Hacker News Reader

Comment by hpoe | Hacker News Reader

So I've been wondering about these positions like those mentioned in the article, people whose sole job is to prevent bias, fight discrimination and what not. Whether they be ML researchers or corporate diversity officer. What happens to these people when full equality is achieved?

Like let's imagine tomorrow everyone she'd all of their sexist and racist and other ist tendencies both conscious and unconscious. Wouldn't these people be out of a job, and as we continue to be a more and more equal society won't these positions have less and less work Todo? Because if they are doing their job they should be solving the very problem they were brought in to counter.

Is it inevitable that these positions will end up continuing to try to justify greater and greater outrage over small and small incidents? What is the future for these positions and the cottage industry that has sprung up around inclusion and diversity?

View on HN

> What happens to these people when full equality is achieved?

What happens to a software engineer when full feature completeness and bug-free-ness is achieved? What happens to a police employee when zero crime is achieved? What happens to a builder when all the buildings that ever need to be built are built?

Even if you get to the point where you can say "there are no problems at all!" (doubtful, except maybe in tiny startups or similar), adding "and there will never be any more and we never have to do anything ever again" is hopelessly naive.

Are you suggesting full-equality, like zero-crime, is something that doesn't exist? Or that our notion of what equality is evolves? These are different things, e.g. like the difference between the policing of laws, and changes in the law itself.

For the former, then accept a threshold, "reasonable level of equality" such that differences are within the norms of individual diversity. Then equality is achieved.

Or focus on "equality of opportunity" which is much simpler to gauge, as it only requires that opportunity be similar e.g. a person covered by fire insurance, whose house burnt down; had the same insurance "opportunity" as someone covered by the same policy whose house didn't burn down - same opportunity, if different outcomes.

It's clear that some I&D initiatives exist because of pre-existing issues. At the very least, some of it depends on the notion of "race", and that isn't guaranteed to exist forever.

> Are you suggesting full-equality, like zero-crime, is something that doesn't exist? Or that our notion of what equality is evolves?

Both, to an extent. Avoiding problems in a large company is presumably a constant uphill struggle, and certainly society's view on it does evolve (often quite rapidly; I personally went from strictly speaking being outlawed before 1993 to having employment protections in 2000 to being allowed to marry in 2015, for instance).

> as it only requires that opportunity be similar

Even if you restrict it to that, it remains a huge problem for employers. To take the most basic example, there have been _endless_ studies showing that blinding resume review by removing names etc substantially changes outcomes in the very first stage of hiring (by impacting who gets called in for interview).

Blind reviews and auditions are considered problematic now: https://www.nytimes.com/2020/07/16/arts/music/blind-audition...

"What happens to a software engineer when full feature completeness and bug-free-ness is achieved?"

Try working in the games industry all the way to the end of the project and find out for yourself. Your work doesn't even need to be "bug free" but "good enough."

Yeah, the games industry is a bit of an outlier there, granted.

> What happens to these people when full equality is achieved?

Thank you, I needed a laugh this morning. Humans are going to be humans, and will always find ways to treat each other in inventively bad ways.

If you are trying to say that these positions should not exist... I disagree. The study of ethics and the profession is not a new thing or a "cottage industry".

Of course people in these positions are sometimes going to attract outrage and cause conflict - if they didn't, I'd suspect they were just there to "rubber stamp" whatever the companies were doing - yep, all good here, sure, we're doing x, y and z thing that people criticize us for, but our ethics committee says we're all good, so you can all put down your pitchforks and go home now.

It's their job to be a thorn in the side of authority. Sometimes it's justified, sometimes not, there's always room for interpretation and disagreement, some of these people are good at their jobs and some are not - but dismissing the entire thing as "fake outrage to promote their jobs" doesn't seem right.

She's a PhD researcher, not some kind of equality police, so presumably if the world suddenly became perfectly just overnight, she'd continue doing research but in a different area. I imagine all the AI biases that she is known for would still exist in this perfect world, the impacts would just be less disparate.

But this seems sort of like overengineering? I'm not sure we're at any substantial risk of the world suddenly becoming fair, so it seems odd to spend time worrying about what might happen if we run out problems to solve.

I mean, most of the actual work-work Gebru did was investigating to see if a bias existed in certain things. As we do more things or as those things change, that testing will have to be expanded or at least repeated.

An ethical review is like QA, you never really "finish" it unless you stop developing altogether.

I'm not sure this is the best argument against these types of jobs. Food used to be completely unregulated and potentially unsafe, but there are still jobs regulating food safety. It's basically a solved problem, but it still requires lots of people in the public and private sectors who's jobs focus on food safety. Same story with cars, same with medicine. You could imagine a world where bias becomes a mostly solved problem but still requires a lot of people to work on it.

This is a really good analogy.

> It's basically a solved problem

It is, or isn't solved depending on how the problem is stated.

The problem of food policing is solved, not the problem of food needing to be policed. In reality, places are constantly shut down for violations. The deterrent only deters people serving food unsafely, it doesn't kill the motivation - people motivated to serve unsafe food still exist, they just choose not to (or are prevented from doing so).

When police arrest a criminal, we say they did their job - we don't say they failed their job b/c the criminal existed in the first place. Sure, good policing deters criminality, but it's not really the only goal. the true eradication of motivation to commit crime are usually totally different social factor that has nothing to do with the police. In a similar vein, it's not really the fire departments job to prevent fires.

The real question is: is bias the same kind of thing. On that basis I'd ask "what kind of bias" - that is, does D&I coverany kind of human bias, or is it more specific. If you are talking about historical generation biases, for example, that might vary based on societal attitudes in general.

If that was my job, I'd be over the moon that I was successfully making myself redundant. I'd happily find myself a different career, proud that I made a dent in a worthwhile and difficult problem, and fulfillment is much harder to come by than money.

That assumes AI and technology in general will stagnate and never change or grow or create new previously unforeseen technologies. Unless that occurs the job of analysing the ethical and sociological impacts of technology will continue to be relevant.

For Gebru individually, she has a PhD from Stanford's AI Lab and is accomplished in the field outside of ethics. If the utopia you're describing some how occurred in her lifetime she'd have no trouble finding another job.

"It is difficult to get a man to understand something when his salary depends upon his not understanding it."

"Something" and "It" here would be "full equality achieved".

(Not that I am expressing a view on whether "it" could be achieved, just a comment on human nature).

https://science.sciencemag.org/content/360/6396/1465.full

Do we think that a problem persists even when it has become less frequent? Levari et al. show experimentally that when the “signal” a person is searching for becomes rare, the person naturally responds by broadening his or her definition of the signal—and therefore continues to find it even when it is not there. From low-level perception of color to higher-level judgments of ethics, there is a robust tendency for perceptual and judgmental standards to “creep” when they ought not to.

I think that worry is like worrying what infosec engineers will do when full security is achieved, or whether Tesla's Full Self Driving team will be laid off once Full Self Driving has been accomplished.

> What happens to these people when full equality is achieved?

The bubonic plague still kills a few people every year in the United States. Worldwide, cases number in the thousands. Granted, biases and discrimination aren't proper diseases, but they come close given they spread with misinformation.

Isn't an expectation that we can 100% eliminate all that ills us irrational? Doesn't that make this question you ask a leading one?

A panacea for the human condition is not a thing. Suffering exists because of impermanence and will continue to do so through at least next week. That's not to say it's not worth fighting for, but I think any expectation that it can be "eliminated" puts us on the wrong footing for improving things over time.

At the least, it blind us to other truths which may be more evident, such as corporate greed and misuse of the technology we all carry around in our pockets now.

But no more smallpox deaths. It is possible.

Crime free can happen without laws.

> Crime free can happen without laws.

This is true in a very literal sense.

Laws aren't necessary in extremely smaller groups. Do you and your dog need laws?

You and your family probably have rules.

Very small communities don't even have police.

The larger the group the more need for laws to protect each other against ourselves.

What do you mean by law? If you mean a verbal or written rule, then any number of dogs have no laws by virtue of dogs being illiterate.

Presumably, they'll find something else to work on.

> Is it inevitable that these positions will end up continuing to try to justify greater and greater outrage over small and small incidents?

No it's not inevitable. To think it is implies everyone doing these jobs is acting in bad faith or will be corrupted by something unspecified and act in bad faith.

Does a financial incentive (employment) not automatically result in "bad faith"? Why would anyone be trusted in evaluating their own job worth?

No? It's one factor among many. It's probably more telling on the belief system of people that do think everyone else thinks that way.

Who doesn't think that way? I assume your front door has no locks?

https://science.sciencemag.org/content/360/6396/1465.full

And I've been wondering what happens to all the HN posters once every sensible comment has been written. Will they become repetitive? Will they start posting absolutely inane comments? Will they all congregate, post one last comment and celebrate a job well done?

Reply on news.ycombinator.com