to best sell a given political concept. Does person x believe in christiantiy, and if so, do they believe in the stewardship of god's earth model? If so, they are prime for pushing climate change policy, but can only be tipped in favor if framed in that form, "How would god feel about your disdain to the gift he's given you?" etc. I was actually writing an article for some blog a friend runs on the notion of authenticity and the Turing test in light of these very technologies. Adobe VoCo is exactly one of the types of technologies you are concerned about. Introduced at a conference in 2016, it uses neural networks (what doesn't?) to develop a catalogue of phonemes and the other constructors of heard language. So, let's say I want to make a fake recording of Obama saying "Fuck Trump." With about 30-40 minutes of recorded speech being analyzed and broken down by the model, I can type "Fuck Trump" into the speech assignment panel, and it will spit out a seamless recording of Obama saying it. Couple that with the video manipulation techniques along the same lines, and yes, you have very real political concerns. I think one of the broader points about authenticity is about voice. Even if someone put together a recording of me saying some wild shit I would never say, for it to be believable, it would need to be in my specific behavioral patterns for stringing words together. We are very good at even reading tone in written text and determining whether someone would or would not speak that way. Generative NLP models like this are something I'm super interested in. That said, I think the concerns you are espousing are pretty inline for the direction we've been heading a long time. Piecemeal creation of the structures and models of our own cognition eventually giving way to the singularity. The problem of bad actors in a space leading up to that point though... well, I feel like a lot of us have felt fucked for a long time. When the general populace can't agree on even the value of objective fact vs subjective opinion, then how can I imagine some piece of shit Trump voter is really gonna be digging in to the subterfuge of geo-political strategy.
  16. Things are lookin good. Soon as the wifey gets a job up n running, I'll pick up a few things. Those patches are fire.
  17. I remember seeing those posts. Yeah, me and 50mill ended up working together at a well respected national grocer for a while till I was surreptitiously fired for some buuuullllllshit.
  18. HA. Honestly, just tried to look on wayback machine to see whats there in general, but I don't think you can view threads without logging in, but I don't think any of the code to authenticate the login works through wayback. So outside of scans of the thread lists, there's no way to see what was in threads when Raven locked the whole forum behind login.
  19. Caligula, Were you at the dolores park meetup years ago? I popped in briefly, but my then gf now wife had just dropped the fact that she was pregnant on me and was way to fucking high to interact with that information and seeing a bunch of 12oz heads, so I have no real memory of that other than meeting Suki for all of 15 minutes before we bounced to go talk out wtf we were gonna do.
  20. You are not buggin. I go to an annual party out in VT every year, thrown by a handful of people, and every year there's a core group and a rotating cast. This place has been a refuge for many of us, be it intellectual, be it the dumbest shit ever, be it illegal, etc. I have a feeling people would show up. It would certainly be a fascinating experiment. Even if for a one shot. I know there's lots of y'all I'd love to shoot the shit with over a beer IRL.
  21. Yeah, great questions. 1. It really depends on the company. The company I work for goes through great lengths to both be compliant with international law, and make sure that people are incapable of matching non-anonymized data with anonymized identifiers. Put one way, let's say you have a singular person ID of XXXXX which is considered identifiable with both your account, and your behaviors. That ID will be anonymized and the various behavioral data silo'd by department, use, and user. As a contractor, I have to pass through many layers of requests to even get anonymized data in most cases. 2. The only way this would work is if there was a non anonymized set of data that was standardized against the available data in the anonymized set, and then used to train a machine learning method that could find both enough differences in the non-anonymized set with enough reliability to predict a person's identity in a validation set, and then pushed against the anonymized data. This is really just a couple steps in an otherwise broad. Ultimately, this isn't incredibly likely. There are means that one could imagine. You could train a network to create profiles of diction for a given set of people, and then feed it quotes with no person attached to see how accurately it predicts the owner of the quote. 3. Case by case basis. Really depends on the contractual agreement in the terms of service, what the terms of the acquisition or merger are between companies A and B, and then what happens after that. Presumably a good faith agreement would exclude the idea of changing the ToS without notice. That said, new international laws are forcing companies to be able to reply for requests for data deletion upon consumer request, and to set up processes that can accomplish this within x period of time after the request. There's a lot of internal instability and unsuredness around those laws, however, mostly due to questions of implementation, and what level of aggregation can we keep data past. E.g. is it within the bounds of law that says you can't keep user behavior data after 60 days to aggregate it to daily values rather than individual user values, etc? Ultimately, these are questions and cases of law not having caught up with technology. 4. I think, ultimately, the notion of psychometrics is somewhat wishy washy. If you were to ask me whether the use of cognitive science is part of certain procedures? Sure. Why wouldn't it be? Are these all nefarious attempts at mind control through suggestive presentation? No. Data's use can be split into a couple means of thinking: I. Reporting II. Optimization III. Prediction What you are questioning is both II and III. To wit, most of what you are thinking about is what's referred to in marketing and sales as customer segmentation. In the technical sense, it's the application of unsupervised clustering algorithms along a number of dimensions to create behavioral profiles. Not of singular people, but rather of broader concepts of people writ large. Generally all these techniques have specific application but their base structures are agnostic to semantic content of the data but rather just difference in data itself. The end result is something interpreted in machine form, but based in the semantic value of a human interloper. Here's an example with Facebook: Based on my language, the complexity of sentence structure, the general topic of things I post about on facebook etc, not to mentioned self proffered data such as educational institutions I've graduated from, and countless other dimensions and conceived metrics, Facebook can make a highly probably guess that I'm: - Well educated - Politically liberal (their terminology, but lacks depth) - Highly Social - High value user - Long time user - Amenable to suggested posts - etc. These general reference points are useful in terms of "selling advertising" but whether those advertising dollars are well spent would rely on how much I give a shit to click on some served up ad next to me news feed. The better the model at understanding the broad strokes of my behavioral patterns the better ad serving will be in showing me something that is actually relevant to my life. Whether or not this is something that should be done at all is more a philosophical argument about the underpinnings of american democracy and where the boundary between it and economic philosophy begin and end. That said, facebook is a product, google is a product, etc. We engage in an active participation of a three party transaction where our eyes and behavior represent our respective payment for these products. Personally, I don't mind targeted advertising. I don't really believe in privacy so much as security, and I think American views on privacy are as much tied to puritanical notions in religion as they are oppression. For those that are concerned about manipulation of experience and information, the same concept that has always governed interactions with institutions still applies: critical thought. In the question of whether or not there's some cabal of technocratic string pullers attempting to nudge people one way or another with the levers at their disposal, the answer is unequivocally no. Conspiratorial thought fails in perpetuity against the sheer size of people in the companies with enough reach and information to attempt to do so. That said, everyone's experience of the internet is literally different. Big companies such as mine, facebook, microsoft ,apple, etc. all invest in large scale experimentation platforms. All of this in better attempts to refine both user experience and internal metrics for growth. Is that nefarious? Really just depends on your perspective.
×
×
  • Create New...