Idea generator

Words are prime tools in the ideation process. But with the advancement of artificial intelligence, AI, the distinction between natural language, as spoken and written by humans, and formal or built language, as used in computing, is narrowing. For example, recent machine learning research has revealed a new text generator, the GPT2, which is capable of performing goal-oriented tasks including translation and summarisation, and pass simple reading comprehension tests giving the model more knowledge of how to understand written text. Moreover, the unsupervised transformer-based language model, created by nonprofit company OpenAI, and trained on a huge dataset, can be used to generate infinite positive – or negative – reviews of products. That is, the model is capable of generating samples from a variety of prompts (human written) that feel close to human quality and show text coherence. This, in turn, may suggest that if the model were to be trained on a dataset of designs, it is conceivable that the text generator, given sufficient and contextual relevant data, could turn idea generator. This raises furter questions, such as, what is the most authentic, or the highest valued idea: human generated ideas - the human connection or ideas generated by the algorithms in machines? That is, questions about the impact of AI system design on ethics, trust and humanity. See also: https://blog.openai.com/better-language-models/ ; https://www.cam.ac.uk/research/news/