For starters, it’s extremely flexible, and it’s relatively easy to train deep-learning systems (which we call models) to take on a wide variety of tasks. Deep learning has several key advantages for creative pursuits. These experiments in computational creativity are enabled by the dramatic advances in deep learning over the past decade. In the realm of language and literature, a text-generating system called GPT-2 from the research lab OpenAI proved able to generate fairly coherent paragraphs of text based on a starter sentence. AI-generated paintings have been auctioned off at Christie’s, the DeepBach program has composed convincing music in the style of Bach, and there has been work in other media such as sculpture and choreography. Our efforts fall within the booming research field of computational creativity. Poetic forms such as sonnets have fairly rigid patterns when it comes to rhyme and rhythm, and we wondered if we could design the system’s architecture so that Deep-speare would learn these patterns autonomously. Our goal was to see how far we could push deep learning for natural-language generation, and to make use of the interesting qualities of poetry.
Instead, Deep-speare independently learned three sets of rules that pertain to sonnet writing: rhythm, rhyme scheme, and the fundamentals of natural language (which words go together). We didn’t give it rhyming dictionaries, pronunciation dictionaries, or other resources, as has often been the case in previous computer-generated poetry projects. Our “poet” learned how to compose poetry on its own, using the AI approach known as deep learning-it cranked through the poems in its training database, trying again and again to create lines of poetry that matched the examples. Our team, composed of three machine-learning researchers and one scholar of literature, trained our AI poet using about 2,700 sonnets taken from the online library Project Gutenberg. As our research team discovered when we showed our AI’s poetry to the world, that’s enough to fool quite a lot of people most readers couldn’t distinguish the AI-generated poetry from human-written works. Here’s a stanza from a sonnet written by William Shakespeare:Īnd here’s one written by Deep-speare, an artificial intelligence program that we trained to write sonnets:ĭeep-speare’s creation is nonsensical when you read it closely, but it certainly “scans well,” as an English teacher would say-its rhythm, rhyme scheme, and the basic grammar of its individual lines all seem fine at first glance.
0 kommentar(er)
