Transcript – Deep Learning – Exploring the Limits Beyond Low Hanging Fruit

Hi, this is Phillip Galanter from the Department of Visualization. In the realm of computer science and artificial intelligence, deep learning networks are current topic of intense research. In this project, we explore the use of deep learning neural networks in three very different disciplines computer science, visual art and music. With the explosive growth in the realm of artificial intelligence has come a body of ethical concerns regarding human impact. However, my work explores our actions towards A.I. systems. The concept of machine patency is the notion that humans may have moral obligations towards AI systems as they become sentient. Five bodies of knowledge are inspected in order to set the landscape for future machine patients research. These are the history of human encounters with sentient others, topics from philosophy of mind, topics from moral philosophy, mission specialists who study AI and ethics, and the nascent field of complexes ism. In the talk presented at the College Art Association with focus on the question as to when computers should be credited as true authors. The noted Excavates paper adds further developments during the conversation on the general problem of machine patency. That paper closes with a provisional affirmation of machine patency is plausible on the basis of both natural charity and rational non contradiction. And now here’s Nima Kalantari commentary with a few words about his work in computer science.

The proposed planning approaches for various industries applications, specifically the present and planning approach to plausibly reconstruct the missing information in an input image due to sensor saturation. We also propose an approach to reconstruct novel views from a single image which can then be used to generate different focus images. Finally, we propose a deep learning method to reconstruct a high resolution, high frame rate video using drug cameras of the regular cell phones.

And now here’s Jeff Morris with a few words about his work in music.

My first and ongoing approach to native composition for machine learning systems was to give a voice to its errors and its uncertainty. For this, I needed a crib sheet structure in which a classifier system could report its best conclusions about an input. But an absolute correct answer could also be known at the same time. So I can compare the two in real time and let any differences drive the music. The uncertainty or incorrect confidence of the system will become readily apparent and whether the audible errors wobble erratically or lock in on a constant tone.