Prof McCluskey and AI training

-
Ociacia, V. (2016) Artificial Intelligence
Went to the computer science dept at Hud for a brief conversation with Lee McCluskey about AI.
Basically (very basically) there seems to be 2 ways to "train" an AI.

Firstly, Data Driven - you give it vast quantities of Data (music scores, pictures,text etc) and it then extrapolates the next sequence of probable "thing" based on the information you've given it (a little like the IQ tests we did at school) - these are beginning to be available as in-browser softwear chat apps etc - I can probably find one to give pictures to... but it's not really "creative". Data Mining.....

Second is theory driven you give it "the rules" around an idea or artefact, parameters distilled into logistical mathematical equations behind each decision. It then measures its result according to the rules in an ongoing iterative process until an acceptable result is reached. Neural networks.....

Comments

Post a Comment

Popular posts from this blog

The Ecology of Culture

-
Image
Searle, P. (2017) Radium88 in Concert This document, The Ecology of Culture (A Report commissioned by the Arts and Humanities Research Council’s Cultural Value Project) by John Holden Visiting Professor, City University, London Honorary Professor, University of Hong Kong January 2015, was very interesting on a personal level. I paused to consider how the industries I've worked in have been a part of creative ecology model... where they fit into different sections and networks - and realised I've experienced all forms of the 3 spheres that make up his model. Publicly funded, Commercially funded and Amateur/Homemade. I received benefits in the 1990's to be in a band (publicly funded) - which made an enormous difference to my creative potential - I had time to make music (which I still do today - 11 albums and counting) and also created costume and art which led to starting my own business... obviously I hoped at that point that the band would be invested in by
Read more

The Initial Pitch (week 25)

-
Image
Jemas initial benefits and problems list (2018)   We want to find new ways of working for the future that will inspire and enhance the skills of designer and cutter while helping solve the problem of fabric waste and ecological impact. VR (or AR) would allow a cutter to drape in real time on a stand without actually using any fabric. This could be viewed from all angles in real time in the same VR space by a designer anywhere in the world, adjustments made, decisions on fit and fabric enacted instantly with the touch of a button. The technology could then be integrated with patterning software to digitally take apart the toile, make a pattern and send to factory. It would fuse the skills of an already talented workforce with technology that already exists (albeit in multiple forms) Tilt brush has proven the possibilities of intuitive artist creation in VR. Marvelous designer / Optitex already have fabric draping engines used for 3D simulation of 3D patterns. Combining al
Read more

major project 2 (plans plans plans)

-
Image
"Kandinsky" Collaboration with Deepart (J.Hewitt 2018) Best way to get started on something is just to start. so I sat down with the DeepArt IO and doodled exhaustively, using a combination of the basic silhouettes devised with C. Peacey and some vintage pattern blocks to reference the "rules" and history of cutting. I feel there are varying levels of "success" in these. They may or maynot fulfil prerequisites for computational creativity. I think they are certainly aesthetically pleasing rather than simply novel, but how far the algorithm itself is part of the aesthetic evaluation process I need to investigate further. Many are absolutely stunning images, and were I a fine artist who had painted them or had I written the Deepart Program myself  they would easily be worthy of submission as my finished pieces. However, I am a pattern cutter and I didn't write the program, it is simply a design starting point for my 3D pattern creation exercise.
Read more