Publication: Yapay zekanın müzikal Yaratıcılığı : chatGPT örneği
Abstract
1940’lı yıllardan günümüze kadar olan süreçte yapay zeka (YZ) oldukça önemli bir gelişim göstermiş; pek çok sektörde kullanım alanı bularak hayatın önemli bir parçası haline gelmiştir. Bu sektörlerden birisi de müziktir. Nitekim yapay zeka ile müzik bestelemeye dair günümüzde çeşitli uygulamalar mevcuttur. Ancak yapay zeka ile Türk makam müziği bestelemeye dair yapılan çalışmalar yok denecek kadar azdır. Bu çalışmada yapay zekanın rast makamında, sofyan usulünde ve Türk müziği formlarından ilahi formunda bir beste üretmedeki yetenekleri ChatGPT örnekleminde ölçülmüştür. Besteleme sürecinde, yapay öğrenme veya makine öğrenmesi olarak ifade edilen teknik ile ChatGPT’nin ilahi formu, rast makamı ve sofyan usulüne dair bilgisi desteklenmiştir. Metin halindeki nota yazım sistemi kullanılarak üç ilahi örneği sohbet botuna aktarılmıştır. 69 prompttan oluşan süreç neticesinde ortaya çıkan rast makamı ve sofyan usulündeki beste karşılaştırmalı olarak analiz edilmiştir. Analiz; bestelenen ilahinin rast makamı kuralları, ilahi formunun vezin yapısı ile biçim şeması ve tarihsel süreç içerisinde bestelenen ilahilere benzerliği üzerinden gerçekleştirilmiştir. Benzerlik oranı; seyre başlangıç perdeleri, seyirdeki en tiz ve en pest perdeler, seyirde kullanılan çeşniler, karar öncesi yeden kullanımı ve çeşni sayıları gibi hususlar üzerinden tespit edilmeye çalışılmıştır. Sonuç olarak ChatGPT’nin bestelediği ilahinin Türk din musikisi repertuvarında yer alan rast makamı ve sofyan usulündeki ilahileri temsil ettiği düşünülebilecek bir örneklem ile karşılaştırıldığında; makam, ilahi formunun vezin yapısı ve biçim şeması bakımından örtüşen niteliklere sahip olduğu söylenebilir.
Artificial intelligence (AI) has undergone remarkable advancements since the 1940s, permeating various sectors and becoming an integral part of life. Music is one such domain where AI has made significant inroads. As a matter of fact, there are various AI-powered music composition tools today. However, studies on composing Turkish maqām music with AI are almost non-existent. This study investigates the ability of AI to generate a composition in rast maqām, rhythmic cycle of sofyān, and ilāhī form of Turkish music, using ChatGPT as an example. In the composition process, ChatGPT's knowledge of ilāhī form, rast maqām, and rhythmic cycle of sofyān was augmented through machine learning techniques. Three ilāhī samples were provided to the chatbot using a text-based notation system developed by the author. As a result of the process consisting of 69 prompts, the composition in rast maqām and rhythmic cycle of sofyān was analyzed comparatively. The analysis is based on the rules of rast maqām, the meter structure and the composition scheme of ilāhī form, and its similarity to the ilāhīs composed in the historical process. The rate of similarity was determined based on aspects such as the starting pitches, the highest and lowest pitches in the melodic progression, the flavors used in the progression, the use of the sansibl (yadan) before the finalis and the number of flavors. In conclusion, when compared with a sample from the Turkish religious music repertoire that can be considered to represent the ilāhīs in the rast maqām and the rhythmic cycle of sofyān, the composition created by ChatGPT was found to possess characteristics that align with the maqām, the meter structure and the composition scheme of the ilāhī form.
Artificial intelligence (AI) has undergone remarkable advancements since the 1940s, permeating various sectors and becoming an integral part of life. Music is one such domain where AI has made significant inroads. As a matter of fact, there are various AI-powered music composition tools today. However, studies on composing Turkish maqām music with AI are almost non-existent. This study investigates the ability of AI to generate a composition in rast maqām, rhythmic cycle of sofyān, and ilāhī form of Turkish music, using ChatGPT as an example. In the composition process, ChatGPT's knowledge of ilāhī form, rast maqām, and rhythmic cycle of sofyān was augmented through machine learning techniques. Three ilāhī samples were provided to the chatbot using a text-based notation system developed by the author. As a result of the process consisting of 69 prompts, the composition in rast maqām and rhythmic cycle of sofyān was analyzed comparatively. The analysis is based on the rules of rast maqām, the meter structure and the composition scheme of ilāhī form, and its similarity to the ilāhīs composed in the historical process. The rate of similarity was determined based on aspects such as the starting pitches, the highest and lowest pitches in the melodic progression, the flavors used in the progression, the use of the sansibl (yadan) before the finalis and the number of flavors. In conclusion, when compared with a sample from the Turkish religious music repertoire that can be considered to represent the ilāhīs in the rast maqām and the rhythmic cycle of sofyān, the composition created by ChatGPT was found to possess characteristics that align with the maqām, the meter structure and the composition scheme of the ilāhī form.