Machines with Power of Imagination

Utkarsh Mishra (~utkarsh13)


1

Vote

Description:

Living beings have many God-gifted power from seeing and hearing to analyzing and imagining. Nowadays, AI, ML and DL are trending topics where we try to mimic the Cognitive Human functions by learning and applying them to machines. Generally, these topics rely on something called Knowledge Engineering (Processing the real-world knowledge). Especially, in Deep Learning, we mimic the neuro-structure of the brain to do many challenging tasks like analyze speech patterns, recognize objects and predictive forecasting. These things can be called as the replicas of seeing, hearing and analyzing for machines.

With the help of these topics, we can provide machines with human-like context and awareness. The goal of these topics to transfer human expertise to a computer program that would take in the same data and would come to the same conclusions as humans would or sometimes even better. But how about imagination, we humans can imagine or recreate a situation based on the provided physical conditions and the past experience. However, machines can’t do so. The primordial project we make after getting in touch with deep learning is image recognition where the model tries to recognize the numbers after being trained with MNIST handwritten digit. Similarly, the simplest hand on experience project we can think of after the machines have the power to imagine could be reconstructing a faded and white spotted image to its original self. It won’t be perfect but will be more than enough for us to understand and see everything clearly.

I will be explaining the basics of machine learning and deep learning and giving you all a brief introduction of psychology and neurobiology to understand how these things like seeing, hearing, analyzing and even imagining happens. Then we will go through the concepts, and I will be presenting my budding idea of implementation.

What follows is a rough outline of a typical 30 minutes talk:

  1. Basic Important Concepts and understanding of Psychology and Neurobiology mimicked in AI [7 minutes].

  2. The problem that is issued during the talk [3 minutes].

  3. Basics of Deep Learning to assess the problem [10 minutes].

  4. An idea to solve it [5 minutes].

Prerequisites:

Basic understanding of machine learning and deep learning.

Speaker Info:

Utkarsh Mishra is a third year undergraduate at BTU, majoring in Computer Science and Engineering. With more than 2 year of experience in machine learning, he did 2 internships related to ML and like to discover new things related to this field.

He also gave his first ever technical talk on my project related to 'Pothole Detection and Visualization on Google Map' in SciPy 2019 at IIT Bombay.

During PyCon India'2020, he wishes to share a problem and a relevant idea with fellow practitioners.

Id: 1729
Section: Data Science, Machine Learning and AI
Type: Talks
Target Audience: Beginner
Last Updated: