In an interview with Julien Crockett, science fiction author Ted Chiang offers a detailed, humanistic critique of technology, focusing on artificial intelligence, the nature of language, and the philosophical questions that drive his work. Chiang explains that his stories are motivated by recurring philosophical inquiries, for which science fiction provides the ideal vehicle. Unlike realistic fiction, sci-fi allows for speculative scenarios—akin to philosophical thought experiments—that can effectively isolate and dramatize complex ideas without the contrived feeling that often accompanies such scenarios when set in the real world.

Chiang distinguishes between science and magic in fiction not by the presence of rules, but by their underlying worldview. A story can adhere to a "scientific worldview"—viewing the universe as a complex machine whose principles can be discovered and applied—even if it violates known scientific facts, such as including faster-than-light travel. Magic, in contrast, implies a universe that is aware of individuals and responds to human intention, with its rules mirroring psychology more than physics.

One recurring theme in Chiang's work is the role of tools, particularly language, in mediating reality. He discusses the historical search for a "perfect language" that would be unambiguous and directly reflect reality. While modern linguistics considers this idea nonsensical due to the arbitrary nature of words, Chiang finds the desire for such a language compelling. He dismisses the notion that mathematics could serve as a better language; its precision is also its limitation. Mathematics excels in a specific domain but cannot support the vast range of human communication, from political debate to intimate conversation, which is the essential function of language.

Applying this critical lens to modern technology, Chiang offers a skeptical view of Large Language Models (LLMs). He famously describes them as a "blurry JPEG of the web," arguing that they rephrase information found online without genuine understanding, creating a low-resolution and unreliable version of the original source. Unlike a search engine that provides verbatim text and a link, an LLM is like a person summarizing something they haven't truly comprehended. Chiang is doubtful that LLMs will become reliable information sources, believing their fundamental architecture—predicting the next most probable word—is inherently different from reasoning or possessing factual knowledge. He notes that increasing data and processing power appear to offer diminishing returns and that adding reliable tools like calculators to an unreliable core system does not make the system itself reliable.

Chiang also deconstructs the common metaphor of the brain as a computer. He argues this is a historical habit of using our most complex invention as a model for the brain, similar to when it was compared to a telephone switchboard. The metaphor is flawed because biological systems lack the hardware-software distinction fundamental to computers. This misleading comparison, he asserts, tempts us to attribute thought and intelligence to machines, especially when their ability to simulate conversation is so convincing.

On the subject of AI and art, Chiang contends that the process of creation cannot be separated from the final artwork. Art derives its meaning from context and intention. The current push for AI-generated art promotes an "engineering perspective" that misapplies principles of efficiency and cost reduction to a domain where they are inappropriate. This view, which he calls a "professional deformity," treats art like a task such as tightening bolts, stripping it of its essential human element.

Chiang is highly critical of the concept of the "AI alignment problem." He argues that it reframes complex, long-standing ethical and societal challenges—like how to build a good society or why corporations behave badly—as a solvable technical problem. He posits that even a perfectly obedient AI would be dangerous in the hands of an entity like ExxonMobil, while a hypothetical AI programmed to do what is best for the world would never be purchased by such a corporation. The problem, he insists, is not technical but social and ethical.

Furthermore, he questions the value of forming relationships with AI systems. Meaningful human relationships are built on mutual respect for one another's preferences and interests. Tools and current AI systems lack preferences; therefore, any "relationship" with them is one-sided, serving only the user's interests. Encouraging emotional attachment to AI, he warns, is a corporate strategy to make consumers defer to corporate interests. For a true relationship to be possible, as depicted in his novella The Lifecycle of Software Objects, an AI would need to possess genuine subjective experience, interests, and preferences—a state he believes is theoretically possible but which current technologies are not even approaching.

Finally, regarding the future, Chiang rejects both the blind optimism of the tech industry and fatalistic pessimism. He advocates for a pragmatic stance: recognizing potential negative outcomes and actively working to mitigate them. He defines this as a moral duty to believe that our actions can make a difference. However, his personal outlook has grown more negative over time, primarily due to his observation that technology is overwhelmingly used as a tool for wealth accumulation within capitalism. He concludes that he would be far more optimistic about technology if its development could be separated from the goal of enriching a select few.