Philip Ball writes:

Reading The AI Mirror I was struck by [Shannon] Vallor’s determination to probe more deeply than the usual litany of concerns about AI: privacy, misinformation, and so forth. Her book is really a discourse on the relation of human and machine, raising the alarm on how the tech industry propagates a debased version of what we are, one that reimagines the human in the guise of a soft, wet computer.

If that sounds dour, Vallor most certainly isn’t. She wears lightly the deep insight gained from seeing the industry from the inside, coupled to a grounding in the philosophy of science and technology. She is no crusader against the commerce of AI, speaking warmly of her time at Google while laughing at some of the absurdities of Silicon Valley. But the moral and intellectual clarity and integrity she brings to the issues could hardly offer a greater contrast to the superficial, callow swagger typical of the proverbial tech bros.

“We’re at a moment in history when we need to rebuild our confidence in the capabilities of humans to reason wisely, to make collective decisions,” Vallor tells me. “We’re not going to deal with the climate emergency or the fracturing of the foundations of democracy unless we can reassert a confidence in human thinking and judgment. And everything in the AI world is working against that.”

To understand AI algorithms, Vallor argues we should not regard them as minds. “We’ve been trained over a century by science fiction and cultural visions of AI to expect that when it arrives, it’s going to be a machine mind,” she tells me. “But what we have is something quite different in nature, structure, and function.”

Rather, we should imagine AI as a mirror, which doesn’t duplicate the thing it reflects. “When you go into the bathroom to brush your teeth, you know there isn’t a second face looking back at you,” Vallor says. “That’s just a reflection of a face, and it has very different properties. It doesn’t have warmth; it doesn’t have depth.” Similarly, a reflection of a mind is not a mind. AI chatbots and image generators based on large language models are mere mirrors of human performance. “With ChatGPT, the output you see is a reflection of human intelligence, our creative preferences, our coding expertise, our voices—whatever we put in.” [Continue reading…]