29-404, LPU, Jal-Del G.T. Road - Phagwara, PB, India
09:30 am – 05:30 pm

State of Artificial Intelligence Today

Published On: Monday, February 18, 2019 433 views

Hello Everyone!

 In this post, my desire is to pen down few thoughts on elusive artificial intelligence to describe current state and somewhat ponder over big claims made by tech companies competing in this market space. If you look around seriously in tech startup scene, most of the new upcoming tech product companies are ‘claiming’ to have figured out the AI puzzle and worked out-kinks to implement the technology.

 AI-driven systems must have an ability to process large amounts of knowledge. I mention 'knowledge' and not 'information' for a reason in this statement because having knowledge of something enables someone to follow the shortest logical path to attaining success. On the other side having just information about something merely enables awareness at best. A system built using AI must have or be capable to extract and differentiate knowledge from ‘information’ where information must be seen as a fundamental requirement for the AI system to start work.

 Here I would like to also point out two variations of knowledge which needs to harvested on-the-go. I classify it as dynamic knowledge vs. static knowledge. Dynamic knowledge is derived, extracted, and analyzed in real-time from raw data as soon as the system goes live and information is processed for further action. Static knowledge on the other hand is, as on date, the hardcoded knowledge as static information which doesn’t change often. For extracting actionable static knowledge, machine learning techniques have been developed to produce actionable insights from vast amounts of raw data. But we know as humans, that static knowledge won't serve us in every situation we're facing. Imagine as humans if we experiment this static knowledge in a social scenario as a rule of "One should be nice to other people" and apply this rule to all interactions, then even in situations where "One should not be nice" e.g. someone is trying to kill you, you can't be nice then. This means that the knowledge is subjective to the situation. Rules of engagement with other people change.

 Clearly, the emphasis needs to be on being able to produce dynamic knowledge from information gathered. There is a downside to all this and that is the well-known Utility Problem. For an AI system to be smart, same system needs to be designed with AI architecture capable of learning relevant knowledge from raw data or plain information without the nasty consequence of slowdown in overall functionality using clever optimization-related algorithms.

 Thank you for reading and I hope you’ll find this post useful and in case if you wish to add any additional point to the post, then please leave a comment below.

Post by: Sarabpreet Singh
Contact author: click here

Key Words

claim state produce elusive information Everyone capable enable thought down which know extract artificial actionable static tech must from made knowledge desire intelligence post nice describe process current space someone learn point people amount human should system need Hello look work somewhat change company market situation dynamic ponder data compete rule something around seriously