I asked AI to analyse Armenian-Urartian language similarities

OpenAI’s new chatbot called GPT-3 is the largest language model AI at the moment. It has been trained with over 175 billion parameters, including books, web texts, academic papers, and encyclopedias. So I decided to ask it some interesting questions regarding the Armenian language.

We know from recent discoveries that Urartians are genetically identical to modern Armenians. Most modern Armenians would score extremely close to Urartian DNA samples from Van, recently released by Turkish academics. However, most resources available to common internet users will tell you that Urartians were a distinct people with only an obscure relation to Armenians. Despite the countless evidence to the contrary (as I have outlined in my five-part articles on “Who were the Urartians“), it seems linguists in particular find it difficult to relate Armenians with Urartians. This awkward attitude is solely based on one major assumption: “The Urartian language belongs to the Hurro-Urartian language family, while the Armenian language belongs to Indo-European, therefore the Urartians could not have been Armenians.”

I always considered this a weak argument, as I have discussed in my articles. There is clearly a strong relation between the Urartian and Armenian languages, not just loanwords but grammar too. I have also published a thread on Twitter with nine arguments against such an assumption.

Nevertheless, the Armenian-Urartian relation has not been widely recognized. To date, only S. Ayvazyan has published an extensive comparative analysis (Urartian-Armenian lexicon and comparative – historical grammar, 2011), in which he concludes that the Urartian language is most likely a form of proto-Armenian. This conclusion is not always accepted by academics, but a better comparative analysis has not yet been produced.

Therefore, I have decided to ask the GPT-3 AI these somewhat complicate questions and see what conclusions it produces. Its answers are absolutely amazing and could actually provide evidence for a better understanding of both the Armenian and Urartian languages. Let’s first ask the AI who the ancestors of Armenians were and see what it says.

Here we can see that it correctly identified the people of the kingdom of Urartu as the ancestors of Armenians. Now let’s ask it more complicated questions.

This is an interesting reply, as I have (completely independently) also identified several grammatical similarities including the nouns “this and that”. I’ve discussed them here: https://www.peopleofar.com/2022/04/09/who-were-the-urartians-part-4/

Let’s ask for more information on this topic.

Now this is an interesting explanation. Especially considering that most resources would claim that the Armenian and Urartian languages are unrelated because they supposedly belong to different language families. So let’s confront AI with this question.

Apparently a language can be part of two or more language families and Armenian might be such a hybrid. But perhaps the Hurro-Urartian language itself is related to Indo-European. Let’s ask further.

I am amazed by the above replies. I have long suspected that Hurro-Urartian has been wrongly classified, because it has been poorly researched. Arnaud Fournet and Allan R. Bomhard for example have published several works on the subject, among which an extensive book titled: The Indo-European Elements in Hurrian (2010), in which they argue for a genetic relationship between Hurrian and Proto-Indo-European. I’ve previously discussed this subject in part 4 of my five part analyses on “Who were the Urartians”.

In the course of this book, we have attempted to show, through a careful analysis of the relevant phonological, morphological, and lexical data, that Urarto-Hurrian and Indo-European are, in fact, genetically related at a very deep level, as we indicated at the beginning of this chapter…

Arnaud Fournet and Allan R. Bomhard (2010), The Indo-European Elements In Hurrian.

I have a crazy theory. Personally I believe that Hurrian is a form of proto-Armenian related to Indo-European. One thing is for sure though, Hurrian languages have always been attested with IE people. Be it with the Hittites, Mitanni or Armenia. Somehow that language family is related more so to IE speakers than with anyone else. In fact if you ask me I would probably say that Hurrian is indeed the missing link between the Anatolian branch of IE and the Steppe branch as Reich and Lazaridis hypothesised. In their latest papers about “Southern Arc” they claim that the origin of Indo-European came from Armenian Highlands and there must have been a (missing link) language between Anatolian branch and Steppe branch (they hypothetically named Porto-Indo-Anatolian).

This is a novel theory and not much is known about it, but I would nominate Hurrian. Khorenstsi actually mentioned a descendant of Hayk named “Hor” who have founded the Horrian dynasty of Armenia. I believe he is recalling the Hurri tribe one of the Armenian tribes. Mind you the Akkadian cuneiform for the sound “u” is the same the sound “o”.

It is clear that there is a complex relation between Urartian and Armenian. Armenian language might have emerged from such complex interaction of Hurrian with proto-Indo European.

Although tools such as GPT-3 are designed to process and analyse complex data including scientific materials, at this point, we should take caution in the way we interpret these replies. For example, asking the same question twice often produces completely different conclusions. Nevertheless, I think such tools could help us understand such complex linguistic problems in the future or at least point towards (less known) research. For example Meta’s Galactica AI is specifically designed for academic purpose. Unfortunately it was taken down due to various problems and is to date not operational. In the future such models could provide valuable information on rarely researched or poorly understood fields. I would invite linguists who are interested in this topic to keep an eye on such tools.