Everywhere Foist Upon Us - The Darth Mall
We live in capitalism, its power seems inescapable — but then, so<br>did the divine right of kings.
Ursula K. Le Guin
When I began writing this, I intended to make the case that there is<br>no responsible use of “AI”. I was motivated by three<br>separate encounters in a single morning of people arguing that the<br>path forward is to adopt “AI” and find a way to do so<br>responsibly.123 In<br>the course of writing, however, I discovered that my real disagreement<br>with this position was that it is predicated on the belief that<br>“AI” is here to stay, that it is inevitable.
I loathe the term “AI” because it is so broad as to<br>be practically meaningless. There is obviously a difference<br>between using ChatGPT to read or write your emails versus using<br>machine translation software to translate something into a<br>language you can read, although both can be said to use<br>“AI”. Hence my insistence on using scare quotes<br>whenever I use the term.
As I began writing, I realized that the arguments I make for why there<br>is no responsible use of “AI” can just as easily be<br>applied to many other tools in our lives — for example, cars,<br>computers, and phones — and as I began to address the hypothetical<br>what-about-ism, I realized that the fundamental difference was that<br>those arguing for “responsible ‘AI’ adoption” had<br>accepted “AI” as inevitable and I had not. Even though<br>you could argue that your laptop, for example, carries with it the<br>same environmental harms and labor exploitation of which these<br>“AI” tools are guilty, it is harder to exist in the<br>modern world without using a computer. Except in cases where your boss<br>forces you to use “AI” or be fired, it is entirely<br>possible to eschew the use of these tools today.
This is not to say that I think that the environmental harms from<br>resource extraction associated with computer chips and battery<br>technology, nor the exploitation of labor in countries like China is<br>in any way acceptable or excusable simply because some electronics<br>have infiltrated our lives to the extent that they are practically<br>indispensable. We can do better — and we should. My point is that we<br>have not yet reached the point with “AI” that we must<br>continue to use it as we demand better. We can instead refuse to use<br>it until it is ethical to do so.
Fundamentally irresponsible
No matter how diligently you check the outputs, or how cautiously you<br>roll out “AI” features, these tools are built and<br>operated in a way that is unethical and irresponsible. Using them,<br>therefore, can be neither ethical nor responsible.<br>It doesn’t matter if it works.
Training and running these enormous models consumes so much<br>electricity that we are<br>keeping coal-fired power plants online<br>to meet demand. We are facing a serious climate crisis; the only way<br>to stop global warming and stave off catastrophe for many of the<br>people on the planet is to cut carbon emissions. It strikes me as<br>utterly irresponsible to pump more carbon into the atmosphere just so<br>you don’t have to write your own emails or unit tests.
The models are trained by<br>exploiting laborers in the global south<br>and subjecting them to trauma. In what world is that ethical?
And many of these tools are explicitly part of a project to<br>undermine workers and consolidate more power for capitalists. Even if your job is not at risk (for now) and you get to<br>be a centaur<br>using “AI”, the moral thing to do is stand in solidarity<br>with your friends, family, and neighbors who aren’t so lucky.
So, no, there is no “responsible use of ‘AI’”, no matter<br>how careful you are in your application of the tools.
Ubiquity is not inevitability
“AI” may seem inevitable because it is everywhere foist<br>upon us. It feels inescapable. I have never knowingly used any of<br>these large language model-based tools, and yet they are everywhere in<br>my life. Practically every day I encounter a new story about<br>fallout from professionals<br>treating chatbots like<br>Electric Monks.4 It<br>infects the software on which I rely. And I have friends that have embraced these tools whole-heartedly<br>and, man, that can make it hard to have conversations with them<br>sometimes.
It may also feel inevitable because this is just<br>the same tired play that capitalists have employed since the<br>beginning of capitalism: use technology to devalue and deskill labor to lower costs and<br>increase profit for the rich. The fact that we’ve been down this road<br>before — this familiar feeling — makes it seem inevitable. We’ve seen<br>this movie, we know how it ends. Except that it hasn’t ended yet.
Ubiquity and familiarity do not equate to inevitability. The industry<br>is<br>unsustainable both financially and computationally. We’re beginning to see these companies restrict access to their<br>products as they increase prices.5<br>On top of which, they are having trouble buying enough computer<br>memory, storage, and graphics cards to scale up their models. Many<br>companies who replaced huge swaths of staff with “AI” are<br>now<br>quietly re-hiring those...