What the Evolution of Language in Humans Means for AI

0
240
language
MIT Professor of Linguistics Shigeru Miyagawa | YouTube.com

Language is always changing, but there are some basic principles that hold true no matter what language you speak. It is possible that the ways that language affects culture and the brain, can be studied to help steer the direction of the future of AI. 

Same Same But Different

In his new book, Agreement Beyond Phi, MIT linguist Shigeru Miyagawa explores the concept of universal languages by analyzing similarities between a range of languages.

This is a topic in linguistics that gives a fresh take on the science of words and their construction, and Miyagawa is hopeful that his research and analysis will allow him to apply his theory to more languages out of the Indo-European spectrum.

In his book, he makes the argument that all languages have allocutive agreement, which is defined as “a morphological feature in which the gender of an addressee is marked overtly in an utterance using fully grammaticalized markers.”

For the non-linguists out there, it’s kind of like a form of subject-verb agreement that allows for formality to change the way that a verb is spoken. When speaking French, for example, to say to friend that “he has” something would be “tu as” but for a professor, or doctor, the polite version would be “vous avez.”

Shigeru Miyagawa of MIT's linguistic research may shed light on the evolution of AI.Click To Tweet

This is notable in Basque, as well as Japanese, which has a certain method of “politeness making.” In the book, he details the similarities between languages like French, English, Basque, Japanese, Dinka, and Jingpo, but he would like to continue exploring further.

There are tomes, like Globish by Jean-Paul Nerrière, which attempt to provide a “universal language” by picking the dominant business language (thought to be English) and consolidating it for professionals who want to learn quickly.

It is intended to be the lingua franca of the world.  Even without considering the weird colonial implications of language, it is not a functional language that draws on common traits of communication. There is no complex grammar, no analogies, the language is economic in both senses of the word. A new universal language might be a greater step towards our understanding of language and how it affects the brain.

An AI And a Linguist Walk Into a Bar

Because language is like mathematics in a lot of ways, having AI that speaks several languages is common.

But although they AI is capable of speech, there is still something missing in terms of the ability to understand and contextualize. Though AI is often able to easily translate and decipher, they are still not capable of having an intuitive understanding of things that happen in language–like allocutive agreement.

It will be important for linguists to work with AI scientists to really advance the ways that AI can communicate.

As technology becomes more advanced, and with computing being more effectively modeled off of brain processes, the ways that AI learn and use language will also change. How will the AI “brain” be able to respond to, and learn about language in the same ways that humans can?

Imdb.com

In an interview with Slate, Betty Birner, a professor of linguistics and cognitive science at Northern Illinois University discusses the behind-the-scene linguistics science behind the sci-fi film Arrival.

Many people believe that learning languages changes the ways the brain is wired and how it processes things.

This made me wonder if, with the invention of more computers that use brian simulation as a model,  AI will also experience a sort of “rewiring.” Could it be possible that the AIs dominant language could change the way that it computes?

Birner explains that’s not a concern, for humans or AI, because that’s not exactly how language works.

The Sapir-Warf hypothesis has two interpretations in the world of linguistics.

The first is linguistic relativity:

Which means that there is a correlation between language and worldview. For example,  the environment selects for successful adaptive traits: it is believed that Europeans typically have lighter skin because sunlight decreases the further from the equator one travels, and dark skin makes it more difficult to produce Vitamin D in limited sunlight. Or, Sub-Saharan Africans tend to have curlier hair because it protects well from the sun to help decrease the rates of skin cancer, but the wiry shape and more sporadic distribution of follicles also help heat escape from the scalp more effectively to decrease the risk of over-heating in tropical climates. Conversely, Europeans tend to have straighter hair and more densely-packed hair follicles because these adaptations are more advantageous in colder climates.

For example, the natural environment selects for successful adaptive traits: it is believed that Europeans typically have lighter skin because sunlight intensity decreases the further from the equator one travels, and dark skin makes it more difficult to produce Vitamin D in limited sunlight. Or, Sub-Saharan Africans tend to have curlier hair because it protects well from the sun to help decrease the rates of skin cancer, but the wiry shape and more sporadic distribution of follicles also help heat escape from the scalp more effectively to decrease the risk of over-heating in tropical climates. Conversely, Europeans tend to have straighter hair and more densely-packed hair follicles because these adaptations are more advantageous in colder climates.

Or, Sub-Saharan Africans tend to have curlier hair because it protects well from the sun to help decrease the rates of skin cancer, but the wiry shape and more sporadic distribution of follicles also help heat escape from the scalp more effectively to decrease the risk of over-heating in tropical climates. Conversely, Europeans tend to have straighter hair and more densely-packed hair follicles because these adaptations are more advantageous in colder climates.

Conversely, Europeans tend to have straighter hair and more densely-packed hair follicles because these adaptations are more advantageous in colder climates.

Culture is shaped by common behaviors resulting from common genetics. Europeans would logically care for their skin and hair differently than Africans. Also, the fauna and flora that are available determine the group’s diet (yams don’t grow in cold climates).

Europeans would logically care for their skin and hair differently than Africans. Also, the fauna and flora that are available determine a given group’s diet (i.e. yams don’t grow in cold climates).

Studies show that Europeans have more similarities with Neanderthals than do Africans. | Dailymail.co.uk

If we view language as a subset of culture, then each group would logically also have different concepts and words for hair-related tools like combs and brushes.

The second way to interpret Sapir-Whorf is linguistic determinism, which means that language actually shapes your reality. This definition is a little bit more extreme, and most linguists do not accept it.

Can Sapir-Warf Apply to New Language Learning AI?

Language is always changing and evolving. What will it be like to have tech that picks up on slang, and is able to do this in not just your language, but multiple kinds of languages?

As the AI continues to gain a greater understanding of language, it gets closer to sentience. If we were still running with the computer as brain theory, do you think that the language that you speak to your AI will change the way that it perceives the world, or rather, the way it computes?

But, we are still very much affected by our genetics and resulting cultures. If culture is a result of environmental differences and our unique adaptations to them, how would it be truly possible to create and adopt a global culture while those differences still exist?

banner ad to seo services page