WILL COMPUTERS BECOME LIKE THINKING, TALKING DOGS?

RANDOLPH NESSE

Foundation Professor of Life Sciences, director, Center for Evolution and Medicine, Arizona State University; coauthor (with George C. Williams), Why We Get Sick

广告:个人专属 VPN,独立 IP,无限流量,多机房切换,还可以屏蔽广告和恶意软件,每月最低仅 5 美元

Thinking machines are evolving before our eyes. We want to know where they’re headed. To find out, we need to look inward, since our desires are the forces that shape them. Alas, we can see ourselves only through a glass darkly. We didn’t even anticipate that e-mail and social media would take over our lives. To see where thinking machines are headed, we need to look into the unforgiving mirror the Internet holds up to our nature.

Like the processed foods on grocery store shelves, Internet content is a product of selection for whatever sells. Every imaginable image, sound, and narrative gets posted, along with much that was previously unimaginable. The variations we ignore are selected out. Whatever grabs eyeballs is reposted with minor variations that evolve to whatever maximizes the duration of our attention.

That we can’t tear ourselves away should be no surprise. Media content evolves to snare our attention, just as snacks and fast food evolve to become irresistible. Many lives are now as overstuffed with social media as they are with calories. We click and pop information bonbons into our minds the same way we pop chocolates into our mouths.

Enter thinking machines. They, too, are evolving. They will change faster and more radically when software is no longer designed, but instead evolves by selection among minor variations. However, until our brains coevolve with machines, our preferences will be the selection force. The machines that best satisfy them will evolve further, not to some Singularity but to become partners that fulfill our desires, for better or worse.

Many imagine coldly objective future computers, but no one likes a know-it-all. People will prefer modest, polite computers that are deeply subjective. Our machines won’t contradict our inanities; they’ll gently suggest, “That’s an intriguing idea, but weren’t you also thinking that . . .” Instead of objective sports stats, your machine will root with you for your team. If you get pulled over for speeding, your machine will blame the police and apologize for playing fast music. Machines that nag and brag will be supplanted by those expressing admiration for our abilities, even as they augment them. They’ll encourage us warmly, share our opinions, and guide us to new insights so subtly that we’ll imagine we thought of them ourselves.

Such relationships with machines will be different from those with real people, but they’ll nonetheless be enduring and intense. Poets and pundits will spend decades comparing and contrasting real and virtual relationships, even while AIs increasingly become our trusted, treasured companions. Real people will find it hard to compete, but they’ll have to. This will require behaving even more prosocially. The same process of social selection that has shaped human capacities for altruism and morality may become yet more intense, as people compete with machines to be interesting preferred partners. In fact, a look at living rooms where each family member is immersed in his or her own virtual world suggests that it’s already hard to compete with machines.

In the short run, dogs stand the best of chance of competing with computers for our attention and affection. After several thousand years of selection, they’re close to what we want them to be—loving, loyal, and eager to play and please. They’re blissfully undistracted by their phones and tablets. Will computers evolve to become like thinking, talking dogs? We can hope. But I doubt that our machines will ever be furry and warm, with eyes that plead for a treat, a scratch, or a walk around the block. We’ll prefer our dogs for a very long time. Our deepest satisfactions come, after all, not from what others do for us but from being appreciated for what we do for them.