Comment by DoreenMichele | Hacker News Reader

Comment by DoreenMichele | Hacker News Reader

Parent Full thread

My general impression of this entire thing, starting with Gebru being fired, is that the department should have never been named "ethical AI" anything.

It's research into social stuff. How AI interacts with things like faces and how that gets biased by a lot of social issues, like racism and existing class stratification related to race (among other things).

These are things that can have "ethical" implications but when you start talking about "Ethical AI," you inadvertently convince people "We are the Good Guys. We will treat everyone in some extremely above board fashion. Our little department is going to single-handedly right all historical racial and gendered wrongs ever in the history of the universe and all the baggage all of society is living with as a consequence."

Good luck with that. It isn't happening and the implied contract there suggesting it will is just going to piss people off and give them unrealistic expectations.

No, I am absolutely not justifying shitty behavior of any sort, not in this case nor in the general case. Don't come at me with ridiculous accusations of that sort.

It's hard to make progress. It's actively counterproductive to inadvertently set expectations that this one department will meet some crazy strict set of ethics at all times every minute of every day in contrast with the entire rest of the world where we duke it out for small gains on a regular basis.

I feel as though people looking joining such a group would have at the very first question “what is the scope of what we will be studying” or something similar. I don’t think people take jobs like that without having a decent answer to that question and I haven’t seen any accusations of bait and switch, only of poor treatment over a disagreeing researcher ballooning into this.

I'm pretty confident the crux of the issue is "We expected to be treated better in this department than what we have had to tolerate elsewhere due to sexism, racism, etc."

And I think they expected that based on "This is the Ethical AI department. We are trying to make the world a better place. You do that by walking the walk first, not just researching how AI works."

That's what I think. It's not something I can prove. For various reasons, I'm not really wanting to debate it. I tossed it out there because some folks know I post as openly female here (and HN skews rather strongly male) and some folks know I have studied social stuff and sometimes people find my take on certain subjects of value or at least food for thought.

Thank you for the perspective.

Assuming you are correct, is there a way to reset expectations without blowing up the department? Not that we shouldn't be striving to be ethical, but it is clear that Google doesn't like the direction the team is heading.

If Google is truly interested in the thorny issues of making AI as ethical as possible, can it do so without tackling the (seemingly insurmountable) issue of social ethics? And if not, do we just throw in the towel and cross our fingers that AI doesn't hurt us too much?

I would start by renaming the department to reflect the idea that the mission is to explore and understand how and why things go wrong socially with the deployment of so-called AI.

I don't know what that name should be. I've had maybe two hours of sleep and this is likely a thing that would require research.

I've had college classes in Social Psychology, Negotiation and Conflict Management and Homelessness and Public Policy. All three of those classes deal with topics you could think of as falling under the umbrella of human ethics. None of them has that word in the title.

You need to first and foremost understand how things happen. You need to understand that before making value judgements concerning ethics.

Speaking of ethics in AI before you even know how and why you get certain outcomes amounts to putting the cart before the horse.

Seeking to understand how things happen for thorny topics inevitably raises hackles.

I was molested and raped as a child. I speak somewhat often on topics like rape and there is a long history of people accusing me of being a rape apologist.

I would like to see rape happen less often in the world. That goal requires one to understand what goes wrong socially without first assuming "Men are just all rapey bastards with hearts of pure evil who do evil things intentionally, knowingly and on purpose."

A lot of people who have already been hurt by the current state of things don't want to understand what happened. They want "justice" by which they typically mean revenge.

You will be facing that same dynamic for AI and issues of racial justice, gendered issues, etc.

Many people will call you a White supremacist or a misogynist for trying to understand what went wrong as neutrally as possible.

But that is the only useful first step for actually redressing a lot of social ills.

Fiction sometimes covers this. People consume it, get jazzed at feeling "their side" understood for once and then routinely fail to actually apply those lessons in life.

Fiction sometimes gives you both sides of a story and lets you see how a+b went kaboom although neither was inherently bad. Research needs to seek that perspective.

Ethics occurs at the point where you apply the research. Simply taking a look at how it happens will not be about ethics and will, in fact, make most people pretty darn uncomfortable.

Reply on news.ycombinator.com