The AI Reader.
Friday, August 28, 202652 min read
Working Knowledge · What Could Go Wrong

The UN's Libya report never confirms an autonomous weapon killed anyone

A radar-seeking weapon Israel was exporting to China by the mid-1990s meets the 2015 open letter's definition of one.

10 min read17 sources
Working Knowledge · The Instruments

'Human parity' in speech recognition came down to how you count the humans

Word Error Rate counts the word edits between a transcript and a reference, and when IBM re-transcribed the same phone calls the human benchmark on the hard set dropped from 11.3 percent to 6.8.

9 min read8 sources
Working Knowledge · The Mechanics

A model can spell strawberry and still miscount its letters

Its tokenizer splits the word into str, aw, berry before it reads anything, and how much of the spelling survives that split is still partly unsettled.

9 min read10 sources
Working Knowledge · The Mechanics

The token a model can't repeat back

SolidGoldMagikarp reached the model's vocabulary from a Reddit counting forum and never reached its training.

9 min read8 sources
Working Knowledge · The Evidence

Google packed a whole sentence into a single vector

The 2014 LSTM that gave machine translation its encoder-decoder shape leaned on a source-reversing hack that attention made unnecessary within months.

8 min read6 sources
Working Knowledge · When AI Breaks

CNET let an AI draft its finance advice under a staff byline

The in-house engine produced 77 stories, one of which told readers a $10,000 deposit at 3 percent would earn $10,300 in a year.

7 min read11 sources