HackerHow much does a successful demonstration establish?
A program gives the expected output. Between “it did” and “it will” lie the test distribution, formal assumptions and deployment context. Those boundaries are easy to miss under pressure to ship.
Questions in this reading · 4
Does running a model explain its decision?
Executing a model yields an answer. Explaining why that answer followed, or predicting its behavior under a new input, requires further evidence.
Compare original passages and timestamps · 2
you connect that inscrutable thing to important real life things like financial marketswe don't know how these parameters collaborate to actually perform thatWhat if the model has already seen the test?
A public score becomes harder to interpret when the training corpus is unknown. Novel questions probe a different failure mode, without proving what the original corpus contained.
Compare original passages and timestamps · 2
we don't know what gpt4 was trained on I don't know what gpt4 was trained onsome of the questions from these things actually leaks through to some of the training setsWhat can a formal check actually verify?
A tool needs a precise property to check. Translating what someone intended into a formal statement is itself a source of error.
Compare original passages and timestamps · 2
in the limit I'll like you know formally verify all of my protocols.Before we can start
proving new theorems, we need to be able to
state old theoremsHow does cheaper automation become a deployment race?
One account considers incentives to replace costly work. The other asks how competition changes the decision to release a technology whose capabilities are still emerging.
Compare original passages and timestamps · 2
A tenth the speed is still cost effective when it's a hundredth the price.if that technology confers power it will start a race.Try it on your own example
State the conditions on the demonstration: what survives a new question or a change of assumptions?