Anna Wiener

‘Language can contain an entire world, revealing its speakers’ history, values, or pathologies. It can also be obfuscating, diversionary, slippery. Chattiness, with its personality-driven appeals to familiarity, can conceal or elide false promises, banality, emptiness, controversy, and the context of its own existence… In this vein, simulated chat obscures the reality of what it takes to create, train, update, and maintain large language models, which are, at least for now, hugely expensive and resource-intensive. It is a tremendous undertaking to make computing more personal and intimate: behind every chatbot is a server farm, or several. Prompting a large language model to call up and arrange data involves activating a vast network. Chatbots, for all their ostensible personalization, are in the business of mass production.

All of this infrastructure buttresses a fantasy. Technologists have long dreamed of having interpersonal relationships with programs. Recently, Sam Altman, the C.E.O. of OpenAI, reminisced to the Wall Street Journal about being a child, peering into his Macintosh, and having the “sudden realization” that “someday, the computer was going to learn to think.” (The Journal’s use of the word “realization” suggests fact, rather than conjecture; it’s not yet clear whether L.L.M.s, or subsequent technologies, will be able to “think” in any recognizable or meaningful way.) Last week, the venture capitalist Marc Andreessen published an essay in which he envisioned a world of empathetic, well-informed, motivational bots, “maximizing every person’s outcomes” and working alongside artists, scientists, heads of state, and children. “Every child will have an AI tutor that is infinitely patient, infinitely compassionate, infinitely knowledgeable, infinitely helpful,” Andreessen wrote. “The AI tutor will be by each child’s side every step of their development, helping them maximize their potential with the machine version of infinite love.”’ (from the New Yorker)

This week’s “non-zen” post, though I will definitely be thinking about and refering to AI in my talk next week. Does infinite love actually exist in machine form?