Hacker News Reader

Comment by wcrichton | Hacker News Reader

I’m a PhD student focusing on large scale video analysis. It’s an unfortunate fact that a lot of the applied research in this space is motivated by surveillance, particularly in systems, eg what’s come out of Microsoft’s efforts here [0]. A colleague who attended CVPR said most of the industry booths were either surveillance or self driving cars.

I wish researchers would start to think more creatively about what we can do with large video datasets and tools like face recognition. The least we can do is try to use this technology for social good. For example, a USC/Google team used computer vision tools to identify gender bias in modern cinema [1] (for my own research, we’re doing a similar kind of analysis on TV news at 100x the scale).

[0] https://www.usenix.org/conference/nsdi17/technical-sessions/...

[1] https://www.google.com/about/main/gender-equality-films/

View on HN

I rarely post, but I feel it necessary to chime in here.

I studied computational linguistics/natural language processing in grad school, out of an interest in accessibility issues (I had RSI) and related applications, and had naively neglected to look very hard about how I'd be likely use it in a career.

Other people in my cohort went to work for the CIA. I couldn't move to the bay area for... reasons, (this was pre-Google) and it seemed everyplace I could find anything wanted me to have a security clearance, with all the implications that has.

One could say I'm not mercenary enough, or that I took too many philosophy classes, or any number of other rationalizations, but one way or another, none of the jobs I've taken have had much need for what I studied.

I feel morally obligated to use what I know to improve the state of the world. This has, unfortunately, led to a lot of unemployment and serious depression.

Pseudonym for obvious reasons.

> I feel morally obligated to use what I know to improve the state of the world. This has, unfortunately, led to a lot of unemployment and serious depression.

And as you would suspect, I'm a PhD grad working in the Bay area on mas surveillance AI. Half a million dollars a year.

I honestly don't know how to invest the money anymore. Index funds are boring at this point.

What point was there in posting this?

It seems like what he's suggesting is that selling out can be quite lucrative, and if you can deal with it, there are rewards to be had.

I'm a little confused by the beginning of your response, because you don't appear to be the person I replied to.

When you wrote this, what did you anticipate my reaction would be?

Why?

I must admit, I really don't know what it would be like to be working in the bay area on mass surveillance AI for half a million a year.

It doesn't matter how you invest. You can't protect yourself from karma with money.

Join Kaggle. Plenty of linguistics work to be done there. After a few months you'll have some idea where you can apply this in a workplace that is in line with your morals.

At the very least it will keep you plenty busy and is fun.

It's disheartening that so many bright people work on research that will undoubtedly be abused by authoritarian governments and make the world a worse place for everyone. Ask a thousand people what video analysis can be used for and I bet the vast majority will say surveillance, and most would probably struggle to think about something else.

I wish researchers would start to ask themselves if what they're building can be abused by authoritarian governments, and if so, switch focus to a field of research that will make the world a better, rather than a worse, place to live in.

History will not look kindly on researchers who built the tools that enabled Xi the Pooh to achieve a dystopian surveillance state.

History? You mean Google?

Google will glorify it’s history in it’s middle schools just like America glorifies it’s history in it’s middle schools.

Considering the researchers are hired according to their sheltered backgrounds, holding them as indivuals accountable is futile.

If you care, vote accordingly, for politicians willing to refuse corporate donations.

If you care a lot, study revolutionary history, learn from it and organize. I’m just paraphrasing the only thing that seems to ever actually work.

Google certainly played a major role, and it wasn't my intention to shelter companies/governments from blame. However, I certainly don't believe holding individuals accountable is futile. They need to consider if people can be suppressed, stripped of their rights, imprisoned, and murdered thanks to their research/code.

We should encourage developers to fight for privacy, and discourage them from engaging in research/writing code that can be used to commit evil acts.

And yes, I wholeheartedly agree that everyone should vote.

Google wasn't responsible for any of the major breakthroughs in AI. The current AI boom is driven by deep learning[1], due to Alex Krizhevsky [2][3], who was then a Ph.D. student @ University of Toronto. Subsequently, several significant breakthroughs have been in Universities & outside Google or FAANGs (Siri, DeepMind were all questions)

[1] https://en.wikipedia.org/wiki/Deep_learning [2] https://beamandrew.github.io/deeplearning/2017/02/23/deep_le... [3] https://arxiv.org/abs/1803.01164

> I wish researchers would start to think more creatively about what we can do with large video datasets and tools like face recognition.

Some very smart people in China did exactly that, and arrived at a conclusion on what the local market for surveillance tech was lacking. Based on what the government will fund, they came up with an answer we seem to dislike.

I think he means, aside from dystopian population surveillance

yeah, like a thing that would remind me what peoples names are at parties.

probably not as much money in that though.

you can probably scrape your own Facebook friends' pictures & use the GCS or Azure free services

To be fair, the absolute last thing “modern cinema” needs is another formulaic measurement of demographically correlated screentime.

I am not personally against girls seeing depictions of other girls in movies (have you consulted with the anti-bodyshaming sect?) however this claim to “social good” is telling of a desperate effort to ignore actual social problems, none of which are so conveniently morally distinct from the economic disparities Google instigates and defends on a daily basis. Studies show little girls who are homeless have bigger problems.

Can't wait for algorithm homogenised movies!

I think this is the internal slogan at Netflix :)

I'd say you've seen a bunch of them, like everyone (sigh)

> identify gender bias in modern cinema

Looks like a way to sell more ads.

If you asked them, they would effusively tell you all this surveillance is being used for the social good, and probably even believe it. Not many people are brutally honest about how their work destroys the social fabric, and will always point to the benefit it provides. Pornographers think its a social good if we would just be a little less uptight about sexuality, and more porn access is a solution.

And the same people who try to work for good can find hey've just armed others for evil by giving them tools.

> The least we can do is try to use this technology for social good. For example, a USC/Google team used computer vision tools to identify gender bias in modern cinema

That sounds entirely worthless.

A social good involves helping somebody who needs aid.

This proposal merely fills a perceived need by political dogmatics. This is exactly a sociopolitical memory leak.

Please don't take HN threads into political flamewar.

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

You allow posts about mass surveillance technology yet it's supposed to remain apolitical.

Disgusting

From recent posts I gather that you don't want to use HN as intended, so I've banned this account. If you don't want to be banned, you're welcome to email hn@ycombinator.com and give us reason to believe that you'll follow the rules in the future.

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

I will not, but I also didn't begin by introducing the tinder - he did.

"USC/Google team used computer vision tools to identify gender bias in modern cinema" is waving a flag for a specific political belief that is not shared by most Westerners. Outside of the valley this is fringe stuff.

The appropriate equivalent to this would be a right winger expressing concern at the 'appropriateness' of race mixing on camera (the hidden belief) - perhaps pointing out bias in the form of the black woman/white man duo being rare (potential legitimate issue).

Suppose I applauded the use of computer vision to spot the frequency of black man/white woman pairs - I know you wouldn't consider that a neutral political perspective. tldr; Political neutrality is impossible when your attention is being dragged along by the current prevailing wisdom of the forum.

That's true, but part of the art of posting a good HN comment, or at least not posting a bad one, is to resist reacting to the most provocative thing in a comment (other than possibly to downvote or flag it) and instead react to the most interesting thing, or not react at all. That's not easy for most of us, including me.

Reply on news.ycombinator.com