I Spent Two Years Training AI to Be Safer and Now I Hate AI

I worked at a company to train AI on trust and safety, and these are my thoughts:

Preface

In this post, I want to share my parting message with my team of AI trainers.

After fifteen months on a wacky, grueling, and somewhat fruitful trust and safety project, our client has decided to end their collaboration with us. I'm told this was purely a financial consideration and that the client was very satisfied with our work.

As for the details of the company and the client, I'm not at liberty to say. I'll keep this post as broad and as theoretical as possible as a result. However, I think this message does an adequate job of summing up my thoughts, feelings, and general outlook about the industry based on my time in the bowels of it.

Of course, I have much more to say, and I do plan to expand on these thoughts in the future. Today, however, I'll just leave the message more or less as I shared it with my coworkers - for posterity's sake.

Farewell For Now! A Thank You and Some Final Thoughts~

Well, I suppose I'll share some parting words, considering I've been on this project for well over a year now.

You all may not have seen me much - I stayed mostly quiet throughout this journey, and I tend to keep my head down. But I've been here all along, and I've been moved and inspired by so many of you. I want to thank everybody for their kindness, enthusiasm, openness, tolerance, and just plain awesomeness! There are so many cool people on this project, it's actually bonkers.

The leadership has also been fantastic - easily the best management of any team I've ever worked with, not just at this company but in any job. I feel truly cared for by everyone involved, and that means so much to me.

---

All that said, my message is bittersweet.

To put it bluntly, this past year has been one of the most grueling times of my life. The world hasn't gone in the direction I'd hoped it would, or indeed in any direction that I could have even conceived of. All the war, racism, bigotry, sexism - the clamping down on individual freedoms, on expression, the doubling down on old sources of power and legitimacy that derive their control from the subjugation of those beneath them - it's exhausting, and I say that from a place of relative affluence and prestige in this brutally unfair global system. I honestly can't imagine what it's like for those less fortunate.

It's been a crazy year, and so much has changed.

On top of that, I've witnessed some of the most profoundly disturbing things on this project, and - believe it or not - I'm not referring to the heinous content we've had to sift through and root out (although, yeah, that was also unbelievably disturbing, too).

No, what I'm talking about is the transformation of the human soul, every part of it, into atomized components of a machine. Not just the machine of AI, but the entire apparatus - we trainers just spent all those months of our precious lives feeding soulless tech that seeks only to make us obsolete; the consumers we're trying to keep safe are being marketed a poor imitation of the most human of impulses - sex, hunger, and friendship - all while those allegedly steering the ship have been usurped by the demands of the market and the whims of capital.

What Have We Done?

I hate to be a downer, I really do. It's just hard to look around at the world and at the things we've built and feel that it was worthwhile. How do we make sense of it, this peeling away of the human psyche - the stripping down of the most fundamental joys of life only to repackage them as inauthentic products to be paid for, either with a forever subscription or in exchange for intimate knowledge of your behaviors, desires, and background that will be used to pinpoint more potential products to us?

Mark Zuckerberg said the quiet part out loud when, in an interview, he pointed out that

"The average American has fewer than three friends, fewer than three people they would consider friends. And the average person has demand for meaningfully more. I think it's something like 15 friends or something."

To put that into context, he was talking about his vision for the future of AI. Think about that; in his eyes, we are just consumers with rational demands to be fulfilled by an AI-saturated market. And friendship is just another one of those "demands."

I don't know about you, but I don't "have demand" for X number of friends - I simply enjoy fulfilling my innate need to be a social mammal, to connect with the people around me, to build things together. To make music, to make art, to discuss the difficult things in life, to gossip, to share joy, to play...

None of this is a matter of economic demand - it's a fact of mere existence. And the reality that most people in the modern, developed world are brutally lonely is not an issue of economic supply but of the perverse incentives that govern our world at the highest levels of political and economic leadership.

There's something very, very wrong with the world if the most basic, lived realities of social need are subsumed by the all-powerful logic of the market.

---

When I tell people I work with AI, they often give me that look. You know the one.

"Oh..." they always say with a phony smile, trying to muster some enthusiasm, but the hollowness in their apprehensive eyes gives it all away.

To be fair, I don't blame anybody for reacting this way. In fact, I agree. It's no wonder, considering that everybody running AI companies and championing the tech talk like Zuck and Musk, awkward and powerful men with riches beyond comprehension who could not be more out of touch with their fellow Earthlings.

To defend myself, I usually say something like, "I know, I know, but I work in the least bad area of AI: trust and safety. Our job is to protect people, especially children, from harmful content." That usually reassures them.

But there's always a small voice in my heart, the skeptic, that isn't satisfied. Basically, it says, I'm doing damage control.

Imagine we're all on a big boat headed toward a sea of ice. The captain is desperately trying to plot a safe course through the ice, while the crew frantically prepares the passengers for potential impacts, administers lackluster medical care for those who've been injured or panicked, and reinforces the hull with large sheets of metal.

In a chaotic frenzy, we brace for the terrifying journey ahead, hoping and praying we've done enough.

But we can just stop the boat.

Turn off the engine. Turn around. Go home to the world we love.

We don't have to sail into a sea of ice. What could possibly be on the other side that could justify knowingly throwing everything away? And why should we trust a captain who sees us only as the raw material of an inhuman machine?

How did we get here?

---

The Luddite in me says we should just turn off the machine. If we did, I think we'd all wake up from the episode feeling a great deal of relief and a fair amount of confusion. Like, "What on earth were we thinking? Who thought that whole AI business was a good idea??"

But the realist in me says technologies, whether mechanical, digital, social, economic, political, or what-have-you, are easily invented but near-impossible to un-invent. And this cat is out of the bag...

Because we're not really on a boat, are we? Really, we're careening down a mountainside on a bike with no brakes. Stopping isn't an option, and the world we left behind at the top might as well not exist anymore. All we can do is try our best to stay upright, bank the curves, avoid oncoming traffic, try not to hit any children...

And when the time inevitably comes that we crash, and it is coming - we just have to make sure we crash right. We have to make sure that we'll be able to get up again, brush off the asphalt and the dust and the blood, and maybe pick a flatter route next time.

The Future

I tend to be an optimist about possible futures, a cynic about the future that will probably come about, and a Stoic about the present.

When I look around at this team, I see all the best parts of being alive. I see intelligence in droves, but also charisma and heart. I see determination and cunning, but also care, empathy, and genuine compassion. I see humor, joy, and solidarity, even though sometimes these are dappled with sadness, grief, or frustration.

There are so many wonderful aspects about being alive, and we humans have so much capacity for good. I see that good in this team, through and through.

Although I generally can't bring myself to embrace the delusion that things really are going to get better in my lifetime, I find some solace in believing wholeheartedly that a better world is at least possible. And I know it's possible because I see little hints of it in the wonderful people around me, including and especially you all. In the meantime, and regardless of how it all turns out, I just focus on doing the best I can, living in accordance with my values, and trying not to worry too much about the things I can't control.

I Think I'm Done Training AI

As Jean-Paul Sartre liked to teach, and I'm paraphrasing here, we all have no choice but to choose projects for ourselves. But although we get to choose the project and how much of our souls we put into it, the outcome is utterly beyond our control. So why should our happiness be based on the outcome, which isn't up to us, when it could be based on our efforts, which are under our control?

As long as we do our part and believe in what we're doing, we can live worthwhile lives. And all the better if we can find like-minded souls to go it together.

And just maybe, let's try to stop the boat anyway? Or test the bicycle brakes one more time before we accept the inevitability of a crash? We're better than this.

So I want to thank you all for enriching my life throughout this time. It's been an honor and a privilege, and I wish you all the best! Goodbye for now~