Death of Deep Learning?

Source: Deep Learning on Medium


Go to the profile of Param Popat
Photo by Cody Davis on Unsplash

A few weeks ago, I came across an article that discusses the results of an Experiment carried out by MIT and published in MIT’s Technology Review.

They studied 25 years of research papers in AI which eventually led them to conclude that Deep Learning is dying. This is not to scare or to demotivate because it gives even better insights into what the future holds.

I personally feel the reason behind this conclusion is that over the past many years the research has actually stalled within young researchers’ fraternity, what I mean by this is that Almost everybody entering this domain mostly confines themselves to traditional methods and use-cases and end up publishing results on what has already been achieved many years ago. Please keep in mind that this is not to disavow anybody’s research or efforts, but the point is that We as engineers and researchers are going to be the torchbearers of Modern AI and we are currently failing on our responsibilities by sticking to the conventional definitions to work upon.

My suggestion to all of you would be that look beyond what’s ordinary, think beyond what has been achieved or what can be achieved, think and work on ideas which you feel are overly ambitious and 99% are gonna fail, that’s when you’ll come up with something new, something that adds to this domain and that would take deep learning even further( or to be precise even deeper). Explore concepts like Deep Reinforcement Learning, One-Shot/ Few-Show/ Zero-Shot Learning, Meta-Learning, etc.

I am not saying you skip the basics and jump onto these Buzz-Words because everything you’re gonna do is going to be built over these basics. But keep these things in mind, focus on an ambitious idea in your free time and Keep on Learning New Things Every Day.

Something that I always like to believe is that in this world especially in the field of Machine Learning “Quality always beats Quantity”.

The article rightly mentions and I quote:

The 2020s should be no different, says Domingos, meaning the era of deep learning may soon come to an end. But characteristically, the research community has competing ideas about what will come next — whether an older technique will regain favour or whether the field will create an entirely new paradigm.”

All the best. Feel Free to Send a message if you agree/ disagree with my observation and let us all unite to drive this beautiful domain forward.

To put the title picture into context: We’ve only scratched the surface and hardly explored anything. There’s a lot more out there.

[Originally an E-Mail Sent By Param Popat with Subject: Death of Deep Learning?, on Feb 1, 2019, 7:46 PM]

[As an Article on LinkedIn By Param Popat with Title: Death of Deep Learning?, Published on Feb 16, 2019]