Accepting Probabilistic Software Is A Political Choice

On Walking Away

Users that have come to accept workflows of probabilistic automation as "useful" may take offense at the output of such work being labelled as "slop." It feels like a fighting word.

On the other end, we have individuals who see the "artificial intelligence" project as a political project, with very specific effects and goals—some that are easy to spot, but some that are far more subtle.

This makes it very difficult to communicate in some sort of conciliatory way. (That is, if it is conciliation that we are looking for to begin with.)

In social media, we tend to look for shorthand ways of expressing our understanding. That's why terms like "slop" or "booster" or "AI" exist to begin with. But those terms are tricky/slippery, especially at a time where linguistic drift seems to be rapidly shifting.

With all that said, I was initially going to reply to this thread on Mastodon, but decided to blog about it instead, as my thoughts kept expanding a bit.

So here goes.

In my opinion, if we are accepting model output into our "open source" projects, that is a political choice. We know that models reflect bias implicit in their training data, we know that they cannot exist without exploitation of land and labor, and we know that they are controlled by some of the most abhorrent people on the planet.

Politics happens when there are more than one individual in play, because the choices of one person invariably affect the other(s). And who gets to choose? Well, that's where power dynamics unfold, and the idea of governance emerges.

Backdrop

Linus Torvalds (creator and lead developer of the Linux kernel) recently made the case on a mailing list that 1) probabilistic software automation tools are useful, 2) if you don't agree, then you haven't used the "tools" correctly (you're holding it wrong), 3) and, since nobody's perfect, we have no reason to reject the adoption of this technology. Oh, and lastly, 4) he makes a distinction about his ideology of making "better technology" as opposed to sustaining some sort of "social warrior" project.

I won't belabor the point about "usefulness." That point was poignantly well-made by tante over on his blog.

I'll add that this is also the point driven home by Emily Bender and Alex Hanna in The AI Con. We should always ask "useful for whom and at what cost"?

But I did want to touch on the last two points I highlighted above.

His appeal to purity is one I've seen (and wrestled with) a lot. It's true that we live in a world full of contradictions. There is no question that many of us accept, and in some ways, actively support corrosive people, groups, industries, and/or political projects.

But that does not mean we absolve ourselves!

It's a slippery slope, isn't it? Well, since no one is perfect, why do anything that is out of our own comfort zone? Why make any action that benefits a collective whole?

If something is useful to me, what is my contract with society?

Perfection as a prerequisite to action is absurdly short-sighted. It attempts to curtail a valid critique with some version of nobody's perfect.

I can appreciate that having a conversation about the ethical and political implications of a widely embraced behavior is hard to do on a mailing list.

But it is still a vapid defense.

If my loved ones were to intervene due to what they observed to be self-damaging behavior, it would be easy to answer with: "You hypocrites. You all do things that are bad for you. Leave me alone."

It might stop the intervention. But they wouldn't be intervening if they didn't care about me to begin with!

Political

Returning to the political angle.

In many ecosystems, such as Science, Academia, Journalism, and yes, Technology, there is often a move to remain apolitical, as it gives a guise of impartiality.

I could spend some time arguing those points, but I'll leave that for the reader to think about or examine on their own (some of you while sweltering in heat caused by global climate change).

Torvalds makes an interesting choice to invoke the social warrior moniker. Again, likely a shorthand way for communicating with a term that is slippery, at best.

That term, as pejorative, is bathed in political baggage and perhaps slithers toward propaganda at this point. (The stories we tell can, and often do, reflect psychological warfare. For more, check out Annallee Newitz's Words Are Weapons.)

It is a strange contradiction to invoke a highly politicized term, followed by the "apolitical" statement that decisions are made primarily based on "technical merit."

I'll admit that at times, critics of probabilistic software automations and their effectiveness in software development conflate the "usefulness" argument with all the externalities mixed in.

This makes it difficult to deal with questions, such as, does the adoption of this technology improve/deteriorate software as a whole?

But that's not the point I'm making here.

In Torvald's email, he makes this bold directive:

Linux is not one of those anti-AI projects, and if somebody has issues with that, they can do the open-source thing and fork it...

Or just walk away.

Similarly...

If you don't harbor the same human-hating beliefs espoused by the leaders of these tech companies, if you're not opposed to strengthening the agency of tech workers, if you are against surveillance technology, if you favor local communities standing up against datacenter buildouts that threaten their ecosystem and livelihood, if you value the mental health of exploited data workers, and hell, even if you dislike the price of RAM... but you still see these "tools" as useful...

Why not also support those who are risking their jobs, their homes, their rights, their dignity—or join them in opposing these tech companies that are spending BILLIONS to ensure that they remain unopposed in their vulgar project to control the capital, the politicians, and the narrative.

Or, you know. Just walk away.