Category: Ken Liu

Uma by Ken Liu

Uma by Ken Liu (Avatars, Inc., 2020) opens with the narrator discussing his employment-related disciplinary case with a lawyer before the story flashbacks to the incident that caused his problem—the rescue of three children from a burning house while he was operating a UMA for a power company:

A Utility Maintenance Avatar is vaguely humanoid, but only about three feet tall fully stretched out and no more than fifty pounds in weight. For light maintenance tasks such as vegetation management, removal of bird and wasp nests, patching cables, and so forth, you don’t need or want anything bigger—the extra bulk would just get in the way. I had at my disposal small shears, extensible ladder-legs, a general electrical tool kit, and not much else. PacCAP has thousands of these cheap telepresence pods distributed around the state to maintain its hundreds of thousands of miles of transmission, distribution and equipment. With remote operators in centralized offices inhabiting them whenever needed, it’s much cheaper than sending out a whole crew in a truck just to prune an overgrown oak branch.  pp. 134-135

During the rescue the children receive minor injuries (scratches, a sprained ankle, etc.), and subsequently a plantiff’s bar AI suggests they should sue the power company because the narrator wasn’t properly trained, etc. Hence the company disciplining him for safety violations.
Later, after the narrator has refused to sign the legal papers, he is contacted by the power company’s CEO about another emergency—and ends up operating a similar model UMA in Myanmar to save a kid trapped during an earthquake.
This piece is a convincing look at what the future might bring, and it also has a couple of good action scenes—but it feels rather fragmentary, more a neat idea than a story.
**+ (Average to Good). 4,150 words. Story link.

50 Things Every AI Working With Humans Should Know by Ken Liu

50 Things Every AI Working With Humans Should Know by Ken Liu (Uncanny, November-December 2020) takes the form of a futuristic article written about a Dr Jody Reynolds Tran and the neural network (essentially an AI) she creates called WHEEP-3. Tran later publishes a best-selling book about WHEEP-3, and subsequently causes a controversy when she reveals that the neural network was the author. There is more fuss later on when “seeds” of prose supposedly written by WHEEP-3 are found to be authored by Tran.
The story finishes with a reprint of one of WHEEP-3’s seeds, the “50 things” referred to in the title, a mix of statements that range from the obscure to the observational:

25. “I never expected to sell my rational numbers.”
26. Accepting that most humans will never get the joke.
27. That they cannot visualize more than three dimensions.
28. That they cannot manipulate time by slowing down or
speeding up.
29. That they are trapped, but think of themselves as trappers.
30. That they are free, but believe themselves imprisoned.

A moderately interesting look at how future AIs may behave and communicate—but ultimately a slight, fragmentary piece.
** (Average). 1,900 words. Story link.