(Thread IKs:
Platystemon)
|
Main Paineframe posted:Facebook already knows what you looked like five years ago, because they were running facial recognition to identify users more than five years ago. Yes, but they don't know which photos of me from five years ago I'd want them to show me.
|
#
?
Jan 18, 2019 19:13
|
|
|
|
| # ? May 31, 2024 03:49 |
|
|
Trabisnikof posted:Yes, but they don't know which photos of me from five years ago I'd want them to show me. Yes, but the 10 year challenge thing isn't going to show them that. The problem you're talking about has nothing to do with photo quality or style. The problem is that they're either not tying the data together enough (which seems doubtful) or that they aren't making those ties available to the "here's what happened five years ago" algorithm (extremely likely).
|
|
#
?
Jan 18, 2019 20:10
|
|
|
Main Paineframe posted:Yes, but the 10 year challenge thing isn't going to show them that. You're mistaking my point. I'm saying that it seems perfectly plausible to run a campaign like this to train a model to classify photos for preferences of historical human images since that's obviously something facebook has had issues with in the past and would be far easier to create using a narrow dataset for training versus trying to extrapolate from their existing graph. Also consider facebook has a history of running secret and unethical studies on human subjects at massive scales before, so its not that conspiratorial to wonder if they're doing it again.
|
|
#
?
Jan 18, 2019 20:21
|
|
|
you're just listing a bunch of totally unrelated things that have absolutely no connection to each other at all, and then saying "see? it all fits together perfectly" i tried to explain why but you're totally ignoring me, so w/e
|
|
#
?
Jan 18, 2019 20:57
|
|
|
Main Paineframe posted:you're just listing a bunch of totally unrelated things that have absolutely no connection to each other at all, and then saying "see? it all fits together perfectly" No the connection is pretty simple. Facebook has a known and well publicized issue with showing people old photos they don't want to see. Solving that problem using their existing datasets would be vastly more challenging than collecting a large new training set. This hashtag gathers that training set. Facebook has a history of doing secret and unethical studies on people. Thus, it isn't unreasonable to think that this might be an intentional data collecting activity from facebook. It is by no means proof, but it isn't an unreasonable guess.
|
|
#
?
Jan 18, 2019 21:02
|
|
|
https://twitter.com/erbrod/status/1085623194759774209
|
|
#
?
Jan 18, 2019 21:35
|
|
|
Reminder that there's zero credible evidence that it's beneficial and in fact your own body is going to recognize the transfused cells as foreign and kill them. And the more often you infuse, the better your immune system will get at it.
|
|
#
?
Jan 18, 2019 21:46
|
|
|
Does this mean I can sell my blood as a young(ish) person
|
|
#
?
Jan 18, 2019 21:47
|
|
|
Trabisnikof posted:No the connection is pretty simple. I've bolded the parts that don't have any basis. Solving the problem of "showing people pictures of their dead family" using their existing datasets is perfectly possible, and this new dataset of "old picture and new picture side by side" would provide absolutely nothing helpful at all for that problem. You don't need machine learning to figure out that people love to see old pictures of their living kids but hate to see old pictures of their dead kids. It's not an image recognition problem at all, it's an issue of analyzing the text of people's posts and correlating that to unrelated images. Collecting more images wouldn't help with that problem.
|
|
#
?
Jan 18, 2019 21:53
|
|
|
e: Whoops, misread.
|
|
#
?
Jan 18, 2019 21:58
|
|
|
Main Paineframe posted:I've bolded the parts that don't have any basis. You seem to be saying that because facebook could develop an alternative without new data they would go with that alternative. But that's ignoring how it is much more complex it would be to do what you're proposing versus gathering a purpose build training set. Facebook doesn't have any extant data that tells them which historical photos of a user they would want to see again, instead they only have proxies for that preference. Extrapolating those proxies either requires a vastly more complex ML system or a bunch of rule building by engineers, either option isn't a compelling solution of a company in the business of data. Especially compared to just gathering a new dataset with that specific preference embedded. It also seems that your misunderstanding the specific problem that this data would be useful for. It isn't an image recognition problem, its an image classification problem. Collecting a training set of preferred historical photos based on an existing photo is a vastly valuable dataset for ML because of the fuzzy decisionmaking of preference. The point isn't to collect the image, its to collect the preferences, the differences between the two images, and the relative similarities between all the historical images. Again, facebook has a history of doing unethical experiments like this, why give them the benefit of the doubt in light of their past actions?
|
|
#
?
Jan 18, 2019 22:32
|
|
|
Rhesus Pieces posted:Jesus Christ that thread seriously somebody loving screenshot that tweet it will probably save lives e: whoops too late  e2: yes i know it was a screenshot already but more places = better than dont be mean to me has issued a correction as of 23:02 on Jan 18, 2019 |
|
#
?
Jan 18, 2019 22:41
|
|
|
the original form of this nefarious cyberweapon was to post your current profile picture with one from 10 years ago. Facebook has already known, for a decade, that you were happy to display that image, because it was your profile picture. Facebook couldn’t learn anything interesting from this exercise, least of all about algorithmic cruelty, because there are no labeled negative cases (which in the general form makes the problem difficult to train for, among other things). the interesting cues are in the post text and comments, but will likely require signals from multiple posts to be strung together. plus, for example, I’m more than fine seeing pictures of my ex-wife, but that’s not true for everyone and their respective exes. (source/bias-if-you-think-so: I worked with facial rec, the AI research team and folks looking at ways to reduce on-this-day discomfort)
|
|
#
?
Jan 18, 2019 22:41
|
|
|
The Slack Lagoon posted:Does this mean I can sell my blood as a young(ish) person no
|
|
#
?
Jan 18, 2019 22:52
|
|
|
IIRC the evil blood companies buy blood that was donated in the regular manner, just like hospitals do. It's something like $500 a bag.
|
|
#
?
Jan 18, 2019 23:30
|
|
|
Vox Nihili posted:IIRC the evil blood companies buy blood that was donated in the regular manner, just like hospitals do. It's something like $500 a bag. wait so it's just normal blood, at a 1400% markup?  VC are such hilarious grifters.
|
|
#
?
Jan 18, 2019 23:34
|
|
|
Easy Diff posted:wait so it's just normal blood, at a 1400% markup? All donated blood is inscribed with the details of the donor, so presumably they buy blood that matches what they want (from young people) and then yeah, sell it to old people with matching recipient antigens at an enormous markup.
|
|
#
?
Jan 18, 2019 23:43
|
|
|
The hosed up part is that some young person gave part of their body in the hopes of saving lives and had their tissue sold to a huckster and a vampire instead.
|
|
#
?
Jan 18, 2019 23:44
|
|
|
Vox Nihili posted:IIRC the evil blood companies buy blood that was donated in the regular manner, just like hospitals do. It's something like $500 a bag. lol
|
|
#
?
Jan 18, 2019 23:44
|
|
|
Trabisnikof posted:You seem to be saying that because facebook could develop an alternative without new data they would go with that alternative. But that's ignoring how it is much more complex it would be to do what you're proposing versus gathering a purpose build training set. It's not an image classification problem either. Facebook gains absolutely nothing from just looking at the images. "This user likes this one, single profile image from 5 or 10 years ago" isn't something that can be usefully extrapolated to anything useful, even if you look into a mirror and chant "machine learning" three times. The thing you're not getting is that the problem you're describing doesn't really have anything to do with the images in the first place. The point is that people don't want to see pictures that remind them of bad things that have happened in their lives. It's nothing inherent to the images themselves - it's that the images show things that the people would rather forget. Bad relationships, dead family members, whatever. After all, the people obviously thought the image was fine to post at the time, but something happened in their life since then that changed their perspective on the image. No amount of image analysis, image recognition, or image classification can determine that: what's necessary is to analyze people's actual posts for information and tie it to the content of the pictures to influence their algorithms when deciding which picture to put in a festive frame. And, honestly, Facebook should already have that information. loving grocery store loyalty cards can tell if someone's pregnant, there's no way Facebook isn't already mining and tying together poo poo like family status (including whether your kid is dead) for their advertising demographics. They're probably just not making that information available to the particular algorithm that picks a picture from last year to show you.
|
|
#
?
Jan 18, 2019 23:57
|
|
|
https://twitter.com/girlziplocked/status/1086147271270547456?s=19
|
|
#
?
Jan 19, 2019 00:21
|
|
|
A lot of you are assuming Facebook is doing something at all useful or functional with the data rather than it being some middle manager's idea that doesn't actually work.
|
|
#
?
Jan 19, 2019 00:28
|
|
|
This whole article is hilarious: https://www.nytimes.com/2018/05/19/style/biohacker-death-aaron-traywick.html Some highlights: quote:Rich Lee, who created a vibrating penile implant called the Lovetron 9000, recalled that Mr. Traywick wanted to own 75 percent of profits made from sales of his device in exchange for funding. He said Mr. Traywick offered to pay him only $5,000 and was insistent that Mr. Lee presell the implant before it was finished and tested. “I was left with a bad taste in my mouth,” Mr. Lee said. quote:On Feb. 4, Mr. Traywick took the stage at the convention and told the audience that Ascendance Biomedical had developed a vaccine to cure herpes. He also announced that the company’s technology could be used to address a panoply of genetic diseases. Interested people were told to simply get in touch with Ascendance. quote:David Ishee, who first became interested in genetic manipulation because he breeds dogs, said that Mr. Traywick’s approach to biohacking was off the rails.
|
|
#
?
Jan 19, 2019 01:40
|
|
|
Vox Nihili posted:The hosed up part is that some young person gave part of their body in the hopes of saving lives and had their tissue sold to a huckster and a vampire instead. Nationalize healthcare now
|
|
#
?
Jan 19, 2019 01:45
|
|
|
Vox Nihili posted:vibrating penile implant well that's the most terrifying thing I've read all day
|
|
#
?
Jan 19, 2019 01:50
|
|
|
Vox Nihili posted:The hosed up part is that some young person gave part of their body in the hopes of saving lives and had their tissue sold to a huckster and a vampire instead. This is already pretty standard. A large part of plasma collection is to sell it to pharma for bio manufacture - including in countries with nationalised blood collection and health care systems (specifically Aus). hopefully the drugs made that way will also be life saving to some extent, but pharma gets their 600% worth of profit on it which is the real blessing.
|
|
#
?
Jan 19, 2019 01:58
|
|
|
Vox Nihili posted:The hosed up part is that some young person gave part of their body in the hopes of saving lives and had their tissue sold to a huckster and a vampire instead. Do blood donation places keep that much data on the original donor of the blood? It's probably not even young blood. You gotta get the stuff yourself if you want to be sure, like that hungarian countess who filled her bath with young maidens' blood. Vox Nihili posted:This whole article is hilarious: https://www.nytimes.com/2018/05/19/style/biohacker-death-aaron-traywick.html Wait, is that using "jet fuel melting steel beams" as an example of something totally unbelievable?
|
|
#
?
Jan 19, 2019 01:58
|
|
|
Vox Nihili posted:I just wanted to make dogs glow, and it’s taken years. Dogs already glow dude just give them a treat and some attention how hard is this
|
|
#
?
Jan 19, 2019 02:03
|
|
|
Vox Nihili posted:This whole article is hilarious: https://www.nytimes.com/2018/05/19/style/biohacker-death-aaron-traywick.html God I love watching ~biohackers~ own themselves.
|
|
#
?
Jan 19, 2019 02:06
|
|
|
I just wanted too make dogs glow guy is loving rad, sorry
|
|
#
?
Jan 19, 2019 02:11
|
|
|
Like if you've ever lost your dog at night, you'd know what a big help it is.
|
|
#
?
Jan 19, 2019 02:12
|
|
|
Traywick is like the unsubtle Stan Lee civilian name for the super-villain Trainwreck
|
|
#
?
Jan 19, 2019 02:27
|
|
|
Turtlicious posted:I just wanted too make dogs glow guy is loving rad, sorry A noble dream.
|
|
#
?
Jan 19, 2019 05:03
|
|
|
As a biologist, biohacking is some of the most birdbrained poo poo ever. 
|
|
#
?
Jan 19, 2019 05:53
|
|
|
people read snowcrash or whatever and everyones a goddamn expert
|
|
#
?
Jan 19, 2019 07:00
|
|
|
biohacking is so loving lame nerds are the loving worst
|
|
#
?
Jan 19, 2019 07:12
|
|
|
Turtlicious posted:I just wanted too make dogs glow guy is loving rad, sorry Not quite dogs, but obligatory: Emperor X wrote a song about the 'ray cats' proposed by the Human Interference Task Force and it grew into an album somehow. Yup, we're back at the 'not a place of honor' part of the thread loop.
|
|
#
?
Jan 19, 2019 07:20
|
|
|
Low Desert Punk posted:biohacking is so loving lame It's the purest distillation of dweebs reading a wiki article and acting like theyre experts, except instead of being wrong online they start injecting themselves with random poo poo I am in actual awe at how dumb these people are
|
|
#
?
Jan 19, 2019 07:53
|
|
|
Agean90 posted:It's the purest distillation of dweebs reading a wiki article and acting like theyre experts, except instead of being wrong online they start injecting themselves with random poo poo I am in actual awe at how dumb these people are wasn't the other guy in that nyt article the 'crispr my gains' guy
|
|
#
?
Jan 19, 2019 08:10
|
|
|
|
| # ? May 31, 2024 03:49 |
|
|
https://twitter.com/nypdchiefofdept/status/1086666128745730048
|
|
#
?
Jan 19, 2019 18:29
|
|
