Comment by adimitrov | Hacker News Reader

Comment by adimitrov | Hacker News Reader

Parent Full thread

"Good enough" is a curious notion. I'm an advocate of "good enough" is good enough, but sometimes it can be quite tricky to decide when something is good enough. And if. Sometimes, you figure it out down the road; and you'll be in deep trouble if you find out it wasn't!

Anyhow, there have been many "good enough" languages — mostly, it depends on cultural factors. Latin is still present in today's traditional sciences (biology, medicine,) French is the language in arts and cuisine, Italian is dominant in music theory, Greek infests philosophy and logics, as does German, which also has a strong foothold in engineering and mathematics. And what do we learn from that? There is no one language. There cannot be one. At least, we haven't figured out a way.

Over time, dominant languages rise and fade. That's why I loved Firefly for incorporating Chinese :-D English will fade (most probably.) Also most probably, it will stay in the computing sector, just like Italian did in the music sector.

> there are more students taking SATs in China in ENGLISH than in the US+Britain

Yeah, that's like saying more Indians write Java code than Americans. What does it mean? Java is "good enough?" For some tasks, but not for all.

That Rome to WWW line of yours is quite eurocentric. Can't put my finger on it, but "American web pages" look different in Asia than they do in the West. The cultural values of ancient Rome had no real influence there — except you mean to imply Babbage influenced C#.

My knowledge of LISP doesn't go beyond programming-book examples. I somehow never got into it — I blame Haskell!. But you're certainly mistaken that language recognition is sequential. For all we know, it most likely isn't. And for Chinese, for example, it really definitely isn't. Reading a Chinese poem is reading backwards, across, down, and up at the same time in order to get the full meaning.

I don't get how LISP is closer to natural languages than to programming languages?

Anyhow, my take on the situation of the OP: if he stays in the sector he's in and doesn't move out before retirement, he's found his last language. Otherwise, different tasks and developments are going to force him to adapt. There is no one best language. We've been trying to find it for thousands of years, and it's been rather fruitless so far.

English is and will be the de facto language of globalization. All of the world's elites speak it in every country. People will speak their local language, AND English. I think this is a given. http://www.youtube.com/watch?v=6ftF8sXzoWk

English will certainly outlast Chinese. The constitution of India is written in English and India will soon far eclipse China in terms of population. I think it's a given, is assumed by most people, and not overly worth arguing

> But you're certainly mistaken that language recognition is sequential.

The initial input and output of [spoken] language is sequential by definition and structure of the ear and voicebox. I simply meant that you can not speak two words or hear two words at the same time in the manner that you can see two words at the same time. You can see indentation in a text document\program, but you have to add extra syntax in spoken language to convey this.

Internal language recognition\parsing is incredibly LISP like. I'll try to add a more detailed explanation of what I mean by this in the future when I have more time.

[ basically when the human brain processes information, there are limited resources, right. The working memory buffer for the human brain can only hold 5-7 "chunks" at the same time. You have to resort to the phenomena of "chunking" to process more information. You should find this article very interesting: http://en.wikipedia.org/wiki/Chunking_(psychology) So basically we use "operators" to chunk things together. We could have a "phrase" operator that chunks a noun, verb, and object together. What's great or genius is that this "phrase", or function, is ALSO a chunk. and you can combine those chunks together in the same manner even if they are functions. This is like LISP: functions are data and data are functions. This is the "big deal" about LISP. It's similar to in the human brain, where everything is a "chunk". As long as you have processed sub-ideas into a chunk and have given a name for that chunk, you can start wielding bigger ideas. ]

[ Spoken languages are all GOOD and reflective of human thought. Anyone is GOOD ENOUGH. The differences between spoken languages are small. The differences between LISP programming languages and non-LISP programming languages are vast. LISP is to thought as Python is to English. LISP is the most similar to an evolutionarily evolved structure (human thought). Therefore a LISP implementation could also share high survivability. And could survive longer than a human lifespan. ]

> What's great or genius is that this "phrase", or function, is ALSO a chunk. and you can combine those chunks together in the same manner even if they are functions. This is like LISP: functions are data and data are functions.

These two sentences have no apparent relation, except that you said they did. Spoken language sentences do not evaluate each other and none of them are apparently functions. Also, spoken languages don't have punctuation…

I think (s)he's saying that human languages have composable, tree-like structures. There's nothing Lisp-specific about that, though. Prolog would work just as well, or ML (human languages may be statically typed), Joy, etc.

Also, I'm an amateur linguist at best, but the recursive structures in human languages generally don't get more than a couple levels deep - the linguistic stack seems to get smashed pretty easily.

> but the recursive structures in human languages generally don't get more than a couple levels deep - the linguistic stack seems to get smashed pretty easily.

Sentences yes (figure 7 chunk limit), semantic associations tokened by words, no.

> Spoken language sentences do not evaluate each other and none of them are apparently functions

I tried to clarify the proper analogy at the end. LISP is not directly analogous to spoken language (the surface level), LISP is directly analogous to human thought (that interprets language internally). This was really the topic for an article and not a comment rant, sorry.

If I'm hungry I would say "Hey, give me a Hot Dog".

I wouldn't be drawing diagrams of synapses and how the chemical signals flow when I'm thinking about that Hot Dog.

The issue of language parsing and how Lisp-like it is, it's incredibly irrelevant.

>I wouldn't be drawing diagrams of synapses and how the chemical signals flow

I actually think that's a pretty good analogy for what programmers do. You program the give function to accepted the variables nominative (you), accussative(hot dog), and dative(me). Now lets say you want to create multiple functions (give, send, take) that are operators on nominative, accussative, dative and want to reuse code easily. Or you want to use your "give" function to define what a "gift" is. No problem in LISP.

Well yes ... LISP is like a language framework, with which you can evolve your own languages suitable for problem domains.

But I just don't get the correlation between its direct representation as syntax trees and being more natural to people (which the parent was arguing).

A language can have hooks for evolving into what you want, it's just incredibly hard to do it without a LISP syntax. But natural language ain't easy to parse either.

>being more natural to people

My argument was actually that things modeled after genetically evolved structures share in the advantages of increased genetic fitness inherent in that design. i.e. Survivability and future-proofness. [i.e. the OP may indeed not need to ever learn another language before he dies] It was not an argument that genetically evolved structures are easy to grok.

> English will certainly outlast Chinese.

That's a conjecture at best. Chinese has so far outlasted any language still actively spoken on the planet. In fact, the Chinese writing system is the oldest continually used writing system. The cultural heritage behind this is huge.

And while you're certainly right that India has had a lot of English influence, you shouldn't forget the recent colonization history of India. It's still part of the Commonwealth, in fact. And when I say 'recent' I mean that in terms of some couple of centuries. I have no means of predicting the political state of the world in another couple of centuries. Do you? Thus, I'd be careful to make claims such as "X is going to outlast Y," except you've got very good reason to do so.

> You can see indentation in a text document\program, but you have to add extra syntax in spoken language to convey this.

You're forgetting prosody and gestures, which are an integral part of spoken language. Also, natural language has 'pragmatics,' a process of eliciting and cancelling implicatures and presuppositions. Natural language is anything but linear or sequential. In fact there's usually several layers of information-transfer going on at once.

> Internal language recognition\parsing is incredibly LISP like.

Nope, it isn't. First of all, we have no idea what's going on internally. Neurolinguistics and neuropsychology aren't far enough yet, and at this point, it's just guess work. But what we do know, is that LISP and natural languages live in entirely different complexity classes of language. LISP is context free - and natural language isn't. Probably, it's not quite context sensitive yet, but probably somewhere in between (there's dozens of disjoint sets of language complexity classes between those two.) LISP cannot create, for example, cross-serial dependencies, like they occur in swiss German and Danish are in a complexity class not provided for by LISP-like grammars.

I know about "chunking." In fact, I'm doing research in a related subject (computational semantics and formalisms for representing meaning.) It would all be well and good: you have chunks and operators, and then you write the grammar, and voilá, you've got your NL parser. Just that it's not that easy. Anaphora resolution, world knowledge and linguistic knowledge, pragmatics, weird syntactical phenomena (say, pied piping, etc.) have made this infinitely harder. If all you want to parse is "Hey, give me a hot dog," we've been here for the past 30 or 40 years. That's not hard. If you want to parse the English Wikipedia, it's impossible. You can only approximate.

The differences between spoken languages are vast. The differences between functional and imperative languages (because that's what you're referring to, I think) are minuscule in comparison. I can learn the functional paradigm and pick up any functional language within weeks, probably less (for example, I can read and write Erlang, OCaml, ML and LISP, even though I only really know Haskell and Lua.) In order to learn a new language (even a closely related one, like say from English to Swedish) you'd take at least a year in order to even be able to understand basic conversation or news. It's mostly the vocabulary, but also the speed and the multiple layers of information transfer that I alluded to earlier that make this a lot harder.

It sounds like you've given the topic some thought, but your ideas seem (to me!) all too reminiscent of the optimism computer scientists had in the 50s to 70s or so. People thought computer systems comprehending natural language were just a stone's throw away, because, hey we had all the basic machinery! Well, they weren't. It seems language is a lot more complicated than we all initially thought.

> I have no means of predicting the political state of the world in another couple of centuries. Do you?

We can look at futurists who claim they do. We can look at big trends that have occurred since the Big Bang. And the "futurists" all say we are "globalizing" and becoming more "interconnected". Transforming from provincal economies and governments to world governments and super-national economic trading blocks. And that everyone in the world is learning English. All the world's elites speak English. And the other member's of their community, when they see that the elites know English, force their children to learn English because they want them to grow up to be like the current elite. In Indonesia, there is currently growing concern that their local language is dying off as parents overencourage their kids to learn English from a young age.

> You're forgetting prosody and gestures

Which are also not present when written down. Natural language has to have a verbose enough in syntax to be compressed to the intersection of bandwidth features present in both Oral and Written. Without suffering too high of information loss. In other words yes prosody and gestures are important, by they are supplementary to the core feature set. The syntax also has to be verbose enough to not loose too much information when transferred back and forth between paper and speech. This is not an important requirement for programming, which can exploit its single medium more extensively.

> The differences between spoken languages are vast.

The differences in brain anatomy between those speaking different languages is not. [ A function could contain infinitely many points. Yet it is possible for all these points to all have exactly the same derivative. And once you have modeled the derivative, you have a certain degree of predictive confidence in all future positions of the function. LISP is similar to an underlying derivative of a function generating programming languages, while still being usable as a language itself. ]

>and voilá, you've got your NL parser.

>computer systems comprehending natural language were just a stone's throw away

I completely agree is it nearly impossible to write an authoritative model of processing system of the brain from top-down conjecture. Because semantic association is defined by its relation to other chunks, we will NEVER have a perfect database of these relations, unless we map the human brain. That's why we have computational neuroscience, and are simulating cortical columns on neuron emulators, which will be able to map long before we would armchair philosophize it due to the complexity. I assure you I can very skeptical\pessimistic of "intelligent design".

I think what we can agree on in relation to the OP is that when he says LISP is the last language he'll ever learn it's kind of cheating the question he imposes on himself because it's flexible enough tool to morph into any language depending on the problem domain. And a specie's flexibility and capacity for adaptation is the best measure of survivability.

[ please hit me up via email if you want to drop links to journal articles so I'm not as ignorant to your field of research ]

[ Also please check out this article: http://www.google.com/hostednews/afp/article/ALeqM5jz3FEk2KJ... "Wired youth forget how to write in China and Japan" . English by evolutionarily choosing alphabet over characters has proven more adaptable and survivable ]

>There is no one language. There cannot be one.

Especially when you consider that language isn't just spoken words. It's a whole way of thinking. A big part of the culture itself. That's not going to just die off because English speakers are too lazy to learn any other languages.

Reply on news.ycombinator.com