Amazon defeated Rekognition shareholder revolt by a large margin
bbc.com
bbc.com
While that's obviously challenging, targeting Amazon specifically doesn't actually improve the status quo in a lasting way.
[0] https://www.latimes.com/local/lanow/la-me-ln-no-charges-lapd...
Risk scores are computed for every traffic stop, wouldn't you want to take that process away from the office and give it to a computer?
Humans are accountable, changeable, cross-examinable. Computer systems are not. Please do not give enforcement systems over to computer systems, and have the bias calcified forever.
Computer systems are accountable, changeable, and cross-examinable. Arguably more so than humans.
I do stipulate that computer systems can be racist or otherwise biased and can be "just wrong".
The human mind tends to be set in its ways.
The history of junk forensic science has pretty clearly demonstrated that nobody in law enforcement cares about testing their biases, or improving the processes.
If it gets them results, they'll use it.
With algorithms, I think it can be done... but given how subtle the negative effects can be, the process must be stringent and err on the side of caution. In other words, any algorithm should be assumed to be biased until proven otherwise, and with a relatively high bar at that. There should be clear technical standards for what constitutes a valid proof. The testing process should be reproducible by third parties. And there should be a straightforward way to challenge either the process itself, or its results in any particular case.
In practice, given how low our bar for law enforcement accountability is today in other cases - including those where people die or are badly hurt - I'd say that we need to fix those issues first before talking about face recognition and its regulation in this context. And until then, it should be off-limits, because the potential for abuse is so high.
I agree. But the reason that we don't have that right now isn't related to whether or not computers are involved, and introducing algorithms won't solve that situation.
The reason that police accountability is weak is political.
Absolutely. One major positive outcome of the OJ Simpson trial was how Simpson's attorneys completely trashed the police handling of forensic evidence on a national stage, forcing law enforcement across the country to raise their game.
Please never enter politics.
What I want is this process to be cross-examinable. You can't cross-examine an algorithm in any meaningful way at a jury trial. You can't cross-examine a neural network at all.
You know all the bullshit about how police dogs are used to conduct illegal searches, and how you can't cross-examine a dog? It'll be like that, but much, much worse.
https://en.m.wikipedia.org/wiki/Bayesian_inference#Applicati...
What we have today is institutional racism based on people’s gut reactions. An NN can be tested and shown that given all other factors it risks race, or any other protected class higher than it should.
Institutional racism is not based on people's gut reactions. Institutional racism is based on flawed algorithms. Not computer algorithms in this case, but bureaucratic ones.
I don't see how a flawed computer algorithm would be any easier to correct than a flawed bureaucratic algorithm.
To address things like institutional racism, you have to change the institution. That's a human problem for which there is no technological solution (although technology can certainly be one of the tools used to create a larger change).
Since we've not made much traction on blackbox testing bureaucracies, I'm going to go ahead and say that blackbox testing algorithms will be even worse. (For some dumb reason, people are inclined to trust algorithms.)
I'm just intervening to point out that what you're saying is not what the paper you linked implies. Its title is "Semantics derived automatically from language corpora contain human-like biases", and the general conclusion you could derive from it is that AI programs reflect the stereotypes ... of the data they are trained with. This is why they use the right word: stereotype, instead of the charged word you used: prejudice.
See: https://www.vigilantsolutions.com/products/license-plate-rec...
Basically, you punch in a plate and get alerts, nationwide. It's likely that facial recognition technology can associate some car passengers to a vehicle as well.
See: https://www.aclu.org/blog/free-future/license-plate-scanners...
I don't know this, but I would guess that you'd attract attention if you started driving from South Florida and rolled through the wrong neighborhood in a place like Philly or DC on your way home to Boston.
This is why you have to go after the companies first and boycott them/cause PR nightmares for them.
The rebels will still find a way, but it will take years and many lives.
That it's now fashionable to install always-on microphones in people's homes, that we carry always-on cell phones broadcasting our information wherever we go, that the streets and skies are covered in cameras and sensors watching our every movement, yet we mill about with casual disregard for the implications, gives me little hope that, as citizens, we could ever collectively wake up or rise up. We are taught fear and hate and division. If we should ever forget, they'll remind us by writing messages with our blood. We point fingers at a bumbling scapegoat sitting in a temporary seat, while the real power lies off-camera.
This really depends on their definition of "misuse", which can be molded to suit their financial goals. Eventually reports of questionable law enforcement use will come in, and I have serious doubts that any large corporation would willingly cut off sales to a large set of existing government clients. This area desperately needs regulation.
Barret stopped selling to CA government agencies after CA banned "regular peasants" from owning Barret's flagship product.
That said, this situation is very different politically. People aren't going to stop using AWS en-masse because Amazon is helping enable the police state.
I had contacted them regarding these issues and was told that they are working with new models. I haven't tested it recently to see if the situation had improved.
I didn't use the Face Recognition API for privacy reasons, I assume that would be the most used API by law-enforcement; but if it is anything like my experience then I would be very worried for some innocent to be wrongly tagged by the system.
You should. Law enforcement in particular doesn't think that false positives are that big of a deal (after all, they argue, all that would cause is a cop to talk to the wrong person). They prefer to accept more false positives if that means fewer false negatives.
I disagree entirely with that.
What does shareholder primacy even mean anymore?
So really, all those shares are voted according to what corporate management wants, not to what the original retail investors want. It's distorting the power balance in a really perverse way.
It's hard to say at the moment, because the data isn't available yet, but I bet index funds voted >95% against this proposal - that's the historical average. We'll know for sure what the stats are once the fund providers disclose their voting, which happens yearly.
It seems like their duty to the index fund owners is to vote like other shareholders (which is most easily achieved by abstaining) and not like management.
Individuals are exempt for that.
It's interesting to think about index fund providers incentives:
(1) track the market (2) do it very cheaply
Voting shares is a cost center with no benefit towards the business. Fund providers don't want to spend time and money to do research on individual proposals, that would be too expensive. So they do the cheapest thing possible and default to following corporate recs.
(That's actually a simplification, what happens in practice is, they use third party proxy voting advisors, but the end results is mostly the same, except that they vote down really egregious proposals, e.g. crazy CEO pay in the context of a free falling share price)
I've building a startup for that -- you can support petitions like https://www.yourstake.org/ask/vanguard-demand-companies-disc...
Suppose index funds collectively own 80% of the shares. If they don't vote or do something that causes their vote to equal the other shareholders, someone else can buy 10% of the shares and effectively have majority voting power, or even fewer than that since most likely some of the other non-index fund owners wouldn't vote either.
The problem is that you need somebody who is paying attention to be in control of the company. Putting that on management is problematic because they obviously have different interest than owners, e.g. on matters of executive compensation or whether to return profits to shareholders instead of using them to build a personal empire/reputation or funneling money to cronies through mergers and acquisitions.
Instead, they vote the board recommendations, disenfranchise all the individual shareholders, and rubber stamp almost anything the board wants. I'm not sure that's any better.
This startup organizes retail shareholders to use their rights via petition, e.g. https://www.yourstake.org/ask/vanguard-demand-companies-disc... .
So as a verified shareholder, you can signal your support for various issues, and the startup then arranges a meeting with the fund managers if enough people support the idea.
With the rise of index funds also comes the rise of the idea of a 'universal owner' solving the tragedy of the commons.[0] If you own a piece of every company, like Vanguard or BlackRock, your funds benefit when the whole economy does well. Your incentive is to advance society, not to promote any one company at the expense of the others. In that way, you internalize externalities -- like climate change. Or in this case, getting industry-wide action on privacy/security.
That said, this idea hasn't sunk in with fund managers yet. They're beginning to talk about caring about 'social purpose'[1] but still voting against it[2].
So the retail investors will have to demonstrate they care in order for anything to change, and they don't even get a vote through mutual funds.
I built this website to organize verified retail investors around petitions, in order to do that. You can support petitions like this https://www.yourstake.org/ask/vanguard-demand-companies-disc... or this https://www.yourstake.org/ask/amazoncom-inc-facial-recogniti...
----
[0] https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3014952
[1] https://www.blackrock.com/corporate/investor-relations/larry...
[2] https://www.cnbc.com/2019/03/18/large-fund-firms-support-for...
Index funds only target top companies. Externalities tend to b eenefit those top companies while harming little companies as well as poor people, often not even in the same country.
If you mean top as in performance, that is also not the case. Index funds are based on the premise that they do not aggressively pick winners and losers within a sector, in contrast to active funds. This lets index funds hire fewer staff and charge fewer fees. Economists generally agree that over time, index funds outperform active funds on a risk-adjusted basis after accounting for fees.
But I’m not sure that would help against the very real and dystopian future that seems just around the corner. Facial recognition is probably the key cornerstone of said future and I’m not sure what we should do.
Banning facial recognition also feels wrong to me, and violates the aforementioned axiom about information.
Moreover, many cities are already blanketed in thousands of government cameras and facial recognition doesn’t seem like anymore an abridgement of our privacy than the cameras alone.
Maybe we should just ban police cameras on streets? But the ship has probably already sailed on that one.
Clearly (to me at least), you need a system of axiomatic beliefs. If your axioms don't form a coherent system (i.e you can prove a contradiction), then you really don't believe anything at all.
It's surprising to me that very few people think about morality this way. Everyone has a different view of morality of course, but you should at the very least try and distill it into an elegant system that is a) consistent and b) has sufficient predictive power to be useful.
An example of a system that fulfills a) but not b) would be something like: There's no morality anyone can do anything they want. That's only one axiom and pretty elegant as systems go. The problem with it is that it's so trivial that it can't assign value judgements to anything at all, rendering it all but worthless.
The system that I tend to follow has three primary axioms:
1) There are two kinds of things in the world: people and property.
2) People own property and they can do anything they want with said property as long as it doesn't violate the rights of others to do the same. Moreover, the person himself is property and thus is solely owned by himself.
3) By violating the axioms above, one forfeits his right to the above axioms.
While not perfect, this system is internally consistent and also has sufficient predictive power:
If someone punches me, he's violating my property of myself (2), and therefore can be sent to jail (3). If I break into someones house, same thing: violates (2) and therefore can be punished by (3).
There are some things that the system can't assign judgements to, however: What are children? What rights do they have? What about abortion? What exactly violates someone else's property? Does air pollution can't as violation? What about the environment? What rights to animals have (the system says they're property, but some people might have a problem with others torturing animals, for example)? Where does the government get it's legitimacy? Can the government levy taxes? etc etc etc
In truth, the things it can't make judgements about are probably than the things it can. But I personally haven't found a better or more elegant system to frame the world in, so I generally stick with it while being aware of it's limitations.
The answer to your first question is "oftentimes, yes" to the second two.
There is this pervasive myth held by people who tend to gravitate towards "functional" fields like ours--that is, technology, math, the logical sciences, and so on--that the human existence can be distilled into a series of logical statements that can be evaluated at different times and, given the same inputs, will always produce the same outputs. (It's the same origin as the myth of the "perfectly rational actor" in economic systems that makes a lot of assumptions that are invalid in the context of how people truly function.) This myth misses a key thing about humans: thinking animals are not computers.
What feels right or what beliefs someone holds can and usually do change over time as someone's existence progresses and new experiences are had. This is where some of our greatest benefits, like the ability to experience art in new and different ways depending on the context or situation or even just our mood, and some of our greatest tragedies, like holding one set of people to be negatively different based on an arbitrary criteria, originate. Saying that if someone's beliefs can't be reduced to "code" causes them to be invalid is against the human condition.
At best you could distill down to multiple separate self-consistent trees of moral rules and model individual humans as sets of weights for those.
It is possible to do two things at once and provide overlapping protection. When something has both technical and legal defenses, that thing is safer than having only one or the other.
I don't think there are any easy answers.
I'd say that the more general problem is that we all had privacy at one time due to the difficulties in gathering/parsing/storing. People used to be able to expect that their criminal record was somewhat private, that private conversations in public places were private, that their locations were private. 'Forgotten' can be interchanged for 'private' in these cases.
There was no real protection in terms of law, just in terms of practicality.
The great irony of modern times is that the sea change was schlepped in with the market for advertising, not your local Stasi. Of course, the cops are happy to use it if it exists.
It sure does to me.
And that's how facial recog is legal. Its legal to put people in public and write down who they are. So its legal for a computer to do it.
And I even wrote my own for a 2015 maker convention and deployed this. No GPU, no internet. CPU bound and worked very well: https://hackaday.com/2015/03/04/face-recognition-for-your-ne...
Aside lawmakers restricting law-upholders (hah), we've got a snowballs chance in hell in solving this.
FaceRec is here to stay. :(
Americans haven’t had privacy rights in a while and while the technology is new and fancy, American citizens haven’t been as free as they think they are in a while considering PRISM, border control checks that include your fingerprints and have the right to demand access your personal devices...