Slashdot Mirror
When It Comes to Gorillas, Google Photos Remains Blind (wired.com)
Tom Simonite, writing for Wired: In 2015, a black software developer embarrassed Google by tweeting that the company's Photos service had labeled photos of him with a black friend as "gorillas." Google declared itself "appalled and genuinely sorry." An engineer who became the public face of the clean-up operation said the label gorilla would no longer be applied to groups of images, and that Google was "working on longer-term fixes." More than two years later, one of those fixes is erasing gorillas, and some other primates, from the service's lexicon. The awkward workaround illustrates the difficulties Google and other tech companies face in advancing image-recognition technology, which the companies hope to use in self-driving cars, personal assistants, and other products. WIRED tested Google Photos using a collection of 40,000 images well-stocked with animals. It performed impressively at finding many creatures, including pandas and poodles. But the service reported "no results" for the search terms "gorilla," "chimp," "chimpanzee," and "monkey."
More than two years later, one of those fixes is erasing gorillas, and some other primates, from the service's lexicon. The awkward workaround illustrates the difficulties Google and other tech companies face in advancing image-recognition technology, which the companies hope to use in self-driving cars, personal assistants, and other products.
So what do their cars do now when they spot a gorilla crossing the road?
while I kiss this guy.
Google, Facebook and Twitter are curating the world's communications to fit their own political agendas. Each has been found manipulating information, despite denying it. Each of them has video evidence of their employees boasting about it.
And yet people are blindly allowing, even encouraging these tax-dodging global monopolies in their own sphere's to push a single way of thinking, even if that contradicts reality.
Have you ever gone to the zoo and looked at the larger primates? They're fascinating because they're so much like us; I defy you to look a silverback in the eyes and not see a near-human intelligence looking back at you.
To a human, they're obviously not human... but to an algorithm checking out just the facial features? I'm surprised this didn't happen sooner.
Some people look like gorillas. Some people look like reptiles. Get over it.
We ARE apes Google. DEAL WITH IT.
Why is this so hard to accept as not only true, but also a giant image recognition/computer vision challenge?
You go to nearly any zoo with large primates and you're bound to hear someone say "They look so human!" Well of course they do, humans are primates.
Which means that it works in reverse, too, primates look like humans. And it's not surprising that blacks look more like gorillas. I mean, there is the whole black coloration to begin with, but also the flatter nose and other facial features of gorillas which are shared with black more than Caucasians.
Of course no reasonable human would think that a black *is* a gorilla or vice versa. But computer vision? It's like version 0.01 alpha and the similarities are strong enough that it's not surprising at all that it would misidentify blacks as gorillas or vice versa.
they're not wrong...
Maybe they should fix the algorithm to tag anything that looks like a gorilla as lou ferrigno.
http://www.johnperkins.com/Eagle%20Monkeys.jpg
is a particularly niggardly way to avoid dealing with reality.
For fuck's sake google why not just label us all "ugly bags of mostly water" and be done with it? It's not so cold in Silicon Valley that google's ugly bags of mostly water should all be such snowflakes.
So, now we add "There's no such thing as a gorilla!" to the list of SJW fail.
Or maybe it's a hard problem to solve. Don't stop being you, RightwingNutjob.
The gorilla channel is real.
This has absolutely nothing to do with race at all and extremely little with machine learning. It's purely a technical problem and limitation of exposure of images. Looks like they need to refine their algorithms to process images more like the human brain does and quite possibly no amount of machine learning will fix this until they do that.
Black anything is difficult to expose correctly and when you put any bright object next to a dark object something is not going to be exposed correctly. The light subject will be blown out to expose the dark subject correctly or the dark subject will be darker to expose the light subject correctly.
With lighter complexions it's easier to see detail, with darker complexions, not so much. We just read about this 2 days ago in regards to the bird of paradise.
Once Google reaches an extensive presence in the African continent, their sample images used to train their systems will surely increase to level of ability necessary of discerning people from their genetic cousins.
I expect a lot of humans wouldn't get 100% accuracy at telling species from the same order apart either. We have certain hardcoded advantages when it comes to our own species.
I'm as white as they come and Google Photos has tagged several monkeys in my pictures as me. Nobody is writing news stories about that (as well they shouldn't!), but because this guy is black the world ended ?
Who can blame the algorithms for having difficulty telling the difference between one ape (gorillas) and another (humans). Its more of a cultural/religious thing that most people have such a hard time accepting the blatantly obvious truth that we're closely related (on an evolutionary/genetic scale).
And John Cena of course.
Some settling may occur during posting.
Google can just roll up the windows and lock the doors in gorilla neighborhoods.
You do not have the right not to be offended. Generally I shouldn't go out of my way to do something just to offend you but that's not even close to the case here. I seriously doubt many people were offended. I do however think a certain group of people used this as an opportunity to criticize google. This group of people care less about difficulties black people face than they do care about being seen about caring about black issues. There is a reason SJW is a derogatory term.
There are so many actual issues that black or native North Americans face where the solutions are actually hindered by SJWs. It is quite frustrating.
If I had an algorithm that occasionally misidentified people in a way that can cause public outrage, I would filter the outputs to avoid controversy too.
Wake me up when they release a fixed version. Hell, a paper describing the issue in detail would be interesting---even fascinating, if I were any sort of expert.
Googles themselves admitted that their algorithms still make the same mistake. This article boils down to "hard problem takes longer than 3 year to solve"---with excessive puffery.
---
According to the latest ruleset, this post should be modded as Vorpal Flamebait +5.
I wonder if this is why Google is also blind to steam donkeys?
I can only find ONE picture of a steam donkey, and it is not of one in actual operation, loading or unloading a ship, but one that has been half buried in an outdoor exhibit.
The rest are just pictures of the lyrics to the sea shanty "Donkey Riding".
YMMV, as this might only be because I was, in fact, previously searching for audio files of the sea shanty "Donkey Riding". I wanted to see what a real steam donkey actually looks like. Google doesn't seem to have had much luck finding that, and "helpfully" gave me pictures of the lyrics to "Donkey Riding" which is not what I was looking for.
...obscuring contrasts and therefore shapes.
It would be interesting to see how their algorithm did on pics with various color bit depths.
When Fascism comes to America, it will call itself Anti-Fascism, and tell you to give up your guns.
... since some black dudes were tagged as "gorillas", does that mean that some gorillas will be now detected as "black men"?
Because if not, I'll need to hurry up and find something else to be outraged about.
Of course, they are in a multitude of the same categories together:
The above applies to all races, but the skin color adds one more common category for Blacks. Big deal.
Somehow being called a "gorilla" is deemed offensive, which is patently ridiculous. Black panthers was fine, but gorilla is bad? Seriously? And because it is considered bad — irrationally — by someone, we can't have Google's image-recognition to work — they without special-casing? And they must continue to apologize for something?
In Soviet Washington the swamp drains you.
1) IQ tests are extremely culturally biased. There may be average intelligence differences you could correlate with skin colour, but none we can currently measure, and certainly none significant enough to use to prejudge individual ability.
2) Koko is a fraud that has been debunked several times. Koko is amazing, but nowhere near the level of amazing that the involved researchers proclaim.
3) Reality isn't nice, but you're racist.
Nobody has designed IQ tests to be culturally unbiased?
Depends. If it's 2018, yes. If it's 1900, no.
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
And how that 250 years of slavery was perpetrated by the attempted elimination of indentured servitude and abusive labor/pay practices used against 'good christians', as well as black people, native americans, mexicans, chinese, indians, and other ethnic groups. Prior to this, slavery and indentured servitude were often not that different and depending on state laws there were ways of bringing yourself out of slavery, whether changing your religion, economic means (since even many slaves/indentured servants were allowed savings, although just as often their 'masters' would find a way to steal/refuse to pay it), or by doing something sufficient to convince your master or a government official to return you to freeman status, whether due to heroic actions or illegal abuse by your master. The politics behind the full on generational slavery of the colonies is a good read, especially the parts about african americans becoming freemen prior to the instutionalized slavery that permeated the south during the colonial buildup leading to the revolutionary war, and then the years until the civil war after it.
If you fail to acknowledge the history of America, you will find yourself doomed to repeat it. The exchange from mostly economic slavery with racial undertones to outright slavery and racism has simply come full circle to a modern return to economic slavery, only now with videocameras replacing the slave masters and a slow burn towards the masses being indebted so that they will once more take on the yoke of economic, physical, and intellectual oppression that they might some day pay off their debts. Give it a few years and watch as the protections put in place in the 1780s, 1860s, and then 1960s get repealed and replaced as the pendulum swings back today.
Black people look more like gorillas than average in the same way white people look more like white bird poop than average. There are 'bad' things black people will look more like than other groups and there are 'good' things they'll look more like than other groups. All this algorithm did was uncover this relationship in its rough state. As it is refined it will uncover relationships more and more toward what is intended.
You didn't get the memo. There are no more hard problems. The science is settled. In 2018, how could you think otherwise? Any output you don't like is evidence of racism. Appreciation for complexity gets you nowhere when you're on a mission from God.
1) IQ tests are extremely culturally biased
If this is the case, then why haven't they been corrected for culture?
1) IQ tests are extremely culturally biased.
No, they aren't. Pattern matching has nothing to do with culture.
Well, when we have a universally agreed definition of what "intelligence" is, and have shown how it can be accurately and usefully quantified as a single number (a rather extraordinary claim in itself), then maybe someone could start to design an unbiassed test for it. Wake me up when that happens. The HHGTTG joke about the ultimate answer being 42 had it right: there's no point looking for an answer until you have properly defined and understood the question.
I mean, the person at Google who thought "lets automatically, and without consent, tag the public's photos with names as identified by an untested algorithm without any checks on identifying people as animals, celebrities, famous criminals, other people's partners etc. - what could possibly go wrong?" probably aced a shitload of intelligence tests.
In a survey of 100 programmers, 111111 thought that duck-typing was a good idea.
So your self driving car detects 4 gorillas on one side and one man on the other and it must hit one of the two groups, which one will it select?
I don't know what outdated publication you got your ideas from, but intelligence tests these days shed that "how many bottles of milk should Johnny put out for the milkman" problem long ago. They can accurately measure intelligence across cultures. Just so you know, and you can stop pushing that old and busted "cultural bias" line and leave it in the 1970s where it belongs.
Shutting down free speech with violence isn't fighting fascism. It IS fascism!
Better to not try at all than to make an innocent mistake that people take for racism.
What about Koko should be fraud?
Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
Truthfulness != racist but you get a trophy just for participating, snowflake.
Oh, right...
The more they ban and try to control information the greater the pushback when people eventually encounter the truth. Why don't the left realize they are creating their own enemy?
Some people genuinely perceive black people as looking similar to gorillas.
This is due to the way their minds break down the overall image of what they are seeing.
When I recognize a black person, I perceive the relative proportions of their facial features (nose, eyes, mouth, ears, etc.).
The image analysis software is likely strongly color relation biased, rather than 3D inference based.
Cheap software, cheap results Google.
About 20 years ago, jewwatch was in the top 3 listings when searching for jews on yahoo search, due to the way that links were created. Too many people got upset, so things were patched to prevent this, by adding code to explicitly prevent this.
I just went to Wikipedia's page on IQ and right at the top is a picture showing a sample from an IQ test consisting of nothing more than a series of geometric patterns with missing pieces. The idea is that you figure out what shape is missing in each pattern, and the more you get right the higher your IQ.
I'm curious how such a test is "extremely culturally biased" for anybody who isn't visually impaired.
dom
The error is to think that IQ tests only test native abilities. In a large part, IQ tests are about learned skills. Learned skills are culturally (or probably more accurately environmentally) biased. Look at the kind of abstract problems presented in an IQ test - it is quite easy to train about solving them. Does the few days (or even the few months) you spend on training on these tests make you more intelligent ? No, but you'll score higher. Likewise, the education you receive and the environment you are exposed to bias the tests results.
> Well, when we have a universally agreed definition of what "intelligence" is, and have shown how it can be accurately and usefully quantified as a single number (a rather extraordinary claim in itself), then maybe someone could start to design an unbiassed test for it.
IQ is the correlation factor shared by all of our attempts at measuring intelligence. There are different tasks on different tests, sure, but we know that the scores all correlate with each other, so this factor is essentially what we try to measure. It's effectively some measure of processing speed and memory for our brains.
Anyhow, the claim of cultural bias is falsifiable, in that it should be possible to provide intelligence tests that invert the bias against certain groups. I'm curious if anyone has such examples to cite? And for the record, I'm actually asking here, I haven't seen one yet.
The real problem is people looking for racism in everything, perpetuating it generation after generation.
No wonder they found no gorillas or chimpanzees using the term "monkey". It should be "ape".
...and if you're looking for it, you're guaranteed to find it.
I truly hope for this to actually happen, some day!
Gorillas casually using human automation to further their own goals. ... May be how the planet of the non-human apes gets started. ^^
Nor primates. Let alone apes. And apes aren't people either to them.
Even though they themselves don'l even qualify as individuals, or lifeforms even, given them being merely, limb-likes of their opinion maker's swarm body. Many can, at most, be qualified as literal tools.
Not judging though. Just observing. Including the harm resulting from it. It is simply a sad state of most "humans" that may need fixing. (Although you don't "fix" a bee swarm either. So why an ideology swarm? The harm is the problem. Not ther undividualness.)
Did you really go from a cartoon gorilla eating a watermelon to rape and murder in two paragraphs?..
In Soviet Washington the swamp drains you.
As usual, dave420's sole "contribution" to discourse is a personal insult...
Your invocation of "coloring" in vain was racist. Fuck you, racist. Shitposting racist...
In Soviet Washington the swamp drains you.
Ah, I see your problem - you are confusing intelligence tests with IQ tests. Gotcha. Intelligence can and is tested with culture-free tests. It's not 1970 any more, grandpa. There has been a lot of progress in intelligence research.
Shutting down free speech with violence isn't fighting fascism. It IS fascism!
I don't see the original post you are addressing, but your first assertion really stood out.
I would like to see your evidence to support the assertion of cultural bias in IQ tests. Certainly past tests did possess some, but every effort has been made to attempt to remove cultural bias. Are you shouting "cultural bias" because there are differences between the scores of various peoples? The purpose of the test is not to evaluate everyone as equal, but to equally evaluate people so the results can be useful. Indeed the test does serve as a predictor of successful behavior we generally consider to require intellect. When a test as simple as pairing a picture of a sheet of paper with fold marks and the resulting folded shape gives very similar results to IQ tests then the "bias" is for capability, not culture.
You are welcome to cite the actual historical facts of which you find me ignorant. The above statement is offensively condescending and indicative of your lack of any specific arguments.
Citations? Citations supporting your implied claim, that such outcomes were not merely potential, but likely — and that the cartoon-depictions increased/contributed to the said likelihood.
Jim Crow laws are fully irrelevant to our subthread — these laws did not deny Blacks police protection, nor have the cartoons you found offensive made in the formerly Confederate states, where these laws were in effect. Please, stay on topic.
It being ugly does not mean it was widespread. And it being widespread does not mean, a programmer today must apologize for his algorithm's failure to reliably distinguish between Homo Sapiens and other hominids. It certainly is a feature worth improving or a bug worth fixing, but it certainly is not a manifestation of racism or even of "insensitivity".
In Soviet Washington the swamp drains you.
Actually you don't need an agreed upon definition. Some endeavors undeniably require intelligence. If the test results in a statistically useful prediction of who will succeed in such endeavors then the test is valid. This does not mean it is infallible. It does not mean that all manner of intellectual capacity is included in the assessment. But once the test becomes a valid predictor then the tests efficacy has been proven to at least some degree that can be statistically determined. I see no evidence presented to assert there is bias. Are you suggesting that European/American cultural participants conspired to make east Asian as well as Ashkenazi culture the preferred cultural "bias?" That notion is absurd but east Asians and Ashkenazim do better than Europeans and Americans. How can this be?
Additionally, a simple test such as successfully pairing a folded object with a piece of paper with fold marks provides very similar results. Now how culturally biased can folding something be? Every effort has been made to reduce cultural bias. The only "evidence" of bias is the insistence that any difference in average scores between any identifiable groups must be bias. There is no other evidence. In other words it is an unfounded belief that is violated by the test rather than any bias. Of course the purpose of the test is not to proclaim intellectual capacity in a bold single number. The purpose is to predict successful behaviors. The test does reasonably well at this. People who do poorly and have their own countries and self-determination are not particularly successful at some pursuits while people in very difficult circumstances who test well demonstrate success in those pursuits. Personally I believe the fact of high scores of Asians completely destroys accusations of bias. The notion is utterly absurd.
>>1) IQ tests are extremely culturally biased. There may be average intelligence differences you could correlate with skin colour, but none we can currently measure, and certainly none significant enough to use to prejudge individual ability.
If IQ tests are culturally biased, presumably there have been IQ tests developed on which blacks outperform whites? Iâ(TM)ll wait for you to provide the links to the findings.
The mistake is thinking the test measures intelligence. The test predicts the odds of successful actions in certain activities where people generally believe intelligence is required. So if it is true that it's easy to train to solve the test answers (it isn't) then it doesn't matter. By demonstrating the ability to train for the test the desired measured behavior is demonstrated and scored. Truly simple tests such as paper folding give very similar results. Perhaps some specific training in origami may help but there are people who reach a certain level of difficulty and give up. Training will only skew results so far. The test is statistical in its evaluation so any such training will get lost in the noise. Again it does not measure intelligence. It is only a predictor of some behaviors. It is not perfect and doesn't pretend to be. People just misuse it.
That said I agree with your assertion that learned skill may be involved. If so then so what. If "learned skill" helps make people more capable and the test attempts to predict capability then attempting to remove the changes in a person gained by experience from the test is like trying to outrun your own sweat. Thinks of it this way. If Einstein had been born on the Australian aboriginal plain long prior to the arrival of Europeans, then how good of a physicist would he have been? If he was magically transported to 19th-20th century Switzerland from that aboriginal plain would he have been as good a physicist as he was? Of course not. If the test reflects that difference then it is a good test for predicting successful actions in certain sets of behavior. That's what the test does.
Of course the real issue is why can the software not tell the difference between gorilla and human? There are definable and measurable differences. Seems like shoddy work.
Not to mention that, even if we established that US blacks were 10 IQ points less than US whites, it wouldn't necessarily mean anything biological. It would mean the average US black was 10 IQ points smarter than the average US person in 1930.
"When you have eliminated the unacceptable, whatever is left, however improbable, must be the truthiness" - Holmes
If it was that easy to train for them everyone (at least everyone who would gain from it) would have an IQ of a trillion.
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
Nothing in there backs up the assertion that was made. You really should try to read whole sentences, not individual words.
You should read this and learn about how exposure works.
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
Among other things, the researcher lied about what koko was 'saying' in order to get hot female co-workers to show him their tits. He said the gorilla would be angry if they didn't show the gorilla their nipples.
Simple fact check: search google images for "gorilla". Lots of results. All (that I saw) are, in fact, gorillas.