A neuroscientist explains why it’s inconceivable for AI to grasp the language

We use or as a meaning of meaning or Signed language To understand our experiences on this planet around us. The emergence of Generative artificial intelligence like for instance Chatt (Use Great -speaking models) Call the term on the way to define “meaning”.

A well-liked characterization of AI tools is that they “understand” what they do. Nobel laureate and Ki pioneer Geoffrey Hinton said: “What really surprised me is how good neural networks are in understanding natural language – That happens much faster than I believed…. And I'm still amazed that you just really understand what you say. “

Hinton repeated this claim in an interview with Adam SmithChief Scientific Officer for Nobel Prize. In it, Hinton explained that “Neuronal networks are significantly better within the processing of the language As all the pieces that was ever produced by which Chomskyan School of Linguistics. “”

Chomskyan Linguistics refers back to the American linguist Noam Chomsky's theories concerning the nature of human language and its development. Chomsky suggests that there’s a universal grammar in humans that enable the acquisition of a language from birth.

I actually have examined how people have understood language for the reason that Nineteen Nineties, including More than 20 years of studies on the neuroscientific of language. This included the measurement of brain wave activity When people read or hear sentences. In view of my experience, I don’t have to appeal to the concept that AI can “understand” – despite the growing popularity of this faith.

Create text

First, it’s unlucky that almost all people connect text with natural language on a screen. The written text refers to – but not with the language.

For example, the identical language will be represented by very different visual symbols. For example, have a look at Hindi and Urdu. At the conversation level, These are comprehensible for either side and due to this fact taken into consideration The same language by linguists. However, they use completely different writing scripts. The same applies to Serbian and Croatian. The written text will not be the identical as “language”.

Next, let's take a have a look at the claim that algorithms for machine learning “understand” the natural language. Linguistic communication is principally nose to nose, in a certain environmental context that’s shared between speaker and listener, in addition to information similar to spoken tone and pitch, eye contact in addition to facial and emotional expressions.

The meaning of the context

There is rather more to grasp what an individual says than simply understanding their words. Even babies who are usually not yet experts within the language can Understand the context -notes.

For example, let's take the easy sentence: “I'm pregnant” and his interpretations in several contexts. If I were pronounced at my age, my husband would fall dead. Compare the understanding and response to a youngster who tells your friend about an unplanned pregnancy, or a girl who tells your husband after years of fertility treatments.

In any case, the news recipient writes a special sort of importance – and understanding -.

In my very own Recent researchI even showed that The emotional state of a person Can change brain wave patterns if the meaning of a sentence is processed. Our brain (and thus our thoughts and mental processes) are Never without emotional contextLike other neuroscientists, too identified.

While a pc code in the shape of text can react to human language, it doesn’t come closer to what people – and their brain – achieve of their understanding.

It needs to be remembered that employees in AI speak about neural networks, they computer algorithms, not concerning the actual, biological Brain networks This is characterised by the brain structure and performance. Imagine confusing the word “flight” (as with birds, mountain climbing) in comparison with “flight” (like on flight routes) – this could lead on to serious misunderstandings!

Finally, let's examine the best to process the processing of neuronal networks higher than the theories generated by Chomskyan Linguistics. This field assumes that every one human languages ​​will be understood Grammatical systems (along with the context)And that these systems are related to universal grammar.

Chomsky carried out research On the syntactic theory as a paper and pencil theorist. He didn’t perform any experiments concerning the psychological or neuronal foundations of language understanding. His ideas in linguistics are absolutely silent concerning the mechanisms on which the sentence processing and understanding are based.

However, what the Chomskyan School of Linguistics does is asked how Human infants and toddlers can learn with such easy languageApart from neurobiological deficits or physical trauma.

There are a minimum of 7,000 languages ​​on the planetAnd no one can select where to be born. This signifies that the human brain should be ready to grasp and learn the language of its community at birth.

Chomsky made a (abstract) innate module for language learning from this fact about language development. From a neurobiological viewpoint, the brain should be ready to grasp the language from birth.

There are many examples of Language specialization in infantsThe exact neural mechanisms are still unknown, but not unknown. However, study objects are usually not recognizable if scientific terms are misused or used incorrectly. And that is strictly the danger: the compilation of AI with human understanding can result in dangerous consequences.