How AI hype can be dehumanizing, and demeans us both.

I've really liked what Emily Bender has written about AI. She and her co-authors coined the term "stochastic parrots" to describe how large language models aren't "thinking" but rather are a simply a statistical system that stitches together the most likely string of words in response to a prompt. She has written many other pieces, and has a podcast, on the "hype" around AI. And all this from a Professor of Linguistics!

I was taken by her recent publication in Current Directions in Psychological Science entitled "Resisting Dehumanization in the Age of “AI”." In that article, she describes how many of the ideas explicitly or implicitly embedded in the AI discourse are dehumanizing. For example, the "computational metaphor" that the brain is a computer also implies that a computer is a brain. But it's not. As just one example why: a computer doesn't have emotions, and emotions are a key part of thinking, and being human. I really worry when we anthropomorphize AI - the more we talk like it is "thinking" or "doing" things, the easier it gets to misunderstand what AI is and what it is not (human, goal-oriented, etc.).

She also takes a critical lens to AI discourse, showing how it (further) privileges and normalizes white, affluent, western views. She also shows how AI development involves hidden human labor, often done and compensated in ways that are dehumanizing. Cognitive scientists, her audience for this article, are well-positioned to work against this dehumanization, and she outlines a number of ways they can do so, including engaging in more public scholarship. It's a really thought-provoking article that I encourage you to read.

The article also reminds me of a classic line from The Simpsons, where a "baby translator" reveals a child's babbles is actually them saying: "This leash demeans us both."

I think talking about AI like it is human "demeans us both."