7/27/2016

Intelligent Machines will need 5 Basic Senses

You can train a model by feeding tons of data and tuning each of the layers to decipher tiny dots into images into things into people and such through pattern recognition machine learning tools.

You can do the same with audio.  In fact, you can take some spoken word, interpret what was said and then send it back out in another language.  One input, translator, many output possibilities.

Likewise, you can receive many inputs, translate, send them all out in one specific language.

So we can process sound and sight, apply machine learning techniques and understand the content of the media.

What about smell.  Do we have any device that can interpret smells.  For instance, you've got a bag of popcorn in the microwave for 30 minutes, strong odor emitting, can a sensor pick up the burning smell and trigger an alter through machine learning?

When we capture sound or visual media, can we also pick up the surrounding environment such as odor.  Create a machine learning model that is trained on smells?

How about taste?  Can a machine tell the difference between Coke and Pepsi?

How about touch?  Difference between a dolphin's skin and a porcupines skin?

I believe those are the 5 basic sense we humans possess?

Some say we have a 6th sense, intuition.  Fairly certain we don't have a model for that one.

How about taking a scan of an object, and determining the molecules that make up the object?  Based on the arrangement, identify the object based on the model?

How many levels of attributes can we capture simultaneously?  Sight, sound, touch smell and taste.  How about location, who's in the room with you, time of day, etc..

It seems if we are to build intelligent machines that interact on daily basis, they should at least have the 5 basic senses.  Otherwise, an inherent disadvantage.

What if we could scan a portion of space, like a 2 x 2 x 2 location, identify each molecule contained within that space, label each, and track their movement over time.  What if we did this with every seeable molecule inn the Universe.  Store in big data, play scenario back and forth, then predict future behavior of each molecule.  Suppose we'd need more processing power, enter Quantum Computers stage left.

Having the ability to process more than 1 logical bit decision per millisecond, it can process multiple depending on the number of Qubits, you could exponentially apply probability calculations on possible movement of molecules over time.  And like the monkey typing on a keyboard, eventually write complete work of Shakespeare, maybe we could get better at predicting movement of things over time, for starters, the weather.  Planets.  Galaxies.  Electrons.  Quarks.  You name it.

No comments:

Post a Comment

Root Cause Analysis