Tuesday, August 28, 2018

Optimizing the Verified Good, by Effie Seiberg

[Analog]
★★★☆☆ Honorable Mention

(Robot SF) The cleanerbot picks up robot parts in the arena after each fight. It does its task well—optimizing the good of a clean arena—but it has other goals to try to optimize too. (5,093 words; Time: 16m)


"Optimizing the Verified Good," by (edited by Trevor Quachri), appeared in issue 09-10|18, published on by .

Mini-Review (click to view--possible spoilers)

Pro: In a way, the story illustrates the problem of any sophisticated AI system: it will try very hard to optimize some objective function. If that objective function doesn’t exactly line up with what humans want done, then the customers may be disappointed by the result.

Clearnerbot, having decided that minimizing pain is a verified good, keeps finding “solutions” that lead to new problems until it finally finds something satisfactory to bots and humans alike.

There’s a cute twist that any political activist will recognize: some of the other bots try to insist on a perfect solution—one with no compromises—and have to be talked out of that lest there be no solution at all.

Con: It’s hard to see how management didn’t see this problem and get the manufacturer to fix it some time ago. Just removing cleanerbot's notion that optimizing some other bot’s pain was a good thing would have fixed it right off the bat.

Other Reviews: Search Web, Browse Review Sites (Issue 09-10|18)
Effie Seiberg Info: Interviews, Websites, ISFDB, FreeSFOnline

Follow RSR on Twitter, Facebook, RSS, or E-mail.

No comments (may contain spoilers):

Post a Comment (comment policy)