I came across this Verge article by James Vincent (@jjvincent) and found it very meaningful to share and to get under the hype and verbose of AI / Machine Learning / ANN / Statistical Learning / Deep Learning / Swarm Intelligence. It is a good reminder that under the hood of “Intelligence”, we have sophisticated models that are not so “thoughtful” and still get labeled as “Artificial”.

Over the last 30 years of my association with Neural Networks and Genetic Algorithms, the definition of AI has evolved. As new-tech becomes more mainstream and crosses the initial euphoria and hype(‘cycle’), businesses do not view them as intelligent anymore, one such example being OCR, and these get pushed into the category of standard technologies.

In this primordial soup of Artificial Intelligence models, we observe the evolution of technologies pushing our understanding of intelligence. And we shall wait and see as it approaches singularity to emerge in the form of Artificial “Life”. But till then, it is for us to continuously till, sow, harvest to push the boundaries; but acknowledge limitations that exist today even with the most complex of learning models today:

1. they need a large amount of data to learn to do meaningful stuff;
2. they still represent intelligence which is extremely narrow and can barely compete with a mentally challenged cockroach;
3. it cannot explain its learning (and neither can the architects of AI models explain how and why the AI application took a certain decision)

Read this interesting article which expands on the above points with interesting examples:

These are three of the biggest problems facing today’s AI

In parallel, but not yet strongly connected with AI, we have seen the development of blockchain technologies. Created by man, we see an evolving system with decentralized ledgers, systems continuously competing and rewarding, and unstoppable! Did we accidentally create Artificial Life?

Ten years back, I was debating with myself on the structure of global Financial Systems – supposedly deterministic but with failures of regulation dotted at regular intervals over the last century. In comparison, the evolution of Life over millions of years has leveraged underlying chaos and still emerged as a winner. It seems very plausible that Blockchain will achieve this feat and provide us with insights to re-design existing learning models.