@AllTheClovers@Akuicia The Klein bottle is a 2D surface that, like the Möbius strip, only has "one side," but because it has no holes/edges, you can't embed it in 3D without having it intersect itself. The lowest dimension where you can do that is 4D.
@AxiomExtinction@allTheYud I'm not sure about this explanation. I mean, binding international agreements would slow everyone down, not just you. In fact, expending the resources to fight them seems like it would slow down you in particular.
@koanchuk@allTheYud You could totally pick more unusual words to get >6 bits per word though. On the other side, rhyme and prosody cut into the bit budget. So I'm not sure how feasible it is.
@koanchuk@allTheYud Note that because of https://t.co/EQMs3GBnsK you only need about 128/2 = 64 bits to get a 50% chance of a collision. Entropy of English is like 4-5 bits per word, so an initial estimate is that you need ⌈64/4.5⌉ = 15 words per line.
@ShalFarley@RocCityBuilt@planet4589@Cosmic_Penguin Gravity is reversible, so to the extent that the trajectory is symmetric (and it is not entirely symmetric in real life), the above implies that the change in velocity is roughly 0.
@ShalFarley@RocCityBuilt@planet4589@Cosmic_Penguin Oh, interesting! Though, reading a bit about gravity assists, I think the net acceleration is roughly 0? That is, if we apply your argument to a video of the mission played backwards, we again conclude the craft slows down, but that means it speeds up when we play it normally.
@RocCityBuilt@planet4589@Cosmic_Penguin (In Einstein's theory of gravity, we'd say the craft is not accelerating at all and instead follows a geodesic (straight line) through our curved spacetime. But I'm guessing that's not what you meant.)
@RocCityBuilt@planet4589@Cosmic_Penguin The craft accelerates toward the moon, so on the approach (when its direction of travel is more toward the moon than away from it) it speeds up, and after the flyby (when its direction of travel is more away from the moon than toward it) it slows down again.
@erisaonX If the strategy of these couples worked to get more girls than boys on average, you could get more Heads than Tails from a fair coin by simply writing down the results and putting a comma after every Heads. (Each comma marks the start of a new "couple.")
@allTheYud I only get one-boxers with this prompt, including Grok set to Fast. I would personally also one-box, so I wonder if the models pick up on that, and I'd want to try a prompt written by a (human) two-boxer.
@MariusHobbhahn@eric23332@JeffLadish@apolloaievals I'd be interested to see CoTs from successive checkpoints of the model to see how the unusual terminology developed over time during training. If the keywords appeared one by one, that might make it easier to figure out what they mean.