Power Concentration
At the heart of most discussions concerning the ethics of AI is the question of power distribution and the fact that decisions about AI are often defined by a few. Likewise, its benefits are concentrated while the downsides often impact those who need support, representation, and economic justice the most. The conundrum is that very specific individuals shape the future of AI with their fears and fixations. They control the capital needed for technological development but are often hesitant to share that power. Given the fact that our future will be drastically shaped by AI, it is telling how control of technical systems is already quasi-monopolized. We have the façade of accessibility through cloud computing and hosting services, but the haunting reality is that the supply of these systems could be ripped from the hands of the public at any moment. There is no apparent solution for this in sight, and labor opportunities and conditions will continue to be controlled by the few for the foreseeable future [56].
Within this context, two issues arise. The first is that many artists and activists are reliant upon the aforementioned services, technological platforms, or research institutions that possess this power. While these opportunities may grant them greater resources than they could access on their own, it also means that they are necessarily in an awkward or precarious position if they attempt to criticize the organizations they are reliant on. Gebru experienced this firsthand when she was fired from Google for developing a project that lacked unbridled positivity toward AI. Her experience, while surreal, is certainly not unique. Sacrifices are frequently made in order to preserve access to the resources necessary for increasing awareness of AI and maintaining its progress.
The secondary issue is exploitative labor, which is often employed in the gathering and labeling of data used to train AI. Astronomical amounts of data are needed to train machines and that usually requires a substantial amount of human labor. This work can be mundane and laborious, and it is frequently outsourced to reduce costs. Despite the significance of their work, these individuals consistently receive marginal compensation and no context for what they are doing. Practices such as this are unsustainable and fail to recognize the importance of gathering data responsibly. Unless the reward and the ethical burden of handling data properly is distributed amongst all workers, the sources of data we rely on so heavily are likely to be flawed at their core and unethical to use.