This article was featured in the One Story to Read Today newsletter. Sign up for it here.
During a significant AI conference in India last Friday, Sam Altman sought to tackle a criticism he labeled “unfair.” The CEO of OpenAI faced a question from The Indian Express about the extensive natural resources needed to develop and operate generative-AI models. Altman swiftly challenged this, acknowledging that chatbots indeed consume considerable power, but then questioned the reporter on the vast resources humans have utilized across their entire evolutionary journey.
“Training a human also demands substantial energy,” Altman informed a large audience. “It involves, for instance, two decades of life and all the sustenance consumed during that period before intelligence develops. Furthermore, it required the extensive evolution of the estimated hundred billion individuals who have ever lived, learning to evade predators and to understand science, among other things, to bring about your existence, and then you utilized whatever you, well, utilized.”
He elaborated, “The appropriate comparison is to consider the energy consumption when a trained ChatGPT model answers a query versus a human responding. And, likely, AI has already achieved comparable energy efficiency when evaluated on that metric.”
Altman’s statements are readily refutable. The energy consumed by the human brain is significantly lower than even advanced AI models for straightforward tasks, not accounting for the laptops and smartphones users employ to interact with AI. It is true that humans require physical nourishment to “get smart,” yet this also functions as a convenient misdirection by Altman; the primary concern with AI isn’t solely its resource demands but its substantial contribution to climate change. Atmospheric carbon dioxide levels are at historical highs, unseen in millions of years. This increase stems not from the evolution of humanity and all other species over millennia, but from contemporary human society and the combustion turbines similar to those OpenAI plans for its Stargate data centers. Other data centers are also constructing private, gas-fired power plants—which collectively could generate enough electricity and emit as much greenhouse gas as numerous major American cities—or are opting to prolong the operational life of coal facilities. (OpenAI, which shares a corporate partnership with the business division of this publication, did not respond to an inquiry regarding Altman’s comments.)
However, the truly noteworthy aspect of Altman’s remarks is his very inclination to draw parallels between chatbots and humans. Such a comparison implies his perception of people and machines as equivalents. This was not an accidental slip of the tongue; rather, it represents a widespread, deliberate stance within the AI sector. Altman delivered a nearly identical assertion to Forbes India during the same AI gathering. Moreover, just a week prior, Dario Amodei—Anthropic’s CEO and Altman’s primary competitor—made a comparable analogy, equating the development of AI models to both human evolution and daily learning processes. This perspective extends into product creation. Anthropic is exploring whether its chatbot, Claude, possesses consciousness or can experience “distress,” and permits Claude to terminate “persistently harmful or abusive” interactions where there are “risks to model welfare”—a clear act of anthropomorphizing a software program that lacks the ability to eat, drink, or exhibit independent volition.
AI companies appear convinced either that their creations genuinely rival humans or that promoting this idea serves as effective marketing. Both scenarios are concerning. A sincere conviction that they are engineering a superior entity, possibly even a deity—Altman, during the same event, predicted superintelligence within a few years—could readily rationalize disregarding humans and the environment as mere side effects. Altman further stated, in addressing energy concerns, that the issue is significant due to “the world is now using so much AI”—thus, societies must “move towards nuclear, or wind and solar, very quickly.” An alternative course of action would be for the AI industry to exercise patience.
Should Altman’s parallel between chatbots and humans merely be a public relations strategy, it reveals a profoundly misanthropic viewpoint. He is targeting investors. The idea that AI laboratories are developing digital life has, naturally, always bolstered their foundational narrative, and OpenAI is reportedly in the midst of securing capital that could value the corporation at over $800 billion—a figure nearly equivalent to Walmart’s market capitalization.
Technology companies might genuinely aspire to create AI tools for the betterment of all humanity, aligning with OpenAI’s initial mission, and truly believe they require significant capital to achieve this. Nevertheless, drawing a comparison between raising a child—or, indeed, the evolution of Homo sapiens—and the creation of algorithmic products starkly reveals that the industry has become disconnected, if it ever was connected, from the essence of being human. To “train a human”—which is to say, to experience life—involves enduring challenges, embracing the potential for setbacks, and occasionally wandering simply to discover marvel and aesthetic appeal. Generative AI, conversely, seeks to bypass this entire process, rendering every endeavor as immediate, streamlined, and effortless as possible. While these tools may prove beneficial to us, placing them on an equivalent standing with organic life is lamentable.