I guess a question would be: how do you know your experiences are similar enough to allow understanding? — frank

Not a very clear question Frank. But in 20 years of full time work as a translator I've translated around 10 million words, I very rarely receive complaints about my translations, my work is checked by an editor and I very rarely receive corrections from them, and my customers keep coming back to me and paying for my services. Does that answer your question?

These are different meanings. In A the councillors advocate violence and B they fear violence (which has two meanings in and of itself). — I like sushi

Hm, I can see the point. Why not:

Out of fear of violence, the councilors refused to allow the protestors, whom are known to advocate violence, to demonstrate.

Or something of the like. Granted not everyone speaks casually in such a manner so it is useful for any application that plans to be relevant to be able to recognize as much variation in sentence construction as possible. Which as has been noted, is quite difficult.

Edit: And of course technically both sentences can mean either or, with a little thought. Granted we know and should assume the same meanings as in the OP, but there's nothing that prevents the opposite.

Hm, I can see the point — Outlander

No you can't. You're missing the point completely.

Would you perhaps mind explaining it then, seeing as you now hold the minority viewpoint of 'understanding' in this discussion?

Not a very clear question Frank. But in 20 years of full time work as a translator I've translated around 10 million words, I very rarely receive complaints about my translations, my work is checked by an editor and I very rarely receive corrections from them, and my customers keep coming back to me and paying for my services. Does that answer your question? — Daemon

Is it possible that this is happening in spite of a rift in understanding? Could it be that you're applying certain rules correctly, and so there are no complaints about your service, and yet there is no communing of intent?

How would you prove that this extra thing beyond rule following, this 'understanding' exists?

A. The councillors refused to give the protestors permission for their demonstration as they advocated violence.
B. The councillors refused to give the protestors permission for their demonstration as they feared violence.

In A. "they" refers to the protestors, in B. it refers to the councillors. We know this because of our experience of the world. It's an example of something a computer couldn't know.

How would you prove that this extra thing beyond rule following, this 'understanding' exists? — frank

Well, for example, there are sometimes mistakes in the source text. Maybe somebody writes "the saw blade must be touched with the fingers while it is still rotating". So I write to the customer and say "I think you missed out the word 'not' here". And they say "yes, thank you, you're right".

Does that answer your question?

In A. "they" refers to the protestors, in B. it refers to the councillors. We know this because of our experience of the world. It's an example of something a computer couldn't know. — Daemon

They only refer to the protestors and councilors respectively, because the father, or author, of the sentence determined so. Or I suppose "it simply happened that way" or as you say, that's just how "the world" (generally) works.. There are numerous scenarios, one of which has been posted previously, where it could easily be the opposite.

The same (likely) context recognition could be achieved, albeit haphazardly, with a 'word map' database.

Councilor = government, order, ruler, leader, society, peace, stability

Protestor = worker, grievance, anger, rebellion, uprising, turmoil, injustice

The more general words (violence) matched with context specific words (they), that happen to match a subsequent 'word map' of words relevant/associated with each party or subject(s) can more often than not determine which party to apply said word to. It would take a great deal of finagling, sure. But it's doable. Not with any laser accuracy, of course. Which I suppose was your point.

What exactly are we discussing and for what purpose? I do fail to see the profoundness or any possible fruit of this topic. Computers, AI =/= human comprehension. I doubt there was any disagreement at any point.

That's part of what makes it an interesting example. The sentence parses differently depending on a difference of the meaning of a word alone. And I do not think it works that way, at least correctly.

And I think @I like sushi is going in the right direction. But "The chicken is ready to eat," is simply ambiguous. The examples on the other hand,

A. The councilors refused to allow the protestors to demonstrate, because they advocated violence.
B. The councilors refused to allow the protestors to demonstrate, because they feared violence. — Daemon

have a structural problem. Bottom line is the reader has to read the meaning in instead of getting it from what he's reading. I.e., eisegesis v. exegesis. And people often get it right, but that's in-itself not right. In a perfect world an editor marks them for rewrite.

In A. "they" refers to the protestors, in B. it refers to the councillors. We know this because of our experience of the world. It's an example of something a computer couldn't know. — Daemon

To most people. Why do you keep refusing to accept this? In a place where the councillors are corrupt/vicious why not the opposite.

A computer cannot understand anything. It is a CODED not THINKING. Other than that what is your point? I don't actually see one but I'm assuming there is one somewhere that is why I'm persisting.

In a perfect world an editor marks them for rewrite. — tim wood

Precisely

I have been a professional translator for 20 years. My job is all about understanding. — Daemon

I have been a professional translator (freelance) for the same amount of years!
However, I never thought of my job as something that is all about understanding. Undestanding is of course essential, but it refers only part of the whole process. The most important things in a translation are 1) be proficient in the language you translate in (target language) and 2) be able to relay information as accurately as possible without being literal. The level of accuracy depends of course on the subcect: the more technical is a subject the more accurate one has to be. On the other hand, if the subject is literary, one can relax on accuracy and rely more on expresssion. But still, the meaning of the source text has always to be relayed.

All this is an art and and this is how I see a translation. Writing is an art. So is translation. Only that here the ideas come from someone else than youself.

The CAT tool suggests translations based on what I have already translated. — Daemon

Yes, CAT tools are very good, but mainly for technical subjects. I used them extensively in translation manuals (75% of my total workload!) But on general text, I use Google translation, which I call "pre-translation". Although in the past Google translations were quite inferior --in Greek, which is my native language, it was actually deplorable, because of the complexity of the Greek grammar-- but these days they are really excellent, even in Greek! Most probably because of their hugely increased database of both words/terms, phrases and evem full sentences. So, after that, your task is only to correct minor mistakes and trim the text in general. It's there that your proficiency in your native language comes in as the most important element. Undestanding becomes of secondary importance. It's a fact.

That's enough about translation! :smile:

***

Now, I don't know how you have reduced such an interesting topic as "The important question of what understanding is" into a translation subject! I have a lot to say about "understanding", what it is, how it works, etc. but it seems that it is not what it matters anymore! :smile:

Now, I don't know how you have reduced such an interesting topic as "The important question of what understanding is" into a translation subject! — Alkis Piskas

Because I was responding to something TheMadFool said, which I quoted at the very start of this thread:

That raises the important question of what understanding is and, more importantly, whether it is something beyond the ability of a computer AI? — TheMadFool

We could of course talk about a single language, but discussing computer translation is an excellent way to address the question of understanding.

You still aren't getting the point Tim. The two sentences are there solely to provide an example of the limits of machine translation.

What exactly are we discussing and for what purpose? I do fail to see the profoundness or any possible fruit of this topic. Computers, AI =/= human comprehension. I doubt there was any disagreement at any point. — Outlander

That raises the important question of what understanding is and, more importantly, whether it is something beyond the ability of a computer AI? — TheMadFool — Daemon

Well, for example, there are sometimes mistakes in the source text. Maybe somebody writes "the saw blade must be touched with the fingers while it is still rotating". So I write to the customer and say "I think you missed out the word 'not' here". And they say "yes, thank you, you're right".

Does that answer your question? — Daemon

It wasn't really a question. :razz:

You don't know for sure that you and your client have the same understanding.

In exactly the same way, you don't know that the world is out there as it appears to be.

You get by just fine not knowing these things. Or we could say you know one just as well as you know the other.

Can you tell me how I could get all that stuff about the store shelf and banana bread into my translation memory? — Daemon

I can sketch it out.

You need some bootstrap capabilities outside of dictionaries... things like what humans have; e.g.:

A camera does not see. — Daemon

...the ability to see. Add to that some basic sapience. The general idea is that this should have the ability to interact with reality in real time on scales roughly approximating that of your typical language using naked apes. Some of this interaction would involve exploiting "seeing" (or other kinds of sensations) in the attainment of goal oriented behaviors analogous to how we "intentionally do things"; i.e., at roughly the same levels of abstractions as the "things we do" or, more to the point, at roughly the same levels of abstractions as the "things we talk about".

Once you have such a thing, we need two more ingredients to make it final: (a) a banana, (b) a shelf. All of this, or something akin to it, would need to be in place before you can have something to map "banana" to and call it understanding.

I skipped a few steps, but it's not like I wouldn't have had to skip steps anyway at some point; I have never built such a thing.

A. The councillors refused to allow the protestors to demonstrate, because they advocated violence.

B. The councillors refused to allow the protestors to demonstrate, because they feared violence.

A computer can't understand that "they" applies to the protestors in A. but the councillors in B, because it's not immersed in our complex world of experience. — Daemon

I like this very much. Whether one could somehow, someday develop an artificial system that could deal with such a case, who knows. I lean toward your view, but I wouldn't put money on it either way.

But it is a lovely example of the sort of thing we manage easily everyday, only noticing when it goes wrong for some reason. Funny things, pronouns

The examples I gave were intended to illustrate that semantics isn't simply mapping! — Daemon

Of course it isn't. I'm surprised anyone would think it is. In point of fact, I'm not even sure what it's supposed to mean: people look up the meanings of words in dictionaries, sure, but you can't look up the meaning of a sentence in the sentence-dictionary, so if sentences have meanings, they must not "map" to them, or they must have a different kind of meaning.

To most people. Why do you keep refusing to accept this? In a place where the councillors are corrupt/vicious why not the opposite. — I like sushi

That doesn't affect the point of the example. If there were such a place, the computer wouldn't have access to that external set of circumstances.

The point is that we are able to make a judgement about the meaning of the sentences which a computer can't possibly make.

Can you tell me how I could get all that stuff about the store shelf and banana bread into my translation memory? — Daemon

I can sketch it out.

You need some bootstrap capabilities outside of dictionaries... things like humans have; e.g.:

A camera does not see. — Daemon

...the ability to see. — InPitzotl

This is just a waste of everybody's time. I mean, come back to us when there's a camera that can see and we'll have something to talk about.

ecause I was responding to something TheMadFool said, which I quoted at the very start of this thread: — Daemon

I read this of course. But it's still about undestanding ... and my wondering is still unanswered! :grin:

discussing computer translation is an excellent way to address the question of understanding. — Daemon

Well, in that case, even if you had used a more specific title, like "Computers and understanding" or something like that, it would be still inappropriate because computers do not possess any understaning!

Well, maybe it's not so important, generally and for most people. But it just happens that undestanding and communication are among my favorite subjects. I have studied them extensivley and I was even teaching about them (theory and practice) in the past ...

A. The councillors refused to allow the protestors to demonstrate, because they advocated violence.

B. The councillors refused to allow the protestors to demonstrate, because they feared violence.

A computer can't understand that "they" applies to the protestors in A. but the councillors in B, because it's not immersed in our complex world of experience. — Daemon


I like this very much. — Srap Tasmaner

So do I Srap! It seems to have caused nothing but confusion above though.

Whether one could somehow, someday develop an artificial system that could deal with such a case, who knows.

According to my theory the artificial system would need to be able to experience and interact with the world in the way we do. It would need to experience such things as pain and pleasure in order to understand what "good" and "bad" mean. Do we really want to create artificial beings that can experience pain? Surely there's enough trouble in the world with the experiencing beings we can already create?

I mean, come back to us when there's a camera that can see — Daemon

You were the one who asked me the question. You were also the one opening this thread with your OP, where you wrote this:

matching linguistic symbols (words, spoken or written) to their respective referents — TheMadFool

...and you were the one talking about CAT tools as if that had anything to do with referents.

There's a giant difference between responding to "Can you pick up some bananas from the store?" ...by showing me the phrase translated (poorly or greatly) to Dutch; and responding to "Can you pick up some bananas from the store?" ...by showing up on my doorstep with a bunch in your hand.

The examples I gave were intended to illustrate that semantics isn't simply mapping! — Daemon


Of course it isn't. I'm surprised anyone would think it is. — Srap Tasmaner

Well there are at least two in this discussion, and I was attempting to apply the Principle of Charity, which asks us to:

"Assume that the opponent is making the strongest argument and interpret others as rational and competent."

Despite, I suppose, all the evidence to the contrary.

What are you still wondering?

You were the one who asked me the question. — InPitzotl

I was kinda hoping you'd realise you couldn't answer the question. In other words, you'd realise that you can't get a computer to understand things in the way we can.

You were also the one opening this thread with your OP, where you wrote this:

matching linguistic symbols (words, spoken or written) to their respective referents — TheMadFool

TheMadFool wrote that, I was quoting him. I'm arguing against him.

I was kinda hoping you'd realise you couldn't answer the question. In other words, you'd realise that you can't get a computer to understand things in the way we can. — Daemon

Ultimately that's correct, but the gaps are really in details.

TheMadFool wrote that, I was quoting him. I'm arguing against him. — Daemon

Sorry, I misspoke here... what I meant was that in the OP that was what you quoted. TMF did indeed write that, but he didn't explain what a referent was too well; the way he explained it, a referent could be interpreted as a phrase... so the proposal could be understood that your CAT tool might understand what "water" is if it mapped "water" to the phrase: "cool flowing substance that animals and plants need".

But that's not what the word "referent" means. The referent for "water" isn't another word (it's not "agua"); nor is it another phrase (it's not "cool flowing substance that animals and plants need"). There's no set of shape you can squiggle on a sheet of paper that is the referent for water; instead, you're going to have to go turn your taps on, point to the stuff falling from the sky outside, or go find that stuff fish swim in. Humans that know what "water" means map that word to that stuff... and to do that, we form a concept of that stuff that comes out of taps, that stuff that falls from the sky, that stuff that fish swim in. The idea of such things is an abstraction; it's a model of the stuff we're made aware of by, say, seeing it; swimming in it; drinking it; and so on. And it is that model that we map "water" to when we understand it; not more words.

Humans that know what "water" means map that word to that stuff... — InPitzotl

:100:

and to do that, we form a concept — InPitzotl

Careful now...

The idea of such things — InPitzotl

a model of the stuff — InPitzotl

Are concepts and ideas and models any more harmlessly, less misleadingly identified as the referent of "water" than are phrases like "cool flowing substance"?

The point is that we are able to make a judgement about the meaning of the sentences which a computer can't possibly make. — Daemon

And?

Are concepts and ideas and models any more harmlessly, less misleadingly identified as the referent of "water" than are phrases like "cool flowing substance"? — bongo fury

Oooooh! What a great question! I think this naturally falls out of our agency. We use our senses to sense the world; as we do so, we create world models. We refer to these world models, in real time even, to "do things". But we also as part of this model "project" it as something independent from us and, well, it winds up that's a good theory of what the world is. I think something along these lines (at least for claims about the state of the external world) is what gives rise to intentionality.

ETA: Just to close the loop here... when we act in the world, we're not merely using our world models... we're literally using that world. By this I mean that we don't simply imagine ourselves walking to the sink, we walk over there. These interactions are in real time, and they are updated by real time world sensations... any difference between what our world model is and these sensations is updated by deferring to the sensed world. This is the long form of what I mean by "project" here.

You don't know for sure that you and your client have the same understanding.

In exactly the same way, you don't know that the world is out there as it appears to be.

You get by just fine not knowing these things. Or we could say you know one just as well as you know the other. — frank

I agree that I don't know for sure that my translations can be correctly understood, that's part of my own philosophical position, and I also believe that, in a certain philosophical sense, words do not carry or convey meaning. But I don't tell my translation clients about any of this.

I do think there's an abundance of evidence that we are able to understand a great deal of what we say to one another. You couldn't get people to land on the moon and come back without shared understanding of language. And on a smaller scale, you and perhaps a couple of others have understood at least some of what I've said in this discussion, as evidenced by your coherent responses.

Whether the world is as it appears to be is another (vast) question, and perhaps off topic for the Philosophy of Language forum. Personally I'm satisfied that the world is enough like it appears to be for us to travel to the moon and back, and for me to make the pasta dish I'm going to eat soon.