Using AI for generating new music
------------------------------------------
Hi guys, I am here again with another interesting topic which you will definitely like. From the title you all would have understood by now that this post is about discussing the advancement in AI where an AI generates music with high accuracy for given genre labels such as Hip-hop music, Jazz, Rock, etc.
As you all know that Music is an integral part of human culture, which exists from the time unknown of human civilization. It evolved with time into a wide diversity of forms. It evokes a unique human spirit in its creation, and the question of whether computers can ever capture this creative process has fascinated computer scientists for decades.
------------------------------------------
Attempts in the past:
Previously there were many attempts to generate music automatically but the were not that accurate and they lacked the adjustment of lyrics(Lyrics Conditioning) with the genre types. Although scientists had many different algorithms to generate various instrumental sounds(like generating piano
sheet music, etc.) but they were by produced by different systems and the scientists wanted a single system to reproduce all the genres of music.
------------------------------------------
Surge in AI with many innovative models like generative models:
So finding a solution for the above mentioned problem, where Novelty was a key issue, was very hard to find until came the boom of AI in the market which was majorly due to the reduction of cost of computing devices and also the reduction in Computation power with the reduction of resource consumption by the Neural Networks for their training. This acted as a motivation for the interested ones(mainly scientists), to again try to start to generate new music automatically using the AI.
So with the boom of AI, the field of generative models also made tremendous
progress in the last few years. One of the very primary aims of generative modeling is to capture the salient aspects of the any data
and to generate new instances of it, indistinguishable from the
true data.The hypothesis is that by learning to produce the
data we can learn the best features of the data.
So, scientists started to use generative models to train the neural nets to produce different music sounds according to the genre. While these conditioned models were able to generate
songs of diverse genres and artistic styles, the singing voices
generated by those models were often sung in a compelling
melody by the AI and were mostly composed of babbling(unrecognized English words).
------------------------------------------
Recent Advancement:
In April 2020, a group of scientists led by Prafulla Dhariwal (research source link) at OpenAI found a method to control
the generative model with lyrics by providing more context
at training time and by conditioning the model on the lyrics
corresponding to each audio segment which allowed the model
to produce singing simultaneously with the music.
If you want to go deep into their work you can check the link above.
------------------------------------------
Conclusion:
Hence, I believe that the AI music generation should only serve as a tool for human musicians and those interested in music to learn and compose different melody and genres and it should not hamper their creativity, as in future many people may only be dependent on the melody generated by the AI if the creativity of people degraded drastically.
Also, I think that in the near future most of us can generate our own music without any formal training with the help of the AI to help us relax. But we should remember to not to forget human musicians; as no matter how much the advancement in AI takes place but an AI can't replace humans as humans has a precious gift of consciousness which the AI don't have yet.
-----------------------------------------------------------------------------------------------------------------------------
So guys, Thanks for reading this post and what do you think can this technique to generate music be used for in the real life.
Please let me know your thoughts in the comment box.
