> I sure am surprised that the discussion has already reached the point where tech companies are debating whether they kill people 'incorrectly'. I must have missed the democratic discussion about private businesses assisting in killing people at all, a duty traditionally exercised by states.

Sure, it's valid to disagree with Google's involvement in this project, but I find it incredibly hard to justify the claim that private enterprises traditionally don't equip armed forces. Providing technology and hardware to armed forces is something that private companies have been doing for centuries. Heck non-state groups selling arms to armed forces likely predates the existence of states as we know them. Most armies prior to the modern era were militias that privately purchased their own equipment.

This project is "assisting in killing people" to at most the same degree that developing munitions guidance, military radar systems, and sonar are "assisting in killing people" (I'd argue less since this technology is purely about reviewing reconnaissance data, not about the actual deployment of weapons) and private companies have been developing those systems for over a century.

That's not a very apt analogy. Google is perceived as a benevolent or neutral service bordering on a public utility. They collect vast amounts of personal information / intelligence because people willingly give it to them.

Google assisting in drone strikes is like the military using the electrical grid to electrocute people in their homes or delivering lethal poisons via the Starbucks Rewards program.

What if your favorite cereal brand started participating in assassinations? It's not inconceivable to deliver poison that way. What if, due to a software error, instead of terrorist #41253 your daughter was delivered poisoned Cheerios.

There is a huge ethical gap between private arms manufacturing and a non-militarized company with this much reach participating in a non-wartime, "preventative" military campaign.

> Google is perceived as a benevolent or neutral service bordering on a public utility.

That is not an accident. They worked overtime to construct that image.

Yet it might come as a little surprise that they are more in bed with the government than Northrop Grumman, Verizon, Goldman and Comcast: https://www.opensecrets.org/lobby/clientsum.php?id=D00006782... they spent about $18M in lobbying last year outspending all those other "evil" companies so to speak.

Schmidt was directly involved in Hillary's campaign and there is a picture circulating of him at one of campaign events wearing a "staff" badge: https://www.politico.com/f/?id=00000158-6a12-d435-a9ff-7f5ae... that coupled with the "killing people the right way" is all you need to know what Alphabet stands for and where it is headed.

> There is a huge ethical gap between private arms manufacturing and a non-militarized company with this much reach participating in a non-wartime

I claim that there isn't in the sense that once a company becomes large enough it will start participating in killing people "the right way" as Schmidt put it. That happens regardless if we still want to have that cutesy, colorful, not-evil search company image. Amazon (AWS) is selling billion dollars worth of services to the CIA, I am sure that's not for cat pictures and buying books. Microsoft's hands I am pretty sure are not exactly clean either, etc.

And let's not forget all those car or clothes manufacturers who shamelessly and without any moral qualms offer their products to whomever who can pay for them, be it a terrorist, a government agent, or a common run-of-the-mill criminal, each of who then goes to perform their horrible acts, aided by the purchased product.

These manufacturers are playing both sides and should definitely be kept an eye on, I tell you!

> And let's not forget all those car or clothes manufacturers who shamelessly and without any moral qualms

Good point. Not just moral legal as well. There is a whole bunches countries companies cannot do business with because of embargoes. In many cases it is because of terrorism. Depending on what technology there are also export controls to watch out for.

From Post-WW2 to 90's Gulf War Harley-Davidson built bomb parts in the same factories the bikes were built.

Why are you not outraged about this and saying things like 'Harley assisted the military in killing humans for the past 70 years with metal casings that may be used on your softail?'

I mean why stop at google? The private industry has always been supplying the government with goods and services, some of which go to the military.

Because the precision scored Tomahawk shell casings they produced caused a net decrease in lives lost, or were used to kill members of an utterly vile regime's military.

> What if, due to a software error, instead of terrorist #41253 your daughter was delivered poisoned Cheerios.

Using your analogy, isn't that exactly what the Pentagon is trying to prevent by seeking Google's assistance in more innovative solutions? Otherwise, what's the alternative?

The alternative is not doing it.

Do I really need to spell out why drone strikes are already ethically questionable? Improving automated recognition will not reduce civilian casualties, it will only help justifying the strikes because instead of having humans say they identified the target based on its height they can say the computer did it (based on "deep learning", i.e. based on its height but with computers).

The analogy is highly misleading. Poisoned Cheerio deliveries would be highly targeted. Drone strikes are not. The only way drone strikes avoid civilian casualties at the moment is by defining casualties as terrorists until proven otherwise (e.g. if they're children) and not having sufficient data to accurately say how many people were killed.

> This project is "assisting in killing people" to at most the same degree that developing munitions guidance, military radar systems, and sonar are "assisting in killing people"

Which incidentally, many developers find completely abhorrent.

Literally all of the equipment used to kill people, from the rifles, to the bullets, to the fighter jets, to the parts to repair the fighter jets, are manufactured by private companies in the United States. So I’m confused by your confusion in thinking that Google’s assistance in the drone program is somehow the first private company to participate in providing goods and services that are directly involved in killing people. Is Google somehow more culpable than the companies that make the bullets or bombs? How is providing software and computing power for that any different? You don’t think a private company writes the software that targets things for jets and missiles?

The majority of which will have been produced by armaments companies.

It would be strange to be outraged if, for example, Lockheed was doing something for the military, as they are primarily a defence company. Some might choose not to work for them for that same reason.

Google are a search and advertising company so it's not terribly surprising there's some negative reactions.

Google has honored existing Boston Dynamics military contracts while owning them.

Another thing to consider is how companies like BD get money. A LOT of research money comes through military. It is an unfortunate side of things. Lots of people think the tech is cool and exciting, but stop there. The military says "this is cool and exciting AND we can use it. Here's some money." There will be spinoffs, but this is a big reason a lot of funding comes through the military. They have money.

I remember this very discussion happening at the time of that acquisition.

When did private businesses ever not assist the state in killing people?

I think the abhorrent trend is the outsourcing of the people who should be assassinated, and that those people on the kill list will be selected by an ML model.

Google will just a be a processor in a long chain of steps, the actual killing will get passed to a drone team in Las Vegas.

https://en.wikipedia.org/wiki/Creech_Air_Force_Base

If Google does a better job of defining who should be killed, and it siphons money into the Bay Area, why shouldn’t we be supportive of google’s involvement?

I agree that perhaps we shouldn’t blow people up overseas, but at least while we are at it, why not make the best decisions to avoid killing the wrong people?

We're not talking about killing imminent threats for self-defense. We're not talking about bombing military targets in a legal war. We're talking about remote drone strikes against poorly defined "non state actors", typically in civilian areas.

Maybe we shouldn't try to legitimise it and abdicate the moral responsibility to a computer. Maybe we shouldn't mislead drone operators by showing them a popup stat "this blob you're looking at has been identified as a valid target with 90% confidence" so they feel better when pushing the button.

If we were talking about space lasers that make an individual person's head explode and high res camera footage you might have a point. But we're literally talking about lobbing bombs at people in public spaces and doing so based on grainy low res data.

The last thing we need is more talking points to justify this.

Extrajudicial killings are cool, as long as they are profitable for the Bay Area because the one thing the Bay Area is missing is.. money? Jesus..

>I agree that perhaps we shouldn’t blow people up overseas, but at least while we are at it, why not make the best decisions to avoid killing the wrong people?

I think you have to begin a little further back in the logic train than that state-sponsored murder for hire with the help of closed-source algorithms is legitimate or acceptable.

Or. You know. Stop killing people for oil?

I think that oil was the main cause of wars in world during the last decades. And it is one of the rare subjects where people can act with their wallet: choose EV each time you can, do not use oil for heating.

Just don't forget to implement the counting of false positives. The current system doesn't work well, given that US still counts al-Mansoura as 30 ISIS fighters killed and not 150 IDPs as per the independent UN commission investigation.

I remember hearing stories about how the British hired toy makers to make guns and furniture makers to make planes back in world war 1.

IBM was put to work manufacturing rifles during World War II. That's not a product I would have expected from IBM. https://www-03.ibm.com/ibm/history/history/decade_1940.html

Well the products you'd expect from IBM, during that period, have also had a rather controversial use: https://en.wikipedia.org/wiki/IBM_and_the_Holocaust

IBM sure must've been spread thin during WW2. [0]

[0]: https://en.wikipedia.org/wiki/IBM_during_World_War_II

World War 1 ended. The entire premise of free society is that war comes to an end, there is a such a thing as peacetime. Somehow that idea has died, and with it morality and social conscience. Comparing our spat with a few mountain dwelling cavemen on cellphones with a global War to threatened the existence of civilization itself is beyond ignorant. It's criminal intellectual negligence.

Please don't post ideological rants to Hacker News. Whatever your point is intended to be, flamebait drowns it out and has the side effect of wrecking this place.

https://news.ycombinator.com/newsguidelines.html

You could argue that the US has done a great job protecting us. Hence why you view it as a non threat.

We shouldnt be as naive as the Romans.

> I must have missed the democratic discussion about private businesses assisting in killing people at all, a duty traditionally exercised by states.

States have contracted out to private individuals and firms for the development of tools for that purpose (and the actual killing) for as long as states as distinct from private individuals exercising power have existed.

To me that sounds like a misreading of the statement. It makes more sense to interpret it to the effect that tech companies are worried that "the military-industrial complex is using their stuff to kill people" and that this is incorrect. Google doesn't really specialize in optimizing lethality, but it does make a lot of noise about ethical application of its tech.

When people refer to the military industrial complex, they are referring to the businesses like Lockheed Martin, Raytheon, etc. that are the private companies involved in designing and building weapons.

Also, private companies are contracted for military services. Remember Blackwater?

Exactly. Those companies are civilian as google is. Lockheed is next door even! Google is part of the intelligence wing of the MIC. The only controversy is that some of the more Berkeley / Burning Man googlers are surprised to learn that.

Actually a lot (or at least a not insignificant amount) of MICs are burners according to the book Stealing Fire

> I must have missed the democratic discussion about private businesses assisting in killing people at all, a duty traditionally exercised by states.

You missed the "using their stuff to" part of that quote, the opportunities for melodrama here are fewer than you might've hoped.

In some way every service you could offer anyone is part of the war machine. When you operate a daycare, you're taking care of somebody's kids, and that somebody either does something for, or is involved with the military. One of the actions typically carried out by an active military is killing people.

Somebody is always using the products of your labour, in some fashion or level of abstraction, to kill people in whatever way their outfit, organization, or mandate deems appropriate.

And if we're to be more honest about what's going on here, you could just as easily say that Google is helping the U.S. government not kill people they never intended to kill. Also, I think it's rather reasonable to assume that the U.S. government's interest in continuing the drone program, or the military operations it replaces, is largely independent of Google's willingness to help them with targeting, which to me indicates that Google has something of a moral obligation to help make that happen if they're called upon to do so.

> I must have missed the democratic discussion about private businesses assisting in killing people at all, a duty traditionally exercised by states.

I agree with you, but I have to point out that nearly all of our wars are motivated to make the world safe for American business. The killing fields run red so that we can mine the green shoots. The last time we voted on military action was right after 9/11, where we gave permission for an unending world-wide war and toppled two countries outright and publicly.

Many defense contractors made a killing off these shenanigans.

EDIT: Connecting this history to now, recall https://wikileaks.org/google-is-not-what-it-seems/

Paradoxically, Google getting involved has the potential, if not the actual goal, of decreasing the amount of mistaken or misidentified targets.

I understand some "conscientious" Googlers feeling uneasy about Google getting involved in this sort of thing, but paradoxically, Google (or other AI resources) getting involved will very likely result in fewer incidents of hitting the wrong targets. In other words a measurable reduction in "collateral casualties/damage"

The problem isn't identifying targets. The problem is not killing civilians. Increasing the confidence of identifying targets may actually make things worse by allowing to justify drone strikes that would otherwise have to be called off.

The military likely won't be interested in identifying civilians. They have a list of people to kill and want to know whether one of them is in the target area. If they are and everybody else qualifies as a potential enemy combatant (i.e. they look somewhat male, adult and able-bodied) the drone goes kaboom.

Improving detection won't reduce casualties, it will increase confidence, which will result in more strikes and therefore more casualties. There's no incentive to improve the detection of civilians when everyone who is not obviously a civilian (e.g. children) can be considered a combatant.

This technology won't result in drone strikes being called off because there are civilians in the area. This will result in more drone strikes being made because they can ask a machine to greenlight them instead of having any personal accountability.

Our drones are already exceedingly good at hitting the targets they intend to. The problem is that most of the targets they are trying to hit are innocent.

That's questionable:

https://theintercept.com/document/2015/10/14/operation-hayma...

At the very least, they are not yet good at hitting only the intended target.

Wouldn't better analysis provided by [AI-]assisted intelligence result in better identification of what things are the correct would-be-targets?

I don't mean getting better at hitting coordinate x,y,z on the mark better but rather identifying which coordinates are actually "real" war targets [munitions factory, command centers, training centers, etc.]

I mean, we literally target community leaders whose sole interaction with terrorists is making sure their village/families aren't in the crossfire. It's not that we don't know this, it's just that we're America, and we don't actually care.

Suffice to say, if Google was deciding who to shoot at, and it came up with a smaller number than they do now, the US would consider that bad design.

I would believe that one of our military's lesser goals (obvs winning and preventing our deaths are the major goals) is to avoid collateral damage.

We are not Hezbollah or Russia or Assad. None of them care much about collateral damage. Pretty sure a few of the above believe collateral damage sows fear and provides them advantage, if short-term.

Here's the thing, even if we magically only killed the people we want to kill, most of whom probably are innocent, and none of which were given a fair trial: We actually still create more violence.

Right now, to a kid growing up in many of these countries, the face of America is our President (...) and the sound of an American drone on the horizon. Now, one of these days, one of those drones pops his dad. In twenty years, that kid is a terrorist.

Our drone program is basically, the exact thing you'd do if you wanted to guarantee that there'd still be terrorists in the Middle East in two decades, because we're practically founding those terrorist groups right now.

What's the alternative? Let them play out their wars in their own style (aka more Assad-style conflicts but with Iran and Saudi as well as Turkey and Israel doing the business of war)?

While I think GWB and Obama both did us a great disservice in stirring up things in the ME. We also cannot just let them "duke it out". It'd be 10x-50x worse.

Pulling out now --would exacerbate the issue incredibly. Only people on the extreme right and extreme left believe that to be our best option.

We killed many, many Germans and many Japanese, in indiscriminate carpet bombing campaigns. Not many of those kids with "dead dads" became terrorists.

The USSR did some pretty nasty things in Afghanistan. You don't have Afghani terrorists going over to Russia and getting theirs.

You'd believe based on what? US doesn't do any on-the-ground post airstrike investigations most of the time, talking to the affected people. So it can't even know what is collateral damage and what is not.

Really, I'd like to see how would a feedback loop that would make US military minimize civcas work in the real world with the current way these things are evaluated by the military.

US had for a long time literaly two people dedicated to processing the hunderds of their civilian casualty incidents reports until recently. They have backlog of a year or so of reports. So after killing thousands of people in Raqqa and leveling the City, US military hasn't even started processing complaints. The operation is over for 5 months and official casualty numbers are like "perhaps we might have killed 2 civilians". "Sorry we have backlog." Meanwhile, hunderd corpses are still being extracted monthly from the leveled buildings in the city months later.

Still, it's mildly better than the idiotic european allies who still maintain that they killed no civilian after droping thousands of ~200kg bombs since 2014. It's really that demented.

It may be a stated goal to minimize collateral damage. But with this crap going on, I don't believe it has any meaning. US military probably employs 100x more people to clean the fucking toilets than to process the civcas reports. Lip service and all.

Wait, what makes you think the US is using drone strikes against "real" war targets? The targets of most drone strikes aren't a formal military, they're "militants".

The problem isn't finding locations to bomb. The problem is being able to tell whether the blob on the screen is the target you want to kill. If any civilians get killed in the process, they're probably terrorists too. If they're identifiably not, well too bad they decided to hang around in a "war zone".

And killing a lot more random unidentified bystanders in doing so.

That's not a standard you should feel proud of.

> kill people incorrectly

"Patches welcome!"

>I must have missed the democratic discussion about private businesses assisting in killing people at all, a duty traditionally exercised by states.

...what? This is America. Private industry makes all killing machines. Last I checked, Colt, Boeing, Lockheed Martin - all private companies.

I sure hope he meant ‘incorrectly kill people’ otherwise that’s pretty macabre

From what I've gathered of the context I think the original presentation is correct. I read it as a joke criticizing an implied absurdity of tech groups being concerned specifically with what we use to kill people rather than whether we should be killing those people to begin with.

> I must have missed the democratic discussion about private businesses assisting in killing people at all, a duty traditionally exercised by states.

Are you from US? You might not be. Sounds like are not aware of the multi-billion dollar (maybe even trillion dollar) military industrial complex.

Also not sure what is meant by "democratic discussion". Who are the constituents here? Everyone, employees, tech people only, HN users?

> kill people 'incorrectly'.

Right. I think that tells us where Google stands. It is already helping kill people, it just helps kill them "correctly" of course. In some kind of a nice, non-evil way presumably

I think what Shmidt is saying is that the concern is about the military using this tech to essentially go out there and kill innocent people.

Yes, the problem of killing innocents via flawed algorithm is noted in this article https://arstechnica.com/information-technology/2016/02/the-n...

Personally, my concern is who or what is defining 'innocent'.

This is really NOTHING. Pretty much everything used by the armed forces are sourced from private companies.

The Dutch East India Company was a private business...

>I must have missed the democratic discussion about private businesses assisting in killing people at all, a duty traditionally exercised by states.

I'm not sure if you're criticizing that this happens at all or if you're criticizing it as a new development. Either way, it's generally been actual killing that's reserved for the state (or its mercenaries). Private organizations building and refining weapons for the state has been a thing in America since the old west.

I think both discussions are worth having. The involvement of private contractors in military affairs seems to have amplified over the last few decades especially as war has become more and more focused on technology.

Then there's I think a unique angle to this specific case. If you're going to work for Lockheed Martin or Blackwater you at least know what you're getting into. Google does not present itself as a military contractor. Did everybody who works at Google really know that their code is used for this? Have they been informed, consulted?

There's something especially weird and shady about the fact that someone writes some tensorflow code for image recognition, goes and gets a smoothy from the office bar while the DoD just hooks itself into the API and bombs the hell out of people at the other end of the world.

That's a lot more opaque and blurs the line between civil and defense work in entirely new ways.

i'm not sure i agree.

google's cloud couldn't host this customer's production application because (afaik) google can't host classified data.

so then you're talking about the fact that your tech is inherently "dual-use." but that's almost anything in computing, absent a license that prevents it. in the early days of postgres, we were a little weirded out by some of the support requests we got [0], given the university prototype nature of the system.

[0] https://www.paulaoki.com/.admin/pgapps.html

I agree that there's definately a need for more dialogue.

>Then there's I think a unique angle to this specific case. If you're going to work for Lockheed Martin or Blackwater you at least know what you're getting into. Google does not present itself as a military contractor. Did everybody who works at Google really know that their code is used for this? Have they been informed, consulted?

When I was reading the article this part jumped out at me too. It's interesting because it's a debate that ultimately focuses on workers rights vs. the rights of the employer. A callous argument could be made that the political views of the employees shouldn't matter because they're not being paid to be policy experts they're paid to make a product. The counter argument of course would be that, even if the first argument is true, an employee who isn't informed about the nature of their work lacks adequate understanding to decide if they feel strongly enough to quit or if they want to take the job in the first place.