/tech/ - Technology

>>26589
>use google?
>nah, I will just harass people
westoids cointelpro-ing themselves out of boredom
https://paulcockshott.wordpress.com/2020/08/08/why-the-lambda-calculus-is-not-really-equivalent-to-the-universal-computer/

>>26590
Alright, my fault for not googling first.

>westoid
my location: https://yandex.ru/maps/-/CDXJV6ic
jannies can check the ip

>>26590
>This forces us to consider computation as a material rather than an ideal process, and actually computer designers end up being preoccupied with very material considerations: minimising power use, disposing of waste heat, ventilation etc.
Claiming any model of computation makes you think "more materially" about existing computers is pure idealism. The abstraction of lambda calculus closely resembles actual classes of programs, posing some essential questions about language implementation like the funarg problem. Turing machines reveal as much about programming as brainfuck. Interest in them stems purely from what they say about the physical limits of computation, which are far from unique to it and don't magically make you come up with better cooling systems. Mentioning a failed research project from the ai boom as some kind of gotcha is frankly embarassing.

>>26592
you didn't get his point. dunning-kruger moment

Cockshott makes a statement so banal that lay people can easily understand it: You cannot transcend the constraints of hardware through the software running on it. And then this thread is full of shitting and farting. Makes me wonder whether the posts are someone with an axe to grind.

>>26592
Well, the von Neumann model is just a physical embodiment of the turing machine. Where's the lambda calculus embodiment (no turing crutches allowed)?

>>26594
> You cannot transcend the constraints of hardware through the software running on it

No, he said that the lambda calculus is not an actual machine and hardware architecture, and that it is a theoretical set of rules, that is basically impossible to be implemented on real hardware, that it needs a turing crutch anyway. This is what's confusing to me, is it really that way

>>26595
The cock man is pretty smart, no shame in getting filtered by him. I like how he's against this lgbt shit (I don't know if hating on them is allowed on leftypol so forgive me jannies in advance)

>>26599
>Well, the von Neumann model is just a physical embodiment of the turing machine.
The von Neumann and Harvard architectures specify the dataflow of a machine. Comparing the PC to a tape head misses the entire point of their discussion in computer design.
>No, he said that the lambda calculus is not an actual machine and hardware architecture, and that it is a theoretical set of rules, that is basically impossible to be implemented on real hardware, that it needs a turing crutch anyway. This is what's confusing to me, is it really that way
No, he is making the former, far more banal statement. Evaluating the leftmost part first (normal-order evaluation) guarantees reaching the normal form of an expression. Lambda expressions in lisp use a mix of applicative- and normal-order evaluation. See Peter Kogge's The Architecture of Symbolic Computers for the abstract implementation of a lisp interpreter. mods fucked up DJVU uploads

>>26599
I personally don't give a shit about the queers, but Paul going on about them does great harm to the movement. it is funny though, watching radlibs get mad at him

Modern superscalar architectures have just barely more in common with a Turing Machine than they have with Lambda Calculus. I think Wangwallop's point is moreso that it's hard to actually achieve the promise of trivial parallelization that functional programming autists have been chasing for decades. This of course isn't what he says in the title or in the talk he's responding to, but it's the argument he makes after conceding the point that UTC and LC can each emulate the other. Could it be that trans derangement syndrome has fried his brain?

>>26590
This is just the usual Marxist obscurantism. The claim that they are equivalent is a mathematical claim. Saying that it does not hold because it holds only in mathematics is completely missing the point.

>>26636
>obscurantism
I found that article very easy to read. He's talking about computers and programs running on them and gives a real-world example how abstract thinking that too removed from the actual hardware lead to lousy performance.

>>26638
Of course you did, it's supposed to look easy to get you to agree with him, while he's misleading you. He's not talking about computers, but mathematical models of computations. His claim is that two specific mathematical models are not equivalent because actual computers resemble one of them more closely than the other. Which is obviously bullshit to anyone who knows anything about mathematics. Actual computer scientists use the "idealist abstractions that the LC encourages" not to design hardware, but to mathematically reason about programs, which is easier to do as the lambda calculus more closely resembles regular mathematical reasoning than Turing machines. We would be in deep shit if the two would not be equivalent, as all the results achieved this way would be worthless. But thankfully they are equivalent as everyone knows. I don't know what Cockshott's problem is and why he tried to portray a mathematical statement as some deep philosophical struggle between materialism and idealism, but it's bullshit and you shouldn't fall for his bullshit just because he sounds like some Marxist textbook from the Soviet Union.

>>26639
<It depends on what one means by equivalent. In an ideal, or idealist, sense they are equivalent. If a function can be computed by the lambda calculus, the function can be translated into a form that a universal digital computer can also compute and vice versa.
<But from a materialist standpoint they are very different.
he addresses that point. reading comprehension

>>26640
I addressed that:
> he tried to portray a mathematical statement as some deep philosophical struggle between materialism and idealism
It's pretty clear what people mean when they say they are equivalent. It's a mathematical statement, not one of idealism or materialism. Him making it a question of idealism and materialism is precisely the obscurantism I was speaking of.

>>26638
>it's supposed to look easy to get you to agree with him
What if the argument actually is easy and you are literally insane?
>He's not talking about computers
I can only ask anybody skimming this thread who is still unsure about this to just RTFA.
https://paulcockshott.wordpress.com/2020/08/08/why-the-lambda-calculus-is-not-really-equivalent-to-the-universal-computer/

>>26642
The argument is easy: they are two mathematical models of computation that have been mathematically proven to be mathematically equivalent to mathematically each other. Saying that they are not really equivalent because you can imagine "equivalent" to mean something other than "equivalent" is not an argument.

>>26641
>>26643
he is not saying that they aren't mathematical equivalents as you claim he is (again, reading comprehension >>>/tech/26640 ). his point is that they are only equivalent in the mathematical sense. then he points out the non-mathematical differences and an example of the effects that those differences had in real life

or rather, that the mathematical equivalence is just a detail
"2 + 2" is equivalent to "4", but they are for example different in the number of characters used to represent them (5 vs 1). I struggle to see how you can't understand that two things can be mathematical equivalents but still different in other aspects

>>26644
They are mathematical models, how else could they be equivalent? There are no non-mathematical differences because they are entirely mathematical. There are no real Turing machines, Cockshott couldn't have done the same experiment that he did in OP's picture because he would have died of old age while constructing an infinite tape.

>>26645
Yes, and if you wrote 2+2 it would have been 3 characters against 1, and that proves what? I could even say that 2+2 is four, now it's 3 against 4! This says a lot about materialism.

>>26599
>the von Neumann model is just a physical embodiment of the turing machine
No it's not, they do not resemble each other in any way.

>>26645
No you see it doesn't make a difference in hardware because our highly abstracted models show that it doesn't make a difference once we have omitted enough differences by simplifying our models that we compare to figure out whether there are differences between the things that these models are simplifications of. When Cockshott refers to the performance of an actually existing physical system that he observed, that is him just imagining things.

>>26646
so you say that there are no differences, but then claim that the experiment would have been different for the turing machine. interesting
math doesn't exist btw, so your first point doesn't really make sense. nothing is entirely mathematical because they are just thoughts in brains

>>26647
you are almost there. now imagine that instead of using decimals, post numbers were represented as series of +1+1+1… it would be the mathematical equivalent, but wouldn't it affect the user experience?
and now imagine that instead of the number of character, we were talking about more complex differences, wouldn't that eventually produce different outcomes in real life implementations of mathematically equivalent concepts?

>>26649
Both the lambda calculus and Turing machines predate computers, they are not simplifications of them, in fact they have very little to do with them.

>>26650
I said that they are equivalent, not that there are no differences. Stop embarrassing yourself.

>>26652
>There are no non-mathematical differences
I accept your concession

big L btw, I don't know how you are going to come back from this one

>>26640
This is the core of his argument:
>The machine that Turing proposes is one directly modelled on what a person does when they compute. So Turing was from the start going in at a lower level than Church. Church relied on there being a human computer to do the actual work with his calculus. Turing asks what does a mathematician do when they apply any calculus?
>He then proposes a mechanism that mimics the essential steps that the human does. He thus comes up with something more general than Church, since once you build the machine and supply it with energy it can do any calculations – including acting as an interpreter for the LC.
>This forces us to consider computation as a material rather than an ideal process, and actually computer designers end up being preoccupied with very material considerations: minimising power use, disposing of waste heat, ventilation etc.
>From the Universal Digital Computer you can move on to see that there are material limits to computation over and above the logical limits to computability originally outlined by Turing in his first paper. Limits to speed set by the speed of light, thermodynamic limits (Landauer limits). If you approach the issue from the level of the LC none of this is apparent.
As already said in >>26592, he emphasizes the more material form of a turing machine and concludes that the associated model of computation leads to "thinking more materially", nevermind that the exact same argument could apply to programming a jacquard loom and that the "idealist" model has already contributed to significant theoretical advances.

tl;dr - While the models are mathematically equivalent, he claims they vary widely in conceptual power in favor of the turing machine.

The problem with lambda calculus is that you eventually run out of other people's imperative code

>>26654
beta reduce yourself, retard

>>26653
There are mathematical differences. But they are equivalent. How hard is that to understand?

I don't want to think about the underlying machine, I just want to execute code.

>>26679
Wdym you don't want the size of the cacheline to run snake?