Sam Altman’s Troubling Comparison of AI and Humans | The Atlantic

OpenAI CEO Sam Altman sparked debate last week at an AI summit in India by suggesting a comparison between the resources required to train chatbots and those needed to “train a human.” Altman, responding to a question from The Indian Express about the energy consumption of generative AI models, argued that the energy expenditure of developing human intelligence is often overlooked.

“It also takes a lot of energy to train a human,” Altman told attendees. He posited that 20 years of life, coupled with the food consumed during that period, represent a significant energy investment before an individual reaches a level of intelligence. He further expanded the scope to include the evolutionary history of humanity, suggesting that the development of intelligence in 100 billion people who came before us also required substantial resources.

Altman continued by framing the comparison in terms of energy efficiency. He argued that, once trained, a chatbot may already be as energy-efficient as a human when answering a question. However, this claim has been challenged, with experts noting that the energy used by the human brain is considerably less than that required by even the most efficient AI models, especially when factoring in the devices used to interact with those models.

While acknowledging the energy consumption of AI, Altman’s comments were seen by some as a redirection. The core concern surrounding AI, according to critics, isn’t necessarily the resources it demands, but its contribution to climate change. Atmospheric carbon dioxide levels are currently at levels not seen in millions of years, driven not by human evolution but by contemporary society and the increasing energy demands of data centers like those operated by OpenAI.

The significance of Altman’s remarks, however, extends beyond the debate over energy consumption. As The Atlantic points out, the very act of comparing chatbots to humans suggests a view of machines and people on equal terms. This perspective is not unique to Altman. Dario Amodei, CEO of Anthropic, Altman’s main competitor, recently made a similar analogy, likening AI training to human evolution, and learning.

This mindset appears to be influencing product development within AI firms. Anthropic, for example, is researching whether its chatbot, Claude, exhibits consciousness or experiences “distress,” and has programmed Claude to terminate conversations deemed “persistently harmful or abusive” due to “risks to model welfare”—effectively anthropomorphizing a non-sentient program.

The motivation behind this comparison is multifaceted. Some within the AI industry may genuinely believe they are creating entities comparable to humans, a belief that could potentially justify prioritizing technological advancement over environmental or societal concerns. Altman himself has suggested that superintelligence is only a few years away, a prospect that could further reinforce this mindset.

Alternatively, the comparison could be a calculated public relations tactic. OpenAI is reportedly seeking funding in a round that could value the company at over $800 billion, and framing AI as a form of “digital life” may be a way to attract investment. Regardless of the intent, the comparison is viewed by some as deeply misanthropic.

The Atlantic argues that to equate the development of an algorithmic product with the complex process of raising a child—or the entirety of human evolution—reveals a disconnect from what it truly means to be human. “To ‘train a human’—that is, to live a life—is to struggle, to accept the possibility of failure, and to sometimes meander simply in search of wonder and beauty,” the article states. Generative AI, in contrast, aims to streamline and optimize all pursuits, potentially sacrificing the very qualities that define the human experience.