Can "Common Sense" Be Given To AI? [How Intelligent Is AI?]

Introduction

Common Sense; like we may have been aware of; is our inner ability of discretive judgement which is actually based on some awareness and perception of facts and of conditions. This ability is an absolute prerogative of almost all humans. Though it has been said that "common sense is not common", but a measure of it is actually common. That is why someone would know that placing his hand on a red-hot iron would cause some nasty burns. But in this advent of sophisticated technology, has AI been able to be taught common sense? Well, we will find out soon. Welcome again to the future of emerging technology.

[Image Source: Maxpixel CC0 licensed]

Common Sense and AI

One of the purposes of Artificial Intelligence is to simulate the human intelligence and experiences; and even better; in an unnatural way. But there has been an area which has been almost sidelined until very recently; and this area is "common sense". This is because, AI, like we may have been aware of, is designed to relate cognitive experiences from the database of information already pre-programmed into its neural network using some complex algorithms.

No doubt, AI has been seen to advance in complexity, which can almost equate or surpass the cognitive ability of the human brain. That is why we have seen AI perform complex activities; but what about simple tasks that a 5year old kid can perform? Maybe the "common-sense-not-common" notion would suffice more for AI than humans; or maybe not. Well, let's find out the advances that have been made to incorporate common sense into AI.

Teaching Common Sense to AI

After so many failed attempts from so many researchers within the last decades, a major breakthrough has finally been heralded in the pursuit to make AI have common sense.

And to lead this revolution, Paul Allen (co-founder of Microsoft inc, and founder of Allen Institute for Artificial Intelligence, AI2) has come up with an idea of "Project Alexandria", which the sole purpose is to teach robots "common sense". Sounds like history is about to be made again.

[Image Source: Maxpixel. CC0 licensed]

Before I proceed, I would like to state that the reason behind the earlier failed attempts to teach machines commons sense is simply because it is way too complex. For example; No artificial Intelligence system currently developed can precisely answer some wide range of Common questions; like: If I light a candle in both ends and keep it under a fan, will it burn vertically or horizontally?". The average man would know that once a lit candle is kept under a fan, it would go off; and you don't light a candle from both ends.

Can I tell you something? In 2016, Google's DeepMind AI; AlphaGo, was seen to beat the human Go champion; Lee Sedol, according to a report from the official webpage of DeepMind. But astonishingly, AlphaGo didn't even know that Go is actually a board game. Yes! This is because it wasn't designed with common sense.

Okay let's continue on our survey of the future advances to teach AI common sense.

How do Scientists hope to achieve common sense in AI?

First of all, we have to recognise the fact that humans can switch between thoughts, and this ability has given our reasoning ability a very unique fluidity. So in the light of this, a method will have to be devised to accumulate huge amount of data from randomly selected people. 

This data would be gotten from texts (with the aid of "machine learning"), from multimedia (aided by "machine vision"), from direct neural link-programming, and from multiple input from crowdsource.

Now, the summation of these various AI data collection methods would be put into a single pool of knowledge or repository, which would be called "Machine Common Sense Knowledge". This is actually the idea behind project Alexandria.

[Image Source: Designed by me @samminator using Microsoft Paint]

Now, with the data and information coupled up in this repository, special algorithms would be run on them, which will be used to inform a host of applications that run on AI (this is just like transferring the "common sense" consciousness from humans to machines). With this, AI can use the available data to draw up inference from real-life scenarios as they present themselves.

Common sense = Simulated

Future applications of this

This technology would come in handy in various places; two of which are:

• In medicine: No doubt, we have seen AI in medical diagnosis and also in robotic surgeries. Incorporating commons sense into these medical AIs would further enhance their efficiency and would also enable them to manage complications in a human-like manner.

• In industrial robotics: Industrial robots have been seen to be bunch of machines programmed to perform a particular task. But with the advent of this tech, their area of application would be broadened. Instead of telling them what to do, they would figure it out themselves. Now this is sounding like some tales from "neverland", but this is the nature of emerging tech.

Conclusion

Common sense has been one of the distinguishing factors between humans and AIs. As a matter of precision, it has been exclusively preserved of humans to possess this ability. But as technology is advancing, all these are about changing, and very soon, we would be sharing common sense with AI. But until then, let us keep an open mind as we welcome advanced technologies.

Thanks for reading

References for further reading: Ref1, Ref2, Ref3, Ref4, Ref5

gif by @foundation