Monday, November 7, 2016

Watson, Beep...Bleep...

Ever since I started using Texas Instruments sensor tags interface with IBM Watson, I have been an audience of this fascinating and uniquely-amazing service. I tried other similar services but Watson-Cloud stood out in terms of it's performance. Digging deeper into this fascination, I went ahead and analyzed my blog posts with Watson tone analyzer.

It seems I have been angry in most of my blog-posts, Excessively Joyous is others, and unrealistically sad in some. I have been consistently open and conscientious with all of my posts, as per Watson, while Confidence varied from 0.20-0.60.
It's difficult to tell if Watson analysis is right or if I really was angry while writing those posts. But that is not a concern. The gap between our understanding may never be resolved. Not unless Watson can let me spawn a NN for my stuff and sync it up only with what's going on in my head.

A global brain may never be able to solve few things in totality which a multitude of local brains can, Perhaps.

Friday, September 23, 2016

Paradigm-Embrace

An equation means nothing to me unless it expresses a thought of God.
- Srinivasa Ramanujan ( The Mathematician ).

It's true some-equations do evoke an expression of a divine thought. Off-lately there is much traction around a paradigm called Deep-Learning. It's a field of AI which has not only revived the idea of artificially designed intelligence, but also, may as well, be the beginning of a whole new world of cognition. This article will not suit well with some who are considering the third wave of AI as yet another wave, and are confident that this one like it's predecessors will get subdued with time.

Revival of this field of thought is attributed to a very smart and subtle tweak by Prof. Hinton and others, which essentially utilizes fractal based layering of deep layers for reduced complexity and increased relevance. Like various other designs in nature fractal nature seems to be a key for information processing.

Information around us is exploding at an exponential rate, but this growth, instead of being random, may be following a fractal evolution path. If so, intelligent designs similar to Deep-learned-nets will provide a framework for following this growth, and most importantly will help nudge this explosion towards a relevant future direction.