The article presents a new method for obtaining analogues of words, characterized by simplicity and the absence of the need for preliminary training on large data as in existing methods. In the method under study, analogues are determined by their syntactic predicates using methods of distributive semantics. In the study, analogues of adjectives, nouns and verbs were obtained and analyzed. This made it possible to obtain a result that is not inferior to the results obtained using the most popular neural (...) network approach as word2vec when qualitatively comparing analogues. The demonstrated method shows that obtaining analogues is possible using methods of distributive semantics using a more interpretable method, which opens up the possibility of studying semantic analogy. This method also allows you to identify analogues on a specific topic. Based on the experimental results obtained, an original definition of analogues and cognitive schemes is formulated. The article also analyzes and substantiates the possibility of a new approach for creating artificial intelligence systems based on the researched method. According to the authors, this provides significant advantages for the creation of such systems. In particular, the proposed method allows for broader generalizations over orders of magnitude smaller data, as well as learning during use, which is not possible for neural networks. (shrink)

We present an analytical study that puts forward a hypothesis about the role of analogy in thinking. The article presents a critique of the structural mapping hypothesis in analogy as an explanation of the analogical argument and will offer an alternative explanation — the functional hypothesis of the analogical argument. It will also be shown how such functions of thinking as recognition and memory, metaphor and syllogism, generalization and ontology, deduction and induction can be realized with the help of the (...) analogy method developed by the author. The article puts forward a hypothesis that these different functions are nothing more than different manifestations of one mechanism for the dissemination of thinking patterns through analogy.In the previous article [1] we described the results of experiments on obtaining analogs of words based on the statistics of predicates extracted from texts by a simple parser. In this article, continuing the topic of analogy, we will analyze two aspects:- how similar expressions (analogous argument) are found when solving problems and learning;- what is the role of analogy in thinking, namely, what thinking processes are realized with the help of analogy.The article critically examines the most well-known approaches to explaining and attempting to implement analogy. The authors propose another hypothesis, called functional analogy, which is distinguished by the simplicity of explaining the analogical argument, greater correspondence to examples of analogy, in the authors’ opinion, and the possibility of programmatic implementation. (shrink)

В статье представлен новый способ получения аналогов слов, отличающийся простотой и отсутствием необходимости предварительного обучения на больших данных как в существующих методах. В исследуемом методе аналоги определяются по их синтаксическим предикатам с применением методов дистрибутивной семантики. В исследовании были получены и проанализированы аналоги прилагательных, существительных и глаголов. Это позволило получить результат, который не уступает результатам, полученным с помощью наиболее популярного нейросетевого подхода как word2vec при качественном сравнении аналогов. Демонстрируемый метод показывает, что получение аналогов возможно на методах дистрибутивной семантики с применением (...) более интерпретируемого метода, что открывает возможность исследования семантической аналогии. Данный метод позволяет также определять аналоги по определенной тематике. На основе полученных экспериментальных результатов формулируется оригинальное определение аналогов и когнитивных схем. Также в статье анализируется и обосновывается возможность нового подхода для создания систем искусственного интеллекта на основе исследованного метода. По мнению авторов это дает значительные преимущества для создания таких систем. В частности, предлагаемый метод позволяет создавать более широкие обобщения на порядки меньших данных, а также обучение во время использования, что недоступно для нейросетей. (shrink)

Large language models in the form of chatbots very realistically imitate a dialogue as an omniscient interlocutor and therefore have become widespread. But even Google in its Gemini chatbot does not recommend trusting what the chatbot will write and asks to check its answers. In this review, various types of LLM errors such as the curse of inversion, number processing, etc. will be analyzed to identify their causes. Such an analysis led to the conclusion about the common causes of all (...) errors, which is that transformers do not have deep analogy, hierarchy of schemes and selectivity of content taken into account in the inference. But the most important conclusion is that transformers, like other neural networks, are built on the concept of processing the input signal, which creates a strong dependence on superficial noise and irrelevant information that the transformer’s attention layer cannot compensate for. The concept of neural networks was laid down in the 1950s by the idea of F. Rosenblatt’s perceptron and did not take into account the achievements of cognitive psychology that appeared later. According to the constructivist paradigm, the input word (or perception) is only a way to check the correctness of the constructed predictive model for possible situations. This is the cause of the biggest problem of transformers, called hallucinations. And its elimination is possible only by changing the architecture of the neural network, but not by increasing the amount of data in training. (shrink)