This makes me think of the ongoing effort by some people to train their pets to use talking buttons to communicate. These include the animals occasionally trying to convey relatively complicated ideas with simple sets of words, like "play help" ("I'm bored, come entertain me") [1] and "dog settle" ("make that dog outside stop being annoying") [2].
The meaning of "where" in your claim is very slippery. AR Luria in The Working Brain points out that damage to virtually every part of the brain can result in language dysfunction. To say that a whole lot of things happen apparently without language doesn't demonstrate that language is an extra. The brain generally isn't very modular and my alternative take would view language as accompaniment that sometimes is "just words" and sometimes touches very specific details (edit: plus allow recursive access these other brain-functionalities, as others point out). We can see how GPT-3's output is a stream of plausible things punctuated by some statement that shows "it doesn't anything" (IE, gets the implicit context wildly wrong).
This may well point to the possibility that animals have a theory of mind.
Sophisticated ToM is likely rare in the animal kingdom, not usually going beyond programmed responses to being observed and possibly empathizing with conspecific suffering.
In humans, it's a much more sophisticated process of inferring motivations, anticipating moves and a number of similar inferential capabilities plus abductive reasoning.
1. Recognize objects in the world, like a keyhole or whatever it is you want to take out of a bin.
2. Recognize your own position -- how close is your hand to the keyhole?
3. Adjust your position until it matches your goal.
My impression of things is that most of our difficulty comes from task 3. It's not so much that we have trouble recognizing where we are, where we want to be, or how to move from where we are to where we want to be. But we have difficulty defining movements that don't cause damage to nearby (or target) objects.
Your human hand has many features that allow it to grip an object securely without damaging the object. And you also have a decent sense for how to move your hand to an object without damaging the object -- or your hand -- though this problem is far from being solved in humans too. Ever stub your toe?
> Animals with very limited language capabilities have the ability to run their own lives, and to some extent manipulate their environment. Systems built from large language databases don't have that.
I'm tempted to agree with this in some capacity, language is probably just one expression of thought, much like how some people have an internal narration going on during their lives, while others think in abstract concepts that aren't necessarily expressed in language.
Here's the first article DuckDuckGo turned up on the subject, i bet someone can share a better one: https://www.iflscience.com/brain/people-with-no-internal-mon...
Furthermore, there are people who don't have the capacity to visualize things in their "mind's eye" either (aphantasia), which is another curious thing: https://www.sciencefocus.com/the-human-body/aphantasia-life-...
Of course, when you get into it, there are also some niche theories about how language could not only help us express ourselves, but shape our thought (not that dissimilar from the "newspeak" concept in Orwell's 1984): https://en.wikipedia.org/wiki/Linguistic_relativity
Perhaps it's helpful for dealing with abstraction, as abstractions don't always map well to visual representations.
What I mean is that language ability of humans and language ability of animals are not comparable; they are so different that in fact, language ability (regardless of the externalization of that language, by speech, text, signs, touch), looks very much like intelligence (meaning the intrinsic ability for language production and not the specific proficiency at some sort of externalization).
If you happen to spend enough time around animals and child humans you will quickly see that after a couple of years, kids can understand and use recursive / referential structures in a way that pets cannot. Pets seem to understand individual words, tones, and maybe "not". Think of "bring the comfy ones that your grandparents bought you to keep your feet warm" - a sentence a 3 year old can perfectly understand. No matter how much you spend teaching/training a dog that has lived in the same environment since its birth, the dog will not react to it bringing you the yellow slippers (unless you specifically trained the dog to associate the word "comfy" or "feet" to some specific slippers). The current state of parametric language models resemble much more that of a very memorious dog than that of a human.
FYI, this is the [Sapir-Whorf hypothesis](1)
> We lack some fundamental understanding about how to interact with the real world. The problem is not on the language side.
This is to a significant extent a sensing issue, not a understanding issue.
But there is a different theory, due to Noam Chomsky I believe, that language in the general sense is in fact our mechanism for higher thought, with communication being just a secondary function[0]. I believe that this is the theory that the GP was combatting with the animal example.
I believe it's a bit more complicated than this.
https://en.wikipedia.org/wiki/Referent
> In fields such as semantics, semiotics, and the theory of reference, a distinction is made between a referent and a reference. Reference is a relationship in which a symbol or sign (a word, for example) signifies something; the referent is the thing signified. The referent may be an actual person or object, or may be something more abstract, such as a set of actions.
> Reference and referents were considered at length in the 1923 book The Meaning of Meaning by the Cambridge scholars C. K. Ogden and I. A. Richards. Ogden has pointed out that reference is a psychological process, and that referents themselves may be psychological – existing in the imagination of the referrer, and not necessarily in the real world. For further ideas related to this observation, see failure to refer.
In this case, what has been ~"proven wrong" is Whorf's specific hypothesis (the reference) of the underlying phenomenon (the referent) that he was observing, whereas the general notion of whether language fundamentally/substantially affects cognition seems to remain unknown.
As a thought experiment: imagine a culture where this psychological phenomenon of "referent" is a "first class" concept in a language, and the culture that uses it insists that the distinction must be at least taken into consideration whenever a disagreement arises between two or more people....after several decades under such a model, might substantial differences arise? Perhaps this isn't a "proper" example of Whorfism, but it seems to be in the neighbourhood.
For example, there are some Australian langauges if I remember correctly where instead of words for left, right, ahead, behind (relative positioning) they use words for something similar to north/south/east/west (absolute positioning) - as in, "give me that cup to your northeast". This of course requires a habit of keeping track of your absolute positioning at all times.
However, you will not gain this ability or habit just by learning the langauge. In fact, your langauge can even express that same thing today. Instead, you will be unable to naturally use their langauge unless and until you get into the habit of keeping track of your absolute positioning.
Similarly, the people growing up in this culture may not easily learn what we mean by left and right, but they would not gain the ability to understand these concepts merely by learning English - instead, they would naturally speak English while avoiding left/right references, or simply make more mistakes than someone growing in a culture where relative positioning is the preferred form of communication.
Regardless of whether the language is a consequence of the culture (say, they imported it, or had it imposed upon them), how might one determine with accuracy that consistently disciplined speaking (which is preceded by cognition) has no causal dependence on speaking this substantially more strict language itself?
> However, you will not gain this ability or habit just by learning the langauge.
You will not necessarily gain this ability just by learning it, agreed - but using it on a daily basis is another matter.
> In fact, your langauge can even express that same thing today.
It can express it, in some fashion (typically with clumsy, annoying phrases and qualifiers that make people angry, rather than dedicated, socially normal words), but whether one actually does it on a regular basis is another matter.
There are various differences between someone who studied 3 months of 1 hour per week of Jiu Jitsu (or math, physics, etc) classes, and someone who practiced it for several hours a day from an early age.
Even in simple scenarios it can be extremely difficult to accurately predict what the consequences of something are, and language, cognition, and culture are not exactly simple. And, from a scientific/epistemic perspective, one should always keep in mind that an absence of evidence is not proof of absence....which brings to mind another thought experiment that is different but similar: what might a society be like where epistemology and logic are considered crucially valuable skills, and were taught from Grade 1 through 12 to all children, and being incorrect or even speaking epistemically unsoundly were considered seriously shameful? Might it be possible that one would notice anything substantially different in people who were raised in such a culture (and, what might be some differences we might see)? I would be quite shocked if it had no effect.
Japanese people show extremely high performance at "spatial awareness" tasks compared to whites.[1] So on average this isn't true. But it's true that there does not appear to be any benefit from either language -- Japanese-speaking Japanese show no advantage over English-speaking Japanese.
[1] On white-normed IQ tests, East Asians generally show a bit of underperformance (relative to European whites) on verbal tasks and gigantic overperformance on spatial tasks, which is traditionally averaged out to an average IQ score moderately above that of whites. Interestingly, Native Americans show the same slant -- much much stronger spatial than verbal performance -- but at a lower base level.
If we kept the same languages and made japanese people write it massively in 26 half sound symbols and made all of what you call a white european population use 3k+ ideogram drawings for word starting at 3 years old, do you think we d see exactly the same performance ?
I bet you the writing tradition is very influencial when it varies so much and has no relationship with language, especially in Japanese where they frankensteined Chinese symbols over centuries which must require an enormous mental gymnastics and redirect large processing power vs someone who can read and write everything he hears relatively simply?
No, neither.
> If we kept the same languages and made japanese people write it massively in 26 half sound symbols and made all of what you call a white european population use 3k+ ideogram drawings for word starting at 3 years old, do you think we d see exactly the same performance ?
Yes, the influence of spoken and written language appears to be zero. I mentioned this.
So SW argues that language is inherently intertwined with human cognition, rather like how the parts of the brain are all interconnected and working together as we think. Strong SWism is pretty thoroughly debunked, weak is tentatively accepted on a spectrum.
I think that they might be trying to refer to generative theory versus cognitive theory. Again, it's been a while, so I might be wrong, but:
- Generative theory is Chomsky et al. who believe that there is a single inherent underlying pattern to language that restricts it, and that one could, in theory, figure out all possible syntaxes, for example. Similar to how phonology and phonetics are obviously constrained by our biological makeup (we can only make certain sounds), the argument is that other structures are as well. Basically that although they all seem different to us, languages are more alike than they are different because they come from a single source. Also Chomsky and co. consider this approach cognitive; the linguists are just infighting.
- Cognitive Linguists believe that language, and the systems that produce it, are under evolutionary pressure and that they change and shift in response to the environment.
Basically, if you took 10k people and dumped 5k of them on one planet and 5k on a planet that was the physical opposite, forbid them to contact anyone, and visited 2000 years later, Generativists say you could find a system in common between the two and cognitive linguists say you probably couldn't, because the systems themselves would have evolved to have different starting points.
Chomsky + GG: Person is born, person hears speech, person reaches for inherent language patterns in brain and uses them to understand/learn.
Cognitive: Person is born, person hears speech, person picks up on the system from the speech and is encouraged to do so/pattern match by social reinforcement.
No linguist that I'm aware of would claim that language is just an I/O system.
Is this not a bit of a false dichotomy (substantially dimensionally compressed representation of the true underlying complexity)? Like sure, one is likely to be able to find substantial commonality, but does cognitive linguistics really unequivocally assert that you will find no commonality whatsoever? If so, that seems a bit crazy.
> No linguist that I'm aware of would claim that language is just an I/O system.
I think you're right. In Chomskyian linguistics there is a distinction between i-language and e-language (https://en.wikipedia.org/wiki/Transformational_grammar#%22I-...). E-language isn't "just I/O" but it's perhaps not far from that whereas i-language is taken as the proper subject for linguists to study.
That's an assumption. The contrary view is that language is just part of the I/O system, not where the hard work gets done.
Animals with very limited language capabilities have the ability to run their own lives, and to some extent manipulate their environment. Systems built from large language databases don't have that.
After 50 years, robotic manipulation in unstructured situations is still terrible. Rod Brooks tried to crack that and his startup, Rethink Robotics, failed. DARPA has had decades of manipulation projects. Fifty years ago, they could put a bolt in a hole. Five years ago, they got a robot hand to put a key in a lock. Amazon is still trying to get robots to pick things from bins fast and reliably.
We lack some fundamental understanding about how to interact with the real world. The problem is not on the language side.