thethirdone

9 ms·

61.
by thethirdone 3y ago
Almost certainly generated spam. They only have one post prior to the release of ChatGPT and within 2 weeks of ChatGPT, they are posting three articles of a similar length to this one in a day. It seems extremely suspicious that they began
62.
by thethirdone 3y ago
Based on the leaderboard screenshot, only 7 people even solved up to day 12. The later days tend to be significantly harder than the early days. Based on top 100 leaderboard times, ~2x as hard for silver star.
63.
by thethirdone 3y ago
The idea is not that ChatGPT will claim neo-feudalism is common, but that it will be more likely to parrot neo-feudalist ideas.
64.
by thethirdone 3y ago
What models in particular do you think it should have been compared to? It seems like they chose some of the most understood / cited models at around 1 B parameters.
65.
by thethirdone 3y ago
> IIRC dividing a subnormal by a subnormal returns infinity. This should not be true in a conformant implementation of IEEE 754-2008. You can get infinity by dividing by a subnormal (due to overflow), but that should only be possible wit
66.
by thethirdone 3y ago
This is not an explanation. I am aware the search results can be different to different people and based on demographic information. My specific question is why would a top result not even end up on the first page. This requires a more sign
67.
by thethirdone 3y ago
That result does not show up for me when I google it and the other one is the top result for me. If someone can come up with a reason why top results aren't even present on others' page 1, I would be very interested.
68.
by thethirdone 3y ago
Notably, that graph starts at 25 which seems misleading.
69.
by thethirdone 3y ago
It is important to note that the one game provided as evidence of 1800ish rating is an opening blunder by stockfish that has occurred in at least 8000 human games. I would expect GPT-3 to be trained on several explanations of how to punish
70.
by thethirdone 3y ago
That article does not provide much evidence to support the police view of the situation. > According to the report, more than "five particles characteristic of GSR" were found when using a "scanning electron microscopy&#x2
71.
by thethirdone 3y ago
It seems like Lichess already has most of the features on the site; they have a better UI on ChessMonitor though. The main killer feature I see is combining multiple accounts. Are there any other features didn't notice?
72.
by thethirdone 3y ago
This actually isn't too hard to make a solver for. Especially in the single white piece case. I made a solver [0] that can solve level 8 (the hardest single white level) in less than 100 ms. I believe that it can solve every single w
73.
by thethirdone 3y ago
Vehicle load will also affect charge loss on flat ground. The entire graph would need to be updated for a different vehicle or load setup. I was simply stating that it is possible to update a graph with negative edges into one with only pos
74.
by thethirdone 3y ago
You can preprocess the battery range problem quite efficiently. If is not possible to gain more charge than the gravitational potential between the endpoints for any path. You can change all of the edge weights from battery charge changes i
75.
by thethirdone 3y ago
Bias is not simply a belief. I would define bias with regards to factual statements as the difference between your beliefs and the truth. With regards to beliefs about how the world should be, I would define it as the difference between you
76.
by thethirdone 3y ago
4 is the only example of a factual claim. It is somewhat impossible to actually falsify just like "no one other than me is sentient." 1 is maybe on the fence as a factual claim, but it is very dependent on what "superior&qu
77.
by thethirdone 3y ago
It seems like some of the difficulty comes from trying to calculate the the contour as a vector rather than a raster image. I would think that a method that only calculates the contour up to a given resolution to be much easier. Rendering t
78.
by thethirdone 3y ago
Surely a Sufficiently Smart Compiler could optimize all of these to only take the a constant amount of memory (and less time). I wonder if there are any correctness constraints that would actually prevent such optimizations. For example, th
79.
by thethirdone 3y ago
> Which suggests a relatively obvious possibility of a simple writing system, where learning to write and read doesn't require memorizing countless (logographic / concept based) kanji. Though kanji seem to be older. So apparent
80.
by thethirdone 3y ago
> (The infinite recursion happens before anything is ever returned --- the return type is irrelevant.) This is not true, the closure is not actually called until the future is awaited. Each instance of do_something does actually return.
81.
by thethirdone 3y ago
> Your understanding of the topic is often completely limited to the way he frames it, and becomes quite useless when you leave that limited framing. I don't think the limited understanding is useless once you leave the simplified f
82.
by thethirdone 3y ago
I definitely agree on the false dichotomy between performance and readable code. > 1) Most code, i.e. at least 80% of the code in a codebase, will never be a performance hotspot and does not need specific optimizations (i.e. as long as t
83.
by thethirdone 3y ago
> As someone who fucking hated and still hates academical math: No, kids do not need math. Not in the sense as taught in schools, anyway. That is a straight waste of precious time. Would you mind explaining why you still hate math? It se
84.
by thethirdone 3y ago
> Maybe I'm totally misunderstanding, but figuring out the "best current path" means re-running A* every time you break a wall, as removing arbitrary walls can give you a totally new path to the goal; to wit, it might be a
85.
by thethirdone 3y ago
I am not explaining an O(n^n) solution. Its an O(E) time and O(V) space solution just like normal A*. I am assuming you are saving the initial A*run and the subsequent reverse run. Then `A* cost from the start + A* cost from the end < be
86.
by thethirdone 3y ago
The SO answer is pretty good and probably the most generalizable pathfinding solution. My first thought was to also run A* from the end to the start. This would allow you to look at each wall in the maze and check if the A* cost from the st
87.
by thethirdone 3y ago
> In addition, the UBI creates bad incentives, and requires enormous taxes if funded at an “acceptable” level. > A UBI experiment, in contrast, might be a good way to convince people of the folly of the UBI. For a modest cost, you co
88.
by thethirdone 4y ago
> Is there some reason the result of "something.iter_mut().filter(|| {...}).enumerate()" can't be saved to a variable and then used as the "something" in the first example? You absolutely can do something like `l
89.
by thethirdone 4y ago
That algorithm (implemented efficiently) is just A* using a different concept of distance. The distance specifically would be `fire*episilon + steps if steps < max else inf`
90.
by thethirdone 4y ago
I don't think Casey would object to "optimizing for human comprehension". If you are willing to give up an order of magnitude of performance, you can probably just do it without much care. I agree that "optimizing for hu

More ›