Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

AI at Google: our principles

blog.google

Post reply on HN
Print view View original

241–250 of 418 posts

Re: AI at Google: our principles

#241

Post by joshuamorton »

rmk wrote: post #233
joshuamorton wrote: post #214

Earlier quoted context omitted.

>itself to be superior to it I'm still confused as to how "we don't want to work on systems that kill people" is in any way "superiority". In fact, reading your comments you seem to suggest that any entity (any individual even!) who has a moral objection to working on military technology, but who uses any modern technology believes themselves superior to the public. But that description describes a large minority (or…

In my comments, I repeatedly say that individuals should not be compelled to work on military technology, except in times of war. Hardly confusing. > In fact, reading your comments you seem to suggest that any entity (any individual even!) who has a moral objection to working on military technology, but who uses any modern technology believes themselves superior to the public. But that description describes a large m…

>Hardly confusing.

Except that much of what you're saying implies that you do think these companies should be regulated in a way that forces them to do this. Or at least, if that's not what you're saying, then you seem to be insinuating a whole lot for no apparent purpose. This is why I'm confused. Your stated words and actions (ie the rest of your words) don't appear to match up.

>Google doesn't just use this technology like you or me, it benefits,

Are you suggesting that you and I don't benefit immensely from the internet?

If I'm understand you correctly, you're saying that it is unethical for an entity to benefit from another entity without supporting it. That is to say, its unethical for Google to benefit from the military's technology without also supporting the military.

Ignoring, for a moment that there's a whole host of debate on whether or not that's even true to begin with, such an objection applies equally to any individual as well as Google. I, personally, benefit greatly from military technology. Is it unethical for me to refuse to work on drone warfare? It seems odd for you to say yes to that, but on the other hand, that's basically what you're saying about Google.

>I do not think that description applies to much of the public, which lives outside the SV bubble. But I will concede that this is something that's debatable.

I would consider more than a third of the US population to be "much of" [1].

> suffers from the hubris that they can simultaneously benefit (i.e., make billions of dollars) off the fruits of research that's quite explicitly geared toward military technology, and they can rebuff those very same benefactors without consequences

But again, except perhaps in terms of scale, this applies to anyone. You and I both benefit, significantly, from military technology, both in terms of safety and quality of life. Yet you've stated that we should not be compelled to give back.

Why should Google (or any other corporation, which again, is really just a set of individuals) be treated differently?

[1]: https://www.politico.com/story/2015/05/poll-support-drone-st...

Re: AI at Google: our principles

#242

Post by cpeterso »

foolinaround wrote: post #62

> Avoid creating or reinforcing unfair bias. AI will likely reflect the bias of its training set, which likely reflects the bias of the creators. So, it is fair to say that AI will be biased?

There is definitely a risk of machines learning to reinforce existing but undesirable bias in real-world training data. There is research into documenting forms of bias so they can be recognized and countered when selecting training data.

https://sloanreview.mit.edu/article/the-risk-of-machine-lear...

https://www.entrepreneur.com/article/279927

Re: AI at Google: our principles

#243

Post by killjoywashere »

marricks wrote: post #200
killjoywashere wrote: post #91

Earlier quoted context omitted.

The emails you're aware of framed it in those terms. Do you think a leaker might selectively leak with the intent to paint a particular picture? As someone who has helped write policy, it is literally impossible to write up all policy in advance. You always end up weathering some angry insights in the comment section (we used to call it the peanut gallery). If you can write out all policy a priori, every CEO, economi…

Why would your default stance be to believe a company over a leaker? The company has millions to gain from the contract and hasn’t shown morals on this issue. The leaker has so much to lose by releasing the documents, everything from their career to a significant portion of their life. You could call that incentive to deceive, but I call it incentive to be just about their leak. Especially when it’d be so easy for th…

> Why would your default stance be to believe a company over a leaker?

Because Sundar Pichai has a strong public track record. Every CEO ends up with warts, but I have some sense of what he's about. The leaker, I have zero information on. Given known vs unknown, I put more faith in the known. Whether I by default believe or disbelieve depends on who's saying what.

Re: AI at Google: our principles

#244

Post by LambdaComplex »

ergothus wrote: post #39
finnthehuman wrote: post #18

>2. Avoid creating or reinforcing unfair bias. They DO realize that the YouTube recommendation algorithm is a political bias reinforcement machine, right? Like, I think it’s fun to talk trash on google because they’re in an increably powerful position, but this one isn’t even banter.

FWIW, since they updated the google news algorithm, I've started getting a pretty diverse set of results. It might be because it picked up that I try to read opposing views, or it might be because the algorithm is terrible, or any other number of reasons, but regardless of the source it's refreshing to see. ...and revealing! The headlines differences between, say, Washington Post, CNN, The Hill, and Fox News for the…

>The headlines differences between, say, Washington Post, CNN, The Hill, and Fox News for the same news blurb is even more dramatic than I'd expect when you put them up side-by-side.

Can you give a concrete example of this? I'm definitely interested in seeing how big the differences are

Re: AI at Google: our principles

#245

Post by skybrian »

kanox wrote: post #171
bobcostas55 wrote: post #37

>Avoid creating or reinforcing unfair bias. I recommend _The impossibility of “fairness”: a generalized impossibility result for decisions_[0] and _Inherent Trade-Offs in the Fair Determination of Risk Scores_[1] [0] https://arxiv.org/pdf/1707.01195.pdf [1] https://arxiv.org/pdf/1609.05807v1.pdf

> Avoid creating or reinforcing unfair bias. This very high up and is written in a way which would explicitly allow "fair bias". This means activists will have a free hand to use their positions at Google to enforce their vision of political orthodoxy.

A search engine with no biases whatsoever is useless. The whole point is to be biased towards articles that users find more useful, not to give a random sample of the Internet (whatever that means).

I'm sure there will be internal debate over what biases are good ones to keep and nobody gets a free hand. But as a policy, it doesn't restrict Google's options very much.

Re: AI at Google: our principles

#246

Post by kerng »

This seems like a PR stunt, but at least something. Nothing prevents them and reverting those newly found principles over time... similar with removing "Don't be Evil" from their mission - which kinda would have covered that. Google's goal is to make money and that's what this is about.

Re: AI at Google: our principles

#247

Post by killjoywashere »

djrogers wrote: post #178
hueving wrote: post #144

Earlier quoted context omitted.

This is a bit dramatic though because these robots can't reload themselves. Automated targeting for a single area is miles away from planning strategy, deploying battle groups, etc.

> This is a bit dramatic though because these robots can't reload themselves Yet...

Seriously, I've already got three ideas cooking on how that could be done. That's essentially the same skill as landing.

Re: AI at Google: our principles

#248

Post by lev99 »

oh_sigh wrote: post #216
lev99 wrote: post #118

Earlier quoted context omitted.

An argument that relays heavily on racial prejudice and emotional appeal is philosophically underdeveloped.

There was no argument - I was only expressing my feelings. And what racial prejudice is there?

61% of Americans have unfavorable views of Islam. [0] Islamophobia is prevalent across the western world.

[0] https://www.brookings.edu/blog/markaz/2015/12/09/what-americ...

Re: AI at Google: our principles

#249

Post by jacquesm »

killjoywashere wrote: post #91
EpicEng wrote: post #55

So, I'm all for giving someone the benefit of the doubt if they have a change of heart upon reconsidering an issue, but this coming after the fact rings a bit hollow to me. I think the only principle at play here is that it became a PR issue. That's fine, but let's be honest about it. Early emails between Google execs framed this project only in terms of revenue and potential PR backlash. As far as we're aware, there…

The emails you're aware of framed it in those terms. Do you think a leaker might selectively leak with the intent to paint a particular picture? As someone who has helped write policy, it is literally impossible to write up all policy in advance. You always end up weathering some angry insights in the comment section (we used to call it the peanut gallery). If you can write out all policy a priori, every CEO, economi…

While selectively leaking certain emails and withholding other might color an issue it won't make a negative into a positive. And if the leaked emails aren't genuine I have not seen any claims to that effect. So either they are real and they paint a real, possibly distorted picture or they're false but as it sits right now they are the evidence that people use to build their opinion on. If there is an email trail that establishes a different narrative Google is of course perfectly free to release that too to offer some counter weight.

Re: AI at Google: our principles

#250

Post by gowld »

ma2rten wrote: post #222
EpicEng wrote: post #55

So, I'm all for giving someone the benefit of the doubt if they have a change of heart upon reconsidering an issue, but this coming after the fact rings a bit hollow to me. I think the only principle at play here is that it became a PR issue. That's fine, but let's be honest about it. Early emails between Google execs framed this project only in terms of revenue and potential PR backlash. As far as we're aware, there…

The article you linked says the email was written by an exec, but the source article in the intercept said it was written by "a member of the sales team". It also says Fei-Fei Li, arguably the only exec on the email chain, was actually against it: “I don’t know what would happen if the media starts picking up a theme that Google is secretly building AI weapons or AI technologies to enable weapons for the Defense indu…

https://www.nytimes.com/2018/05/30/technology/google-project...

That article adds some pre-context to the quote above.

> "Avoid at ALL COSTS any mention or implication of AI,” she wrote [...] . “Weaponized AI is probably one of the most sensitized topics of AI — if not THE most. This is red meat to the media to find all ways to damage Google.”

Post reply on HN