The Future of Art with Machine Learning
Artificial Intelligence(AI)/Machine Learning(ML) has already revolutionized various industries like Manufacturing, Information Technology, Finance, Healthcare, and more in the past few decades.
After witnessing the reinforcement learning techniques of AI, AI seems to be much more than a technology to streamline processes. It is promising to facilitate creativity in areas like art, music as well.
Yes, AI is already making it big in the world of arts and culture. The power of AI has opened up a whole new world of creativity and has produced many digital artists. These artists leverage computer algorithms to bring life into their art creations.
So, what does that even mean? Will these AI artists take over humans to create splendid artworks, enthralling music, inspiring poetry, and rational movie scripts? If so, what will be the future of human creativity in art?
Would AI take over the art field entirely, or will it only act as a tool to augment human creativity? Like in any other industry, technology advancements will revolutionize the art industry too. Be it in art galleries, music production studios, or poetry. AI will make an impact undoubtedly.
Now, let’s move on in understanding the possible role of AI/ML in various art fields:
1) Video Editing can be a Bliss with ML
Runway, a Newyork based startup, has successfully integrated Machine Learning(ML) in the art and creative field, building next-gen video editing tools. Hands down, the company’s real-time video editing tool Green Screen has revolutionized video editing for real.
Right now, rotoscoping, the process of cutting objects from videos, is highly time-consuming. The software allows you to overcome it by creating synthetic content. That is using AI algorithms to generate, modify and edit audiovisual content automatically.
Before proceeding further, the Runway team posted a tweet asking for reviews regarding the potential AI video editing tool. They got an overwhelming response from various tech giants like Google and Facebook within 48 hours of posting.
With the strong support, the company went ahead and now assures that AI algorithms can facilitate anyone to create hyper-realistic animations in a few seconds and edit them automatically. That means, with AI, you can create special effects that you have only seen in Hollywood movies so far.
2) Transform Any Photo Portraits into Masterpieces with ML
ML algorithms are booming as a remarkable tool in painting as well. A Japanese-origin AI artist app transforms the photo portraits of users into masterpieces. This particular AI art generator/AI app is more like entertaining people.
Many people have downloaded this app and tried transforming their photographs into beautiful masterpieces. However, the app has received some controversial comments, like some people who used the app have received a light-skinned painting of their uploaded photo.
However, by using a wide variety of learning data, one can increase the diversity of the output in the future, assured the Japanese app developer. Nevertheless, AI seems to be highly influential in the art field.
3) Music Compositions is made easy with ML
Music composition has never been this easier with AI. OpenAI has introduced its new machine learning model named Jukebox. It allows you to create new music samples right from scratch.
The AI model is remarkable with identifiable melodies and words, though the results are similar to familiar songs. The model works based on raw audio rather than symbolic music that does not contain voices.
The team then uses neural networks to encode and compress raw audio. With the help of a transformer, they generated new, compressed audio and upsampled it to turn it back into raw audio.
Moreover, Jukebox generates its own lyrics as well. However, OpenAI researchers also contribute their fair share of verse. Though these AI models come with spectacular music, there is a substantial gap between these AI systems and human-created music.
For instance, AI-generated songs seem to have musical coherence, follow chord patterns and are good at featuring captivating solos. But, AI is not assuring larger musical structures like choruses that repeat.
4) Recreate Existing Paintings with ML
With recent advancements in the AI field, researchers have started developing algorithms that can recreate art pieces. A Ph.D. student at MIT, Amy Zhao came up with a paper, Painting Many Pasts: Synthesising Time Lapse videos of Paintings, which illustrates how an existing painting might have been painted originally in detail.
Ideally, an image is a mere list of random numbers for a computer. You would need a Convolutional Neural Network (CNN) to make sense of all these numbers. The team trained an algorithm by feeding it with time-lapse videos of artists creating paintings. That way, the system too learned to figure out the steps involved to recreate an existing painting.
Amy Zhao’s team has got their inspiration from artistic style transfer. That transfer involves using neural networks to create art mash-ups of various artists’ work. Or even to make art in the style of a specific artist.
Almost 158 people participated in evaluating the time-lapse videos. The participants compared the time craft videos to the original time-lapse videos during the event. Though most participants preferred the actual videos, around 50% of the time, they got confused with the time craft videos for the real ones.
That indicates that we can use Machine Learning (ML) to discover the individual steps involved in the art piece.
5) Write Poetry in the Style of Famous Poets with ML
Do you want to write poetry but hesitate because you think you don’t have the right skills? Then the Verse by Verse tool from Google comes in handy to guide you. The renowned poetry tool can help you with poetry suggestions in the style of 22 reputed American poets.
You can start by giving some of your favorite poet’s names and select a poem structure. You can choose the rhyme of the poem, the syllable count, and the poetic form.
Thereby, the Verse by Verse tool will request you to compose your first line. Then the AI provides several suggestions on how to enhance or proceed further. However, you can accept, dismiss or alter its recommendations.
The suggestions from the tool are not from the original poetry but are novel verses that resemble lines from renowned poets. Google further explains that their engineers fed and fine-tuned their AI with an extensive selection of classic poetry and writing styles.
Google’s experts trained the system to have a general semantic understanding of what lines suit the best to follow the previous line of verse. Hence, even if you write on topics not common in classic poetry, the system will give you the best possible suggestions and clarify Google.
6) AI Singer can Sing Songs in Multiple Languages with ML
Popgun, an Australian AI-music startup that composes and plays music, has now trained AI to sing as well. With text and midi as input, the startup has generated vocal tracks in various voices.
The CEO of Popgun says that they have made progress but still need to go a long way to reach prime quality. He reassures that AI will play the guitar, the bass, and the piano. And each one of those AIs can listen to each other and play together. Moreover, they can sing as well.
Recently, the Chinese have also come up with an AI model, the DeepSinger system. A research team from Zhejiang University and Microsoft has generated the voice of an AI singer using algorithms that can predict and control the pitch and audio duration.
Moreover, the sound people produce when singing has more complex rhythms and patterns than the sound they make while speaking. Thus, the researchers came with a data pipeline that mined and transformed audio data. Singing clips were extracted from various websites, and the singing itself was isolated from the rest of the audio and divided into sentences. The technique is related to the text-to-speech method that enables machines to speak.
When human artists need to visit the studio to fix mistakes or changes after a recording session, AI-assisted voice synthesis will prevent the need deliberately, saving time and money. The AI technology can create deepfake, enabling it as though specific artists sang lyrics they never did. However, the possible future problem with this AI technology is already here with Jay-Z’s deepfake audio. The rapper appeared to rap a famous song, which in reality, he had never done.
Wrapping Up
Machine Learning (ML)/ Artificial Intelligence (AI) is one of the prominent technologies that can revolutionize the creative field also as they did in every other industry. Yes, be it in music, fine arts, poetry, or architecture. The machines can support humans in enhancing creativity and take it to the next level of excellence.
Art is one such medium that enables humans to express their thoughts and emotions with the world. Artificial Intelligence enables scientists to explore and understand the operating system of the human brain. Though they can’t replicate human actions entirely, they still can augment human creative work and take a step ahead.
ML is still in the early stages in the art field. Time will decide whether AI will remain a tool or will it become a creative entity itself? When we humans use technology in every other industry to enhance our lives or increase our productivity, why not art? We can leverage AI in the creative field also to push our artistic boundaries and create extraordinary artworks.
