So a few days back, I sat down to explore if I could create krithis in carnatic music. In Tamil and Telugu and Sanskrit. Just for fun. I was bored and there was nothing interesting on any of the OTT channels.
My prompts were for a Tamil composition( with a pallavi, anupallavi and charanam) in praise of Lord Muruga. Additional prompts were provided in the iterations to refer to his valour, his compassion and his consorts Valli and Deivanayaki and use the Ramadasan mudra( attempt to create in the style of Papanasam Sivan). I stopped after a couple of revisions . But the output is given below.
This is truly horrific at so many levels. The first thing that bothers me is the superficial mimicking of something truly profound. I am not getting into the content of the krithi at this point which is garbage since one could improve that with some more tweaks. But the basic source code. All of our krithis have been the outpourings from a deep wellspring of devotion . All of the alliteration, the prasa , the references to events in mythology happened on their own without any particular planning or manipulation. Most practitioners would believe that singing Dikshithar's navavarnas is akin to doing Srividhya puja of the Devi. All of them are deeply emotional . They evolved out of an altered state and a singer can still experience that magic when listening or singing some of these.
I stopped after maybe a few revisions because I could feel my skin crawling. You know those horror movies where a mad scientist creates this perfect looking baby in a laboratory with an advanced IQ. But totally devoid of everything that makes it human: empathy, compassion, simple kindness. And that very lack goes on to unleashing a truly evil being on the world. Something like that.
But there is more. We have today the ability to generate videos both from text and also for alteration. We can create audios or videos of a fake composition sung in the style of a music legend by anybody. In fact many years back Kavita Krishnamoorthy has spoken about how all that is needed for playback singers today is attitude. Both pitch alignment and tempo are being corrected in the studio.
So there will be a time when there will be a fake AI generated composition in a AI generated video in the style of some old time master previewing on Youtube and garnering hundreds of likes and comments😢😡😠😟
