Originality of the categories of sense and absurdity concerning neural network modeling

Fundamentally, in this case, the presence/absence of a specifically absurd sense (not as a certain result of work, i.e., “at the output”, but “at the input” - as a determinant of the data received from the developer) is crucial. In such a situation, we speak ofthe presence of elements of traditional sense that discredit each other through refuting arrangements (a classic case of absurd sense). Such capabilities of the Sora neural network model [8] are significant for the development of several areas of deep and machine learning (for example, Google Assistant), machine translation (for example, DeepL), artificial intelligence systems for processing large corpora of texts (for example, natural language processing), etc. and innovative tools for linguistics (for example, computer and corpus linguistics) and some other sciences (for example, data science).

We should note that it is necessary to keep in mind the pragmatic nature of the existence of the language polysystem in general (in particular, the specifics of its actualization in the mind of the native speaker: within the linguistic and national worldviews). Thus, the above-mentioned understanding of the parameterization features of things in the event horizon of ontological reality is representative of neural network modeling (in particular, the work of the aforementioned Sora neural network model [8]). After all, the language poly system, first of all, acts as a means, a tool for achieving certain goals (recall the same contract social [22]), and not an abstracted, ideal phenomenon (we are talking about its realization aspect: speech, writing, and not about language as a construct [20]), which once and for all froze in a form at a certain point in time. In addition, it is advisable to take into account the specific distinction between content (images, senses, ideas) and apparatus (the mechanism of actualization of the latter). This is because the content of consciousness cannot exist without the aforementioned apparatus and outside of it, since it consists of the sign apparatus (we analyzed its specifics at the beginning of the article) and the sensory apparatus.

The sensory apparatus is biological in nature and is inherited by biological inheritance; it includes the brain, nervous system, and sensory organs (vision, hearing, etc.). For us, the inseparability of the sensory apparatus from the human body is important, as it is a part of the body. It is noteworthy that it is this apparatus that is capable of producing sensory images of reality, ranking them in the coordinates of sensations, perception, and representations, and storing the latter in itself (memory), reproducing them without the influence of stimuli. In addition, it is possible to combine sensory images to form new ones (imagination, fantasy). This is especially important in the context of working with the categories of meaning and absurdity in neural network modeling: to understand or at least outline an algorithm for working with textual data, it is advisable, if possible, to act by analogy with biological mechanisms. For example, let's remember that even structurally, artificial neural networks are an artificial equivalent of biological neural networks.

That is why it is significant that the core component of the human sensory apparatus is the language poly system (in particular, linguistic units actualized by the native speaker). Accordingly, since the signifier (linguistic signs), or rather their set, has an individualized nature, then, despite the presence of numerous communicative practices and templates, social contract [22], etc., the main thing is the sense produced by the individual. Thus, the latter (sense) is the pivotal determinant that determines the selection of linguistic signs actualized by the individual.

The above allows us to assert the productivity of neural network modeling in general and the use of specific neural network models. Thus, in terms of linguistic research, Sora [8], in our opinion, is representative in the context of studying the occurrence of the categories of sense and absurdity in online discourse (in particular, political online discourse). The latter is due to the ability of the aforementioned neural network model, as well as similar ones, to reproduce complex linguistic structures while analyzing semantic correlations between them. This, in turn, is productive for analyzing the distorted or distorted sense inherent in absurd statements, texts, etc [32].

Therefore, in the context of the actualization of the categories of sense and absurdity concerning neural network modeling, it is advisable to study the specifics of artificial neural networks. In particular, the peculiarities of their functioning and the consideration in this process of the context and semantics of the lexical units they analyze. At the same time, we consider it productive not to focus on specific linguistic tools in neural network modeling (in particular, recurrent neural networks, large language models, etc.). This is because the aforementioned Sora neural network model [8] has several common features with the work of other, seemingly unrelated, text data processing systems.

For example, similar features are found in the functionality of DALL E2 (in particular, DALL E3) [2] by the same OpenAI, which is capable of generating a picture from text data, or LeiaPix [4]. The functionality of the latter allows you to create a presentation from the text (i.e., again, despite their multimedia orientation, both have an understanding of the language poly system, although not as deep as Sora [8]). Another representative example of similar technologies for working with textual data, but much more imperfect than the above neural network model, is: Gemini (Google Bard) [3], which has similar functionality to ChatGPT 3.5 [1], and Narakeet [5], which is capable of converting text data into speech (indicating the presence of at least some understanding of text arrays, which suggests the “roots” of the latter text-to-speech technology, which is the basis of the “Balabolka”, “Govorilka”, etc. programs, the “@Voice” application, etc.).

Conclusions from this research and prospects for further research in this area

Thus, the above demonstrates the originality of the categories of sense and absurdity and the productivity of using the Sora neural network model [8] to work with them in studies of the corpus, computer, mathematical, etc. linguistics, philosophy of language, and several other areas. This is due, on the one hand, to the integration of the mathematical paradigm into the humanities, and, on the other hand, to the focus of linguistic science on the problems of understanding (lexicon, semantics, stylistics, etc.), generation (philosophy of language, linguopragmatics, metaphorology, etc.), and natural language processing, in which neural network modeling plays a prominent role. We are talking, first of all, about recurrent neural networks, large language models, etc., and, of course, the aforementioned Sora [8].

In our opinion, the use of the results of the latter's work in linguistic research is representative because of its ability to distinguish between the original categories of sense and absurdity, which are the cornerstones of neural network modeling. In turn, the actualization of such categories in the above process is productive in identifying relevant semantic patterns in the language poly system, which, in turn, is an important basis for machine and deep learning, machine translation, text data analysis, and generation, etc. It should also be noted that neural network modeling of linguistic units of online discourse (in particular, political discourse, which is the topic of our dissertation research) is representative of the development of interactive systems that can interact with users through language commands or queries.

For example, the aforementioned neural network model is an innovative tool of linguistic science that can study linguistic constructions (identify patterns, specifics of actualization, etc.) and features of discourse (political, Internet, etc.) in the context of the frequency of use of the categories of sense and absurdity. It is not only about understanding textual data (words, phrases, sentences, etc.) but also about decoding cultural metaphors, idioms, jargon, etc. actualized in the studied environment. Thus, mastering the originality of the categories of sense and absurdity is a direct indicator of the productivity and quality of the results of an artificial neural network. In addition, this indicator is key in the context of its understanding of these categories in different linguistic and cultural contexts (semantic bases, several possible actualizations, etc.).

At the same time, we note that a possible drawback of using artificial neural networks (including OpenAI products (ChatGPT 3.5 (ChatGPT-4) [1], DALL E2 (DALL E3) [2], Sora [8]), Google (Gemini) [3], etc.) as an innovative tool for linguistic science is the process of their training. Thus, the data used by the developer for this process are anthropocentric, i.e., they are created by people and for people, just like the developer and the teacher of the neural network model, who are people. In turn, this creates a tendency to inaccuracies, distortions, biases, etc. that the artificial neural network will “inherit” from them. The latter is especially true in the context of the russian-Ukrainian hybrid war, which is being circulated in the information space around which numerous russian narratives exist. Such narratives function as

part of misinformation, propaganda, and disinformation accusations against the Ukrainian government in general and the Ukrainian population in Ukraine and abroad.

Instead, counteracting and eliminating such inaccuracies in the context of linguistic science, machine, and deep learning is a promising area for further research, especially in the context of such distinctive categories as sense and absurdity. Thus, further research in this area will reduce the amount of such “information noise” (prejudice, misinformation, narratives) on both sides and improve the accuracy and objectivity of the analysis of the language poly system. In addition, in the context of linguistic science, it is advisable to talk about the importance of studying the role of cultural and social contexts (in particular, political online discourse) in understanding the categories of sense and absurdity in the language poly system. Neural network modeling (the aforementioned neural network models and their successors) is a modern and innovative tool for researching the above. The latter is due to their ability to process large data corpora with the localization of implicit/explicit relations in them, simplifying the conduct and deepening the significance of research by Ukrainian and other scientists.

References

1. ChatGPT. OpenAI : website. URL: https://goo.su/IHf5S (date of application: 29.02.24).

2. DALL E2. OpenAI : website. URL: https://goo.su/EdjJMG2 (date of application: 29.02.24).

3. Gemini. Gemini : website. URL: https://goo.su/pkW1vK (date of application: 29.02.24).

4. LeiaPix. LeiaPix : website. URL: https://goo.su/Ibj6bD (date of application: 29.02.24).

5. Narakeet. Narakeet : website. URL: https://goo.su/WzRB (date of application: 29.02.24).

6. Pika. Pika : website. URL: https://goo.su/eRrORM (date of application: 29.02.24).

7. Runway. Runway : website. URL: https://goo.su/xYz3RPl (date of application: 29.02.24).

8. Sora. OpenAI : website. URL: https://goo.su/3Rm1x (date of application: 29.02.24).

9. Уряд підтримав запуск Мрії. Дія : телеграм-канал. URL: https://goo.su/jh6ruhr (date of application: 29.02.24).

10. A graph neural network with context filtering and feature correction for conversational emotion recognition / C. Gan et al. Information Sciences. 2024. Volume 658. https://doi.org/10.1016/jJns.2023.120017 ScienceDirect : website. URL: https://goo.su/mrn8x (date of application: 29.02.24).

11. Ahmed U., Lin J. C. W., Srivastava G. Emotional Intelligence Attention Unsupervised Learning Using Lexicon Analysis for Irony-based Advertising. ACM Transactions on Asian and Low-Resource Language Information Processing. 2024. Volume 23, Issue 1. P. 1-19. https://doi.org/10.1145/3580496 ACM Digital Library : website. URL: https://goo.su/y9CU (date of application: 29.02.24).

12. Artefact. Cambridge Dictionary : website. URL: https://goo.su/OB7pBO2 (date of application: 29.02.24).

13. Bajohr H. On Artificial and Post-Artificial Texts : Machine Learning and the Reader's Expectations of Literary and Non-Literary Writing. HANNES BAJOHR : website. 2024. URL: https://goo.su/v4ALndG (date of application: 29.02.24).

14. Bhaskar Y OpenAI's Sora : This Text to Video Model is Absurd. Medium : website. URL: https://goo.su/ bnBUvk (date of application: 29.02.24).

15. Bilokon P. A. Evolution-Bootstrapped Simulation: Artificial or Human Intelligence: Which Came First? arXiv : website. 2024. URL: https://goo.su/SfuKUUF (date of application: 29.02.24).

16. Bruner E. Cognitive archaeology, and the psychological assessment of extinct minds. Journal of Comparative Neurology. 2024. Volume 532. https://doi.org/10.1002/cne.25583 WILEY: Online Library : website. URL: https://goo. su/FalPM9 (date of application: 29.02.24).

17. Buscemi A., Proverbio D. ChatGPT vs Gemini vs LLaMA on Multilingual Sentiment Analysis. arXiv : website. 2024. URL: https://goo.su/DpnjA (date of application: 29.02.24).

18. Carchidi V. J. Book Chapter for the forthcoming Artificial Knowledge of Language : A Linguist's Perspective on its Nature, Origins and Use / ed. by J.-L. Mendlvil-Giro. ResearchGate : website. 2024. URL: https://goo.su/fijYY (date of application: 29.02.24).

19. Choe Y. Meaning versus information, prediction versus memory, and question versus answer. Artificial Intelligence in the Age of Neural Networks and Brain Computing. Academic Press, 2024. P. 61-76. https://doi. org/10.1016/B978-0-323-96104-2.00003-8 ScienceDirect : website. URL: https://goo.su/zVx5JM (date of application: 29.02.24).

20. Construct. Cambridge Dictionary : website. URL: https://goo.su/xSa1XUx (date of application: 29.02.24).

21. Content. Cambridge Dictionary : website. URL: https://goo.su/UHUKqC (date of application: 29.02.24).

22. Contract social. Cambridge Dictionary : website. URL: https://goo.su/AzsxBH (date of application: 29.02.24).

23. Deschenaux I., Matthews W. Homo anthropologicus: unexamined behavioural models in sociocultural anthropology. Anthropological Theory. 2024. LSE: Research Online : website. URL: https://goo.su/vy7mSL1 (date of application: 29.02.24).

24. Diversity and language technology: how language modeling bias causes epistemic injustice / P Helm et al. Ethics and Information Technology. 2024. Volume 26. Springer Link : website. URL: https://goo.su/4GMRyH (date of application: 29.02.24).

25. Doroudi S. On the paradigms of learning analytics : Machine learning meets epistemology. Computers and Education: Artificial Intelligence. 2024. Volume 6. https://doi.org/10.1016Zj.caeai.2023.100192 (date of application: 29.02.24).

26. Gozalo-Brizuela R., Garrido-Merchan E. C. ChatGPT is not all you need. A State of the Art Review of large Generative AI models. arXiv : website. 2023. URL: https://goo.su/kcV771s (date of application: 29.02.24).

27. Interpreting black-box models : a review on explainable artificial intelligence / V. Hassija et al. Cognitive Computation. 2024. Volume 16. P. 45-74. Springer Link : website. URL: https://goo.su/lJyBX (date of application: 29.02.24).

28. Jaeger L., Dacorogna M. Artificial Intelligence from Its Origins Via Today to the Future : Significant Progress in Understanding, Replicating, and Changing Us Humans or Solely Technological Advances Contained to Optimising Certain Processes? Where Is Science Leading Us? And What Can We Do to Steer It? Cham : Springer Nature Switzerland, 2024. P 207-235. https://doi.org/10.1007/978-3-031-47138-4_10 Springer Link : website. URL: https://goo.su/Pq3QKB (date of application: 29.02.24).

29. KERMIT: Knowledge-EmpoweRed model in harmful meme deTection / B. Grasso et al. Information Fusion. 2024. Volume 106. https://doi.org/10.1016/jJnflus.2024.102269 ScienceDirect : website. URL: https://goo.su/bz1Arz (date of application: 29.02.24).

30. Kondurkar I., Raj A., Lakshmi D. Modern Applications With a Focus on Training ChatGPT and GPT Models : Exploring Generative AI and NLP. Advanced Applications of Generative AI and Natural Language Processing Models. IGI Global, 2024. P. 186-227. IGI Global : website. URL: https://goo.su/tbZ2Uxg (date of application: 29.02.24).

31. Kozikowski D., Zema T., Sulich A. Artificial Intelligence Usage and Ethics in the Choice Theory. 2024. P. 7438-7449. Springer Link : website. URL: https://goo.su/m6u3UEs (date of application: 29.02.24).

32. Kumar R. Open AI SORA : A Game-Changing Text-to-Video Model, How to Access. LinkedIn : website. URL: https://goo.su/QcwpzU (date of application: 29.02.24).

33. Meaning. Cambridge Dictionary : website. URL: https://goo.su/eQE4 (date of application: 29.02.24).

34. Model-Theoretic Logic for Mathematical Theory of Semantic Information and Communication / A. F. Saz et al. arXiv : website. 2024. URL: https://goo.su/MAi7 (date of application: 29.02.24).

35. Sense. Cambridge Dictionary : website. URL: https://goo.su/SUIYVr (date of application: 29.02.24).

36. Sora AI Video. X : website. URL: https://goo.su/XVpZTR (date of application: 29.02.24).

37. Trustllm : Trustworthiness in large language models / L. Sun et al. arXiv : website. 2024. URL: https://goo.su/ a1jUg (date of application: 29.02.24).

Размещено на Allbest.ru