thesz

27 ms·

511.
by thesz 3y ago
The reliance on the tool in front of you can constrain you. Some tools are fads, some make you less of a thinking person. When I faced a second wave of IDEs I figured that any IDE is a fad - there will be better IDE in near future and your
512.
by thesz 3y ago
> But, you can generally prompt around this sort of thing by asking the model to adopt a preferred context such as "you're an expert xxx", providing of course that is did see such expert-tagged material in the training set
513.
by thesz 3y ago
> You can always prompt it to respond differently though. No, you cannot. I rarely do anything with LMs [1], but if I do I ask them to code "blocked clause decomposition" [2] in Haskell. Blocked clause decomposition is very, ve
514.
by thesz 3y ago
MathCraft [1] is little more [2] than chat. [1] https://www.cyc.com/mathcraft/ [2] https://www.youtube.com/watch?v=pbrp7MzBDm0
515.
by thesz 3y ago
https://en.wikipedia.org/wiki/Cyc#MathCraft Quote: One Cyc application aims to help students doing math at a 6th grade level, helping them much more deeply understand that subject matter... Unlike almost all other
516.
by thesz 3y ago
digit late Middle English: from Latin digitus ‘finger, toe’; digit (sense 1) arose from the practice of counting on the fingers. "Digit" is the decimal digit.
517.
by thesz 3y ago
Thanks! You saved me. ;)
518.
by thesz 3y ago
His page [1] is pure awesomeness! He posed such an uncomfortable questions in his (and his students) reseach.
519.
by thesz 3y ago
Every CPU and GPU now has popcount [1] instruction. You can thank US three letter agencies for that [2]. [1] https://en.wikipedia.org/wiki/Hamming_weight#Processor_support [2] https://vaibhavsagar.com&#x2
520.
by thesz 3y ago
MIPS (R2000) did not outdone Intel and/or Motorola at the time, those last, but not least two enjoyed all "economics of scale" of the world. MIPS was slower clock-frequency-wise than competitive offerings from Intel and/
521.
by thesz 3y ago
> regular RISC platforms were riding an economy of scale I've found a fascinating page [1]. [1] https://computer.fandom.com/wiki/Instructions_per_second#Tim... ARM (1986) has 4 Dhrystone MIPS, and Motorola 6802
522.
by thesz 3y ago
I recently made a comment that neural models cannot provide chain of reason, while symbolic methods can: https://news.ycombinator.com/item?id=39759033 The vulnerability in the post is directly linked to that inability, in m
523.
by thesz 3y ago
a^2 = 1, first base vector is a regular one b^2 = -1, second base vector is "imaginary" ab = 0, base vectors are orthogonal (a+b)^2 = a^2 + 2ab + b^2 = 1 + 2\*0 + (-1) = 0 Trick is taken from conformal geometric alg
524.
by thesz 3y ago
To illustrate: https://arxiv.org/abs/2401.05566 One can teach LLM to write good code now and bad code in future (when everyone lower their guards). And no one can prove who and how made LLM to do that. Also, no one can
525.
by thesz 3y ago
What cannot I do myself? Look at the output of gcc -ftree-vectorizer-verbose=2? Run the verifier after picosat? I did both.
526.
by thesz 3y ago
I think I have to know what chain of reasoning is behind this or that fact and/or deduction. i would like to be able to verify that. For example, the proof of the absence of solution in SAT should be accompanied with the easily verifia
527.
by thesz 3y ago
The ability of compilers to make code faster is 12 (twelve) times slower than Moore's law: they double the program's speed in 18 years [1]. [1] https://proebsting.cs.arizona.edu/law.html This might seem discouragi
528.
by thesz 3y ago
Superoptimizers: https://en.wikipedia.org/wiki/Superoptimization Also, program distillation: https://www.researchgate.net/publication/220989887_Distillat...
529.
by thesz 3y ago
Can you elaborate on where the delays came from?
530.
by thesz 3y ago
> Addition was not the bottle neck for the 386. It is a bottleneck for MIPS, SPARC, Alpha and not for 386. How so?
531.
by thesz 3y ago
My assumption is that speedup is less than lambda's ratio.
532.
by thesz 3y ago
And you are also confirm that in order to have higher clock frequency you need more than just pipelining. Thank you. I also think that 1.5x difference in clock speeds cannot be directly attributed to the difference between node size (lambda
533.
by thesz 3y ago
My point is that you cannot get design much faster in terms of clock frequency by just pipelining. Pipeline unrolls state machine and overlaps different executions of the state machines. But the bottleneck, which is addition, is there in al
534.
by thesz 3y ago
They do not cite [1], a paper on (learned) variable computation in RNN, applied to language modeling, that predates their work by almost 8 years. [1] https://openreview.net/pdf?id=S1LVSrcge Microsoft also had something alik
535.
by thesz 3y ago
You are not quite right about pipelined design being faster. At least, not without substantial effort. https://en.wikipedia.org/wiki/R2000_microprocessor "The R2000 is a 32-bit microprocessor chip set developed by
536.
by thesz 3y ago
> ensure consistency and reliability across two different database stores. This is main known known. And this is hard thing to attain. My favorite story on that is testing of tendermint consensus implementation [1]. The testing process f
537.
by thesz 3y ago
Karel Čapek: R.U.R. His short stories are beyond fascinating and adorable, BTW. One, for example, equates accounting to hunting and detective work. The Stars My Destination by Alfred Bester. Works of James Herriot.
538.
by thesz 3y ago
You can have target language to be as far fom host language as you like. For one example, again, borrowed fom Haskell universe, is Atom [1]. It is a embedded language to design control programs for hard real-time systems, something that is
539.
by thesz 3y ago
Quite the contrary. You need to use existing facilities (type checking, pattern matching combinators, etc) of a good implementation language as much as possible before even going to touch yacc or something like that.
540.
by thesz 3y ago
> actually creating a working language is a big overhead Languages, with first class values, pattern matching, rich types, type inference and even fancy RTS, often can be embedded in Haskell. For one example, it is very much possible to

More ›