thesz

29 ms·

541.
by thesz 3y ago
Let me introduce you to the wonderful game that is The Talos Principle: https://en.wikipedia.org/wiki/The_Talos_Principle It discusses whether it is possible to evolve AGi using... computer game engine! And that is Joh
542.
by thesz 3y ago
It has to do with the following part of the comment I replied to: > guaranteed... that programs are: Non-blocking Data-race free Deadlock free Optimally scheduled Scalable by default And the call that "an
543.
by thesz 3y ago
SQL with SERIALIZABLE isolation level. It also has an upshot that you can execute your quieries on GPU [1]. [1] https://github.com/BlazingDB/blazingsql
544.
by thesz 3y ago
> It is interesting that things still train even when various parts are pretty wrong — as long as the sign is right most of the time, progress is often made. https://forums.fast.ai/t/how-to-do-reproducible-mo
545.
by thesz 3y ago
John Carmack pointed out (and I learned it here at HN) that what training really needs is the *sign" of each individual gradient parameter. I.e., you can quantize gradient to -1, 0 and 1 and still have neural network learn much of the
546.
by thesz 3y ago
You are mathematician, am I right? Your answer is correct (sorta, as far as I am concerned) and does not add to the discussion I tried to open. Your answer does not show anything related to the ability of transformers to prove absence of so
547.
by thesz 3y ago
Can transformers prove absence of the solution? E.g., did they tried to train them on the unsolvable problems? I once tried my hands on the PCB routing and here's a simple problem of simultaneous multipath search that is unsolvable:
548.
by thesz 3y ago
> darcs before it. Darcs is older than git.
549.
by thesz 3y ago
> people who already have Non-alcoholic fatty liver disease. That is 25% of US adults. Fuits contain fructose (a poison) which is transformed to glucose by liver, because our bodies cannot use it directly. From this perspective, beverage
550.
by thesz 3y ago
> sleep is good, anxiety is bad. Sleep is not all good: https://en.wikipedia.org/wiki/Hypersomnia Anxiety is not all bad: https://www.intechopen.com/chapters/43757 Sattvic food is not good for
551.
by thesz 3y ago
Words can be read as single hieroglyphs, basically, it is matter of experience.
552.
by thesz 3y ago
"Lifted pretty fast" is a vague statement. When where these restrictions lifted? Before or after wildfires season?
553.
by thesz 3y ago
I remember (from videos of VivaFrei, a Canadian citizen) that it was prohibited to walk outside area of few blocks away from your home address. And there were curfews. If so, going off in the woods was one of the things you could definitey
554.
by thesz 3y ago
More or less so. The resulting vocabulary will be structurally same even if you correct for that, only indices will differ slightly. When I spotted this in my code, I did an investigation whether this affects my encoding. Turned out, it doe
555.
by thesz 3y ago
I do not understand why your solution skips to the next unique elements. Consider sequence [1,1,1,1,1,1]. We should count there 3 (three) (1,1) pairs. Karpathy's (and mine's) algorithm would count 5 pairs, yours' will count 1
556.
by thesz 3y ago
Hehe. https://github.com/karpathy/minbpe/blob/master/minbpe/base.p... This code will incorrectly count pairs in [1,1,1] as two pairs of (1,1), despite there is only one, if you consider what merge f
557.
by thesz 3y ago
Do you have a link?
558.
by thesz 3y ago
An application may not "pwn" device directly, but it can execute an attack on the out-of-order CPU and steal valuable data.
559.
by thesz 3y ago
I tried to learn some tai chi long ago and found that it is not possible to perform tai chi correctly if you are thinking about anything, including thinking about performing tai chi correctly. And "not thinking about anything" is
560.
by thesz 3y ago
Blasphemy! Everything is a tables and SQL! (hint: tables and queries implement content-addresable memory [1] and, consequently, dataflow architecture [2], which is also Turing complete) [1] https://en.wikipedia.org/wiki&#x2F
561.
by thesz 3y ago
It does not mention Array of Structures and Structure of Arrays dichotomy, which is extremely important if you design a structure to hold data for analytics of any kind.
562.
by thesz 3y ago
"A stack is certainly the most intuitive data-structure with which to implement a lambda calculus." This depends on the evaluation order. I think you are talking about call-by-value (eager) or call-by-name orders. But normal order
563.
by thesz 3y ago
The question asked not for me (although...) or you, but for wider audience. For one immediate example, I do not know what or who Finsta is. The answer to that question, as well as preceding dialogue, would help me to navigate through Facebo
564.
by thesz 3y ago
What one considers a "profoundly stupid question" can be a question that shines a light to a heart of the problem or provides necessary context. For one such example, the question "how Facebook makes money" provides cont
565.
by thesz 3y ago
> Are there any type systems that work well for the array programming problem domain? Do a search for "haskell strongly typed tensor programming."
566.
by thesz 3y ago
This is the implementation of hashtable_rand_obj: cache_obj_t *chained_hashtable_rand_obj_v2(const hashtable_t *hashtable) { uint64_t pos = next_rand() & hashmask(hashtable->hashpower); while (hashtable->ptr_table[pos]
567.
by thesz 3y ago
Cube is sphere in L_{\inf}.
568.
by thesz 3y ago
You cannot add a loop skew optimization to compiler before compiler needs a loop skew optimization. Which it would not need at all because it is loop skew optimization (it requires matrix operations) that need a loop skew optimization. In s
569.
by thesz 3y ago
Please, do add the "two random choices" to the list. It is a noticeable omission, let me explain why. Here are some comparisons of 2-random-choices with LRU in context of CPU caches: https://danluu.com/2choices-evi
570.
by thesz 3y ago
I provided a link, you can read more about differences there. LRU suffers from linear access pattern, where it cannot degrade gracefully. Try to cache N+1 items in N-item LRU when items are accessed linearly or at random with removal. The &

More ›