Description
A writer hides half a sentence written by AI in his novel, then dares his readers to find it. Who will?
Meanwhile, another novel was pulled amid suspicions. Between software that guesses and readers who get it wrong, where is the proof?
A lab has AI agents prove that the code protecting your messages keeps its promises. Then it breaks that code on purpose, just to see.
At Stanford, the machine can do the homework. So one professor changes the test: handing it in is no longer enough, you have to defend it in front of someone who can say no.
In 1790, a padlock waits in a London shop window, with a challenge posted beside it. It tells you what a proof that holds looks like.
And to close, three things to try this week, including one to check the security of your own conversations yourself.
Three signals. One thread. One questionβ¦
ποΈ Every week, SynthAIcast listens to the world's best AI podcasts so you don't have to. Synthesised in ten minutes, verified by a human, in English and French
π Also available in French: SynthIAcast
π THE SOURCES, BY SIGNAL
Signal 1 - a sentence nobody finds
β’ RTL (FR) - the winner reveals his half-sentence, 07/11
ββββββββββββββββββββββββββββ
Signal 2 - a proof you let people attack
β’ Google DeepMind - AlphaProof Code and Signal, 10/11
ββββββββββββββββββββββββββββ
Signal 3 - proving means defending
β’ Latent Space - Mihail Eric, Stanford, 08/11
ββββββββββββββββββββββββββββ
Perspective
β’ Wikipedia - the Bramah lock (1790-1851)
β’ Linda Hall Library - Joseph Bramah
ββββββββββββββββββββββββββββ
π§ͺ TRY THIS WEEK
β’ Signal - what is a safety number?
Also cited, without a link: RTL (the Goncourt Academy and detectors, 23/10) and Google DeepMind's technical paper (10/11).
Hosted on Ausha. See ausha.co/privacy-policy for more information.





