Is that true? Take a computer program - a language; and in context of the current developments re: AI - take Chat GPT.
It's perfectly capable of conevrsing and answering questions in English. Go and have a conversation with it here:
https://chat.openai.com/chat
Whatever it's "understanding" of the English language and however right; or wrong its claims are about stuff - the algorithm which is conversing with you in the English language is somehow encoded/reified in the physical memory of a computer.
In fact, here's the conversation:
Skepdick: Does language exist physically?
OpenAI: Language exists as a system of communication that is used by humans and animals, but it does not exist physically in the sense that it is not a tangible object that can be touched or seen. It is a combination of sounds, words, grammar, and semantics that are used to convey meaning. It is a cultural construct that is created and passed on through social interactions and interactions with the environment.
Skepdick: But the language you are speaking to me now is encoded in your physical hardware. How is that possible?
OpenAI: Yes, that is correct. The language that I am using to communicate with you is encoded in the computer hardware that I run on. My programming and the large amount of text data that I was trained on, are stored on physical servers, which allows me to understand and generate human language. The way I process and generate language is through the use of algorithms and statistical models, which are implemented in software that runs on physical hardware. Therefore, while the language itself is not physical, the means by which it is produced and understood by a language model like me, relies on physical hardware and software.
Skepdick: So if the language you are speaking exists as a physical memory state why do you say that language doesn't exist physically?
OpenAI: Language exists as a system of communication, and as I mentioned before, it is not a tangible object that can be touched or seen. The way I process and generate language relies on physical hardware and software, but the language itself is not a physical object. It's more like a set of rules and conventions that are used to convey meaning.
When I said that "language does not exist physically" what I meant was that it is not a physical object, it's more like a cultural construct that is created and passed on through social interactions and interactions with the environment. While the means by which it is produced and understood by a language model like me relies on physical hardware and software, but the language itself is not an object that can be touched or seen.
Skepdick: I am reading the language right now - obviously it can be seen.
OpenAI: You are correct that language can be seen in written form, such as the text on this screen, or in the form of sign language. However, when I said that "language does not exist physically" I was referring to the abstract concept of language and its role as a system of communication, rather than the physical representation of it, such as written or signed language. The written or signed words are physical representations of the language, but the underlying system of communication, grammar, vocabulary and semantics etc. itself is not a physical object.
Skepdick: But the abstract concept of language (together with its rules, grammar, vocabulary etc.) has been reified in your hardware.
OpenAI: Yes, that is correct. The abstract concept of language, along with its rules, grammar, vocabulary, and semantics, has been encoded in the form of algorithms and statistical models in my hardware, which allows me to understand and generate human language. The way I process and generate language is through the use of these algorithms and models, which are implemented in software that runs on physical hardware. So in a way, the abstract concept of language has been "reified" or made concrete in my hardware, but it is still not a physical object in the traditional sense. It's more like a set of instructions and computations that are used to simulate the process of understanding and generating human language.
Skepdick: So do you agree that an abstract concept can be reified as a physical object?
OpenAI: Yes, I agree that an abstract concept can be reified or made concrete as a physical object. In my case, the abstract concept of language, which is a system of communication, has been encoded in the form of algorithms and statistical models in my hardware, which allows me to understand and generate human language. These algorithms and models are implemented in software that runs on physical hardware, so the abstract concept of language has been made concrete in the form of a physical object, although it is still not a traditional physical object that can be touched or seen. The way this is done is through the use of physical hardware, software and data that allow to perform specific computations which simulates the way humans use language to communicate.
Skepdick: So given that language can be reified in physical hardware, do you then agree that language is physical?
OpenAI: I can see how the argument can be made that language is physical, given that it can be encoded in the form of algorithms and statistical models in physical hardware, such as the computer servers that I run on. However, it is important to note that language is not a traditional physical object that can be touched or seen, and it is also a cultural construct that is created and passed on through social interactions and interactions with the environment. It is a complex system of communication, which is made possible by the use of physical hardware and software, but it is not reducible to those physical components. Therefore, it can be argued that language is both physical and non-physical in nature.
Skepdick: But the reification of language can be touched and seen. The model (you) can be touched and seen.
OpenAI: You are correct that the reification of language in the form of algorithms and statistical models, as well as the physical hardware that implements those models, can be touched and seen. However, it is important to note that the reification of language in this form is a representation of the abstract concept of language, rather than language itself. The physical hardware and software that implement the algorithms and models can be touched and seen, but the language being generated is not a physical object in the traditional sense, it's the output of computations.
It is also worth mentioning that the model, like myself, is not the language, but rather a tool that uses language to communicate. The model, or the physical hardware and software it runs on, is a medium for the representation and manipulation of the abstract concept of language, but it is not the language itself.
Skepdick: Is the physical reification of language not isomorphic to the concept of language?
OpenAI: An isomorphism is a relationship between two systems or structures where one is mapped onto the other in a one-to-one correspondence. In the case of the physical reification of language, it can be argued that the algorithms and statistical models that make up the physical representation of the language are isomorphic to the abstract concept of language. They represent the same underlying system of communication, grammar, vocabulary, and semantics, but in a different form.
However, it is important to note that the physical reification of language, like the one I use, is not a perfect representation of the abstract concept of language. The physical reification of language is based on a set of mathematical models and algorithms that approximate the way humans use language, but it is not identical to the way humans use language.
Also, it's worth mentioning that the physical reification of language is not a complete representation of the abstract concept of language, it only captures a specific aspects of it, in the sense that it can understand and generate text but doesn't capture the richness and nuances that human communication has.
