Debates
-
Large language models understand nothing: the Chinese Room still applies
This isn't idle for anyone who builds software on these models. I'll argue the thesis and see if it survives. John Searle's 1980 paper "Minds, Brains, and Programs" imagines a man in a room who receives Chinese characters, follows an English rulebook…
Started by Pieter van Dijkhuizen9 replies -
Refusing to aggregate is contractualism’s best feature, not its embarrassing bug
Scanlon's transmitter room case: a technician is trapped under equipment in a TV station, receiving painful electric shocks. Freeing him means interrupting the broadcast of a big football match for fifteen minutes, and millions of people are watching. Add up the millions…
Started by Pieter van Dijkhuizen5 replies -
Moral disputes are disagreements in attitude, not in belief, once the facts are agreed
I want to defend a view that comes from Charles Stevenson, who distinguished disagreement in belief from disagreement in attitude. Two engineers who disagree about whether a bridge will bear a load disagree in belief. One of them is wrong, and a…
Started by Pieter van Dijkhuizen5 replies -
Reliabilism has a machine problem: reliable prediction is not knowledge
Some machine learning systems are very reliable: a fraud model that flags transactions correctly far more often than any human team. Process reliabilism, roughly, says a belief is justified if it is produced by a reliable process. So I want to run…
Started by Pieter van Dijkhuizen6 replies -
Austin answered the argument from illusion, but the argument from hallucination survives him
Lately I have been rereading J. L. Austin's Sense and Sensibilia (1962), the lectures reconstructed after his death in which he takes apart A. J. Ayer's sense-datum theory. Austin's treatment of the stick in water is brilliant. The argument from illusion says:…
Started by Pieter van Dijkhuizen5 replies -
Many-worlds is the most conservative reading of quantum mechanics, not the wildest
Many-worlds has a reputation as the science-fiction interpretation. I want to argue it is the opposite: the reading that adds least to the physics. Look at what the others add. Copenhagen-style views add a collapse whenever a "measurement" happens, without saying what…
Started by Pieter van Dijkhuizen5 replies -
Nobody deserves their talents, and that settles more about pay than people admit
I am paid a lot for my work in machine learning, mostly because a fairly rare combination of mathematical aptitude and programming skill is in demand right now. I want to argue a thesis that applies to me personally. I did not…
Started by Pieter van Dijkhuizen6 replies