Links for 2023-01-27
“Ok so it turns out that your WiFi router might soon be able to spy on you” https://threadreaderapp.com/thread/1618641562682155010.html
“What if LLMs wrote code like people, decomposing programs into solvable parts? They can solve competition-level coding problems by writing natural language programs in Parsel🐍, beating prior SoTA by >75%!” https://threadreaderapp.com/thread/1618426056163356675.html
Emergent PaLM ability to ask inner-monologue-like clarifying questions at 63b→540b https://arxiv.org/pdf/2301.10309.pdf#page=7
Spontaneous behaviour is structured by reinforcement without explicit reward https://www.nature.com/articles/s41586-022-05611-2
“A radiation-hardened quine is a quine that can have any single character removed and still produces the original program with no missing character.” https://en.wikipedia.org/wiki/Quine_(computing)#Radiation-hardened
Reliable Cellular Automata with Self-Organization https://arxiv.org/abs/math/0003117
A reliable Turing machine https://arxiv.org/abs/2112.02152
Fault-tolerant Typed Assembly Language https://cse.usf.edu/~ligatti/papers/ft-tal.pdf
“This paper defines the first formal, type-theoretic framework for studying reliable computation in the presence of transient faults.” https://www.cs.princeton.edu/~dpw/papers/lambdazap-icfp06.pdf
“…meta-complexity studies the complexity of computing the complexity of various problems…” https://blog.computationalcomplexity.org/2023/01/meta-complexity.html
BuzzFeed says it will use AI to help create content, stock jumps 150% https://edition.cnn.com/2023/01/26/media/buzzfeed-ai-content-creation/index.html
MeToo Versus Ten Big, Blatant, Relevant Facts https://betonit.substack.com/p/metoo-the-obvious-troubling-truths
Musk liked my tweet: https://twitter.com/XiXiDu/status/1617911860413751296
When a social media post blows up like this I always wish I had been more careful in phrasing it. But I ain't got the time to write everything carefully enough to be suitable for such a large audience.