317070

10 ms·

121.
by 317070 1y ago
Note that here they are working via phone apps, so if you are navigating with a local navigation app like Karta GPS, Flitsmeister or Sway, these apps can let the lights know you are heading that way and will be there in X minutes. Therefore
122.
by 317070 1y ago
At a quick glance, the minimal memory requirement is the O(N) partial solutions you need to keep, with N being the number of rows. Is your comment that this is higher than the memory requirement of the recursive solution, because that doesn
123.
by 317070 1y ago
> Even just keeping the partial solutions in memory that will later be needed can use much more memory than a bottom up solution. Can you give an example of this? I don't think that is correct. But we might be saying the same thing,
124.
by 317070 1y ago
I think "cached recursion" is too broad. I tend to go with "Minimal Memory Memoization". 1) the recursion solution is often too slow, but uses little memory. 2) the memoization solution can make the algorithms from 1 a l
125.
by 317070 1y ago
Does someone know the problem/challenge being solved?
126.
by 317070 1y ago
It's not the prairie dogs themselves, but the fleas on the dogs. The carriers for the plague are fleas.
127.
by 317070 1y ago
Earth without land masses would have around 0.5m to 1.5m of tides, depending if it is springs or neaps. All the higher amplitude tides are caused by resonance in the sea. The North Sea and the Channel and Irish straight are quite resonant.
128.
by 317070 1y ago
The issue is clarified with a map like this: https://i.redd.it/rontertecjqd1.jpeg Most regions have very small tidal ranges. That doesn't mean they have small tidal currents (think of fjords or straights for example),
129.
by 317070 1y ago
What you're thinking of is much more like the Genie model from DeepMind [0]. That one is like Veo, but interactive (but not publically available) [0] https://deepmind.google/discover/blog/genie-2-a-large-scale
130.
by 317070 1y ago
It might be computable, but is it practical? It sounds like it requires a stack of exponentials in compute time, and is impractical even for very short messages.
131.
by 317070 1y ago
> Let a semantic frame be defined as Ω = (Σ, R) But if we let an AGI operate on Ω2 = (English, Science), that semantic frame would have encompassed both Newton and Einstein. Your argument boils down into one specific and small semantic f
132.
by 317070 1y ago
I'm sure this is a reference to alpha stable distributions: https://en.m.wikipedia.org/wiki/Stable_distribution Most of these don't have finite moments and are hard to do inference on with standard statistica
133.
by 317070 1y ago
There are so many more of these. A common one is everything to do with babies. * In the UK you should not start solids before 6 months, in France you can start at 3, and should at 4. * Baby bed room temperature: in the UK 16 degrees Celsius
134.
by 317070 1y ago
But relatively few countries follow common law with its focus on precedents. In fact, civil law is a lot more common, and that one cares less about precedents. [0] > The primary contrast between the two systems is the role of written dec
135.
by 317070 1y ago
lol at having 1 gram of antimatter. They are moving around dozens of atoms here. 1 gram of antimatter is off on humanity's capabilities by roughly a dozen orders of magnitude.
136.
by 317070 1y ago
It's a pet peeve of mine, but "publications" and "peer-review" are not really part of the scientific process? Just like "academia" they have kind of grown onto the term, almost claiming it for their own. I
137.
by 317070 1y ago
Ironically, Britain is now an island with an extremely busy train connection to continental Europe. Soon, you might go from London to Köln without switching trains!
138.
by 317070 1y ago
I do hope people are still innocent until proven guilty? If you did not use copyrighted materials for training, people will not be able to prove that you did, and that should be good enough.
139.
by 317070 1y ago
Why the golden ratio? Because the continued fraction of the golden ratio is all 1's [0]. So it is uniquely hard to approximate with a rational number. The golden ratio is the bound on Hurwitz's theorem [1]. And avoiding a rational
140.
by 317070 2y ago
Way back when, most of a GPU was for graphics. Google decided to design a completely new chip, which focused on the operations for neural networks (mainly vectorized matmul). This is the TPU. It's not a GPU, as there is no graphics har
141.
by 317070 2y ago
Yes, that sounds right.
142.
by 317070 2y ago
Or the Montgolfier brothers! That was the first time humans went up in the air. https://en.m.wikipedia.org/wiki/Montgolfier_brothers I was surprised the first time I heard Americans attribute flight to the Wright broth
143.
by 317070 2y ago
Why would I want it to be bad? I'm afraid I don't understand what you mean.
144.
by 317070 2y ago
2^0 and 2^{-1}. other positive integers will end on an even number, other negative integers will end with the numbers 25.
145.
by 317070 2y ago
I take my quip both ways, so I would wager that even with finetuning, these models are only 1 generation ahead in esoteric language performance and therefore _still not very good_. Am I correct?
146.
by 317070 2y ago
I work for DeepMind on project Astra. Not to dwell too deep into confidentiality of what capabilities I have been looking at, but it has been the theme since the flamingo model that you only gain about 1 model-generation by fine-tuning ver
147.
by 317070 2y ago
I've been finetuning these models since before chatGPT, and the one lesson I've learned is that by the time you have set up everything to fine-tune a model, you can expect a newer model to do as well with prompt-tuning. So, unless
148.
by 317070 2y ago
No, 10x less sampling steps. Whether or not that means 10x faster remains to be seen, as a diffusion step tends to be more expensive than an autoregressive step.
149.
by 317070 2y ago
Sure, but the lesson is that the big story is what matters, way more than detailed RoR calculations. Bob made huge mistakes, but even then compounding was the stronger force and he ended up ahead.
150.
by 317070 2y ago
Meet Bob. Bob is the world’s worst market timer. https://awealthofcommonsense.com/2014/02/worlds-worst-market...

More ›