In Before Sunrise (the acclaimed 1995 Richard Linklater film about a couple falling in love over the course of a night in Vienna) Jesse and Celine, the characters played, respectively, by Ethan Hawke and Julie Delpy, are strolling along the Donaukanal when they stumble across a young man.

“Instead of just asking for money,” the young man says,

I’ll ask you for a word. I will then write a poem in which that word will be used. . . . If you like it, if you feel it adds something to your life in any way, then you can pay me whatever you feel like.

Celine gamely offers the word “milkshake,” and the man proceeds to churn out some maudlin verse that includes the lines “Look at those big eyes on your face / See what you mean to me / Sweet cakes and milkshakes.”

After duly chucking the poet a couple schillings, the couple stroll off, and Jesse snidely—but accurately—observes, “You know, he didn’t just write that poem. . . . He just didn’t write it tonight. He just plugged in that milkshake.”

This scene, when considered nearly 30 years later, evokes nothing so much as a critique of ChatGPT, the now-ubiquitous OpenAI chatbot that, over the past several months, has generated terabytes of content and commentary.

ChatGPT advocates have waxed lyrical about the service, and many users—myself included—have managed to fool family and friends into thinking they’ve composed works that the chatbot actually wrote (incidentally creating an enormous potential problem in schools—ChatGPA, anybody?).

Critics of the Jesse variety, on the other hand, contend that the bot, like the cinematic canal-side poet, is simply assimilating the inputs it’s fed into a preexisting framework and regurgitating the combination as a supposedly new work.

These reactions appear to reflect fundamentally different understandings of who or what artificial intelligence (AI) actually is: a genuinely autonomous entity capable (now or soon) of cognition or a representation or avatar of its programmers? A “who” or a “what”? An “it” or an “us”?

This division between what I’m calling the “autonomy” and the “automaton” schools is likely to drive not just philosophical and ethical but practical, legal, social, and economic debates as AI continues to mature.

For instance, if we apply the autonomy-automaton distinction to the copyright dispute I examined earlier this month over AI image generators, it appears that the plaintiffs, who labeled the robotic art-makers “21st-century collage tools,” fall into the automaton school of thought that considers AI essentially a human-programmed regurgitator that mindlessly uses and copies their works with abandon. Getty Images, which filed a separate suit in the UK, acknowledged that “artificial intelligence has the potential to stimulate creative endeavors” (emphasis added) but not necessarily to create independently.

Meanwhile, the folks at Stability AI and DeviantArt defending the suit rely on the fair-use doctrine—which, among other things, allows for limited use of a copyrighted work so long as the result is transformative and does not cause excessive harm. The autonomy school believes that AI image generators employ authentically creative impulses (“thought” may not yet be the correct term) to produce genuinely creative, transformative art.

But who’s right? Is AI itself, or is it us? And what are the implications of how we resolve this tension?

In future posts, and longer articles, I look forward to exploring issues of intellectual property and innovation, including awarding patents to AI and determining ownership of AI-generated works, through this autonomy-automaton lens.

The post Who Is AI? Toward a Framework for Understanding Intellectual Property and Robotic Innovation appeared first on American Enterprise Institute – AEI.