@boshsiz@noMukh The logic goes like this: previously we made experiments to show electric laws X and separate experiments to show magnetic laws Y. And relativity came along and says if relativity is true, then X must imply Y. This strongly suggests relativity has better explanation strength.
@boshsiz@noMukh The order of discovery doesn't have to agree with the order of teaching, or the order of explanation. I'm talking about how relativity explains magnetism from electricity (which adds credence to relativity since it makes more correct predictions with fewer assumptions).
@noMukh It might seem pretty complex but if we take space and time to be a 4-dimensional whole the formulas turn out to be very elegant compared to separating them. (PS. the total charge if a moving object is slightly larger, but this is second order so I ignored it in the explanation).
@noMukh So for us (not moving) to explain why the particle isn't being pushed as strongly, there has to be a separate force that's perpendicular to the velocity of a moving particle. Indeed if we have a magnetic field the force does exactly that. So the equivalence pricniple is saved.
@mijemoder@royllovians It does sound like you're speaking with an accent (and maybe trying to say "rén sài" instead). I think it's implicitly learned that these are actually impossible syllables even though when you make a table they look like obvious holes
@lefthanddraft I never understood why these special tokens can be input by the user at all. Shouldn't the tokenizer just never emit those? Or is the model confusing the string "<thought>" with the actual token?
@EChatav Would this allow us to make parser-pretty printer pairs simultaneously? I feel like people have tried multiple times combining optics with parsers, but with no usable libraries yet
@thesilenceturns@ZachWeiner I interpret the rows as statements about AI in general, and the columns are about this particular wave of AI development. This means E1 is a coherent position (although I don't agree with it): The LLM approach to AI is overhyped, but AI in general is good.
@_Fredrek@nirlichtman_ Imagine when humans have AGI and can travel the stars, and every program still starts up by checking if it is in DOS mode and politely rejecting to run
@A1987dM@ereliuer_eteer The continuum hypothesis is more often discussed in the context with choice because otherwise cardinals behave a lot weirder. (That's not to say it is not interesting without choice though.)
@yuchenliu42@yachty66@deepseek_ai There is literally a single photo of him somehow on the whole internet and everything else is AI/Photoshopped from that one