ChatGPT, OpenAI’s flagship AI product, has captured public imagination about artificial intelligence and beckons the question, “Are we in the AI age?” And, if so, what are the new costs of this new age?
Reaching 1 million active users within one week of launching in November 2022, ChatGPT reached 100 million active users less than 2 months later. An artificial intelligence innovation has become the fastest growing software app in history.
Seemingly endless articles are guessing, ideating, and pontificating what ChatGPT means for what the future holds. But, few are contemplating what a post-ChatGPT world holds against a pre-ChatGPT world. And, more importantly, what are the costs of ChatGPT’s benefits?
LinkedIn is aflutter with lovestruck techies and entrepreneurs salivating about how ChatGPT-based applications could enhance productivity, output, and save cognitive energy and time (the hallmark red flags to your neighborhood political economist). Certainly, ChatGPT presents a worrisome advancement in the exploits of neoliberal capitalist domination. And, as such, it advances the existing problems of neoliberal capitalist domination.
Paying attention to how we narrate — tell stories and instill beliefs around — AI is important for keeping grounded and sober about what this new normal means for what has been normal. Often, AI is narrated as a salvatory silver bullet that will extinguish the sins and preserve the boons of time immemorial.
Many believe AI will be emancipatory for society. We must be vigilant of being incarcerated (and deluded) by that belief.
In this spirit, I responded on LinkedIn to a post widely circulating touting ChatGPT as a harbinger of free goodies. While ChatGPT certainly benefits those already with the privileges of a cosmopolitan datafied society, it brings enormous costs — beyond the financial — that are important to acknowledge but commonly explained away in service of sustaining the narrative that ChatGPT has an emancipatory potential.
ChatGPT may be “free” to those who enjoy high-speed internet, which tens of millions of Americans still live without, let alone the rest of the world.
The immense, non-financial “costs” of AI are detrimental to all people and threaten human rights like:
ChatGPT is predicated on surveillance brokering, a core practice of surveillance capitalism. Attention-worthy content is baited to commodify users’ privacy into troves of data sold to organizations around the world (of different political persuasions) to further capture users’ attentions and wallets.
Immense carbon costs like water and electricity are required to train and maintain high-parametered LLMs like ChatGPT. Using material resources in such ways is simply unsustainable. The unintended consequences of such resource expenses are incalculable.
While much of AIs like ChatGPT are facilitated immense algorithmic processing, crucial parts of these systems rely on humans to “enrich” and “refine” them through screening and feedback. This includes the dirty work leveraging low-wage workers in places like Kenya and India to perform the harrowing task of removing toxic data featuring harmful content from datasets that ChatGPT are trained on. Humans are exposed to horrific media to make ChatGPT more human-friendly.
ChatGPT is very costly — both financially and otherwise. It is important to acknowledge them, even at the cost of imagination, hope, and hype.
Aarshin Karande writes about AI Ethics & Psychopolitics. He studies at the University of Cambridge — formerly at the London School of Economics, University of Oxford, and UW Bothell. He is also an Indian Classical musician.