Friday, April 28, 2017

Why do we need the Democratization of Machine Learning?



https://medium.com/@matelabs_ai/why-do-we-need-the-democratization-of-machine-learning-80104e43c76f

We are living in an era of hype. In this article, I am trying to discover the hype around Artificial Intelligence. The First thing I want to clear is that ML/DL are algorithms, neither conscious nor intelligent or smart machines.
There’s more to Artificial General Intelligence than just Machine Learning or Deep Learning. I agree that Deep Learning has penetrated industries and it holds the potential to disrupt industries, but it is nowhere near to being conscious or an intelligent machines.
The term Masses is so powerful that once aligned together, it brings revolutions.
Singularity, AI taking over the world, End of the world were one of the most used phrases in the media last year. Media being a primary source of information for most of the people, including investors and financial institutions which makes it a vicious circle fuelling the hype and adding air to the bubble. It can be good or bad, but if it bursts it is going to affect us all (includes you too). If you talk to the researchers or the experts, their views on this are poles apart. They echo AI is over-hyped and AGI is far away. We have a long way to go to achieve true Intelligence. Current AI(Deep learning/Machine Learning) applications can only do what they are trained to do. When the knowledge flows from the source to the publications, it gets distorted.
AI is a pie and everyone wants to have a bite of it.
AI/ML tools are limited to the researchers, industry, colleges and labs. They are not accessible to the masses in a simple easy-to-use form. In the last 2–3 years we have seen some pretty interesting use cases of modern deep learning. Apps likes Prisma showed us that we just need to be creative to make it available to the masses. It uses deep learning to extract styles from images(“Vincent Van Gogh’s Starry night”, “Picasso’s Self Portrait 1907”, “Frida Cahlo”) and apply extracted styles on to your images. AI as a lawyer, AI as a painter and AI as a doctor and so on, just name it and you will find a startup/company working on it.
If I talk about industry use cases, Baidu runs a food delivery service and it uses AI to predict how long the food will take to get to the customers. Google uses deep learning for machine translation, search and for other products. Facebook uses it to recognize faces in your images which you upload on Facebook. GANs (Generative Adversarial Network) are the type of neural networks which learns to imitate and produce original content. Google Research used GANs to invent an “encryption” protocol. Andrew Ng, Chief Scientist at Baidu research says
AI is the new electricity
In the next 3–5 years AI/ML is going to affect almost every industry. Infact it is a new industry and it is worth billions of dollars right now. A report by Bank of America Merrill Lynch forecasts it to be USD 14 Trillion by 2025, and it is a HUGE number.
Source: Wall Street Journal
61% of companies in this survey by Wall Street Journal are planning to use AI within 5 years for business analytics, 45% for Machine Learning and 21% for self-learning robots. According Stackoverflow’s Report, the global economy is seeing an explosion in the demand for Machine Learning expertise this year as much as 3700 times more than year 2016. This is HUGE!
And, we are missing out on one very important user group, which is general people. What if users can participate in the creation of the AGI. Creative users, coming out with use cases which we had never thought of. What they need is easy-to-use tools. AI with everyone is the future of AI. Personal computers revolutionised industries and powered people with tools that amplify one’s creativity and productivity. Internet came out and connected us all together. Handheld devices like Mobile and tablets have given us all of that power in our hands. Now it’s time for AI to go into everyone’s hands, next revolution is getting in shape now and everyone has to participate in shaping it. AI can find the cure for diseases which we were unable to cure for centuries like Cancer, Sclerosis and other diseases. It holds the power for a better future, a future which we have always dreamt of.
As far as the world is progressing, most of the programming might not even exist in the future. AI will code for us in the future. Data is one thing which I bet on, second is how we think and how we use AI for our betterment and betterment of the society. We dream of humans and AI working together as companions making world a better place.
It is our responsibility to shape it well, so that it doesn’t go rogue. Our curiosity will lead us there. One day we will reach there, true AGI.
This is what we need:
  1. Open Source tools and libraries
  2. Open source datasets to encourage developers
  3. Open educational material like videos, tutorials, articles and easy to unserstand description of research papers.
  4. Open groups to discuss AI/DL and community building.
  5. All of the above do exist in some form, what is not there is a tool for masses to play around with ML/DL and contribute back to the society. Until now.
A revolution is happening right now, the whole world is embracing it. Mate Labs wants to be there as one shoulder for people to bank on and fuel this revolution by enabling everyone with the right tools, libraries, datasets and educational materials. We believe in building a community of enthusiasts, curious and thirsty folks standing for innovation and a bright future.
We have open sourced the implementation of All Convolutional Networks which is available on Github. We have also published a suite of easy-to-run scripts to install Tensorflow and Docker on your system to get you started. More to come soon.

Previously we’ve shared.
  1. A very crisp explanation of All-CNN’s implementation. How these researchers tried something unconventional to come out with a smaller yet better Image Recognition.
  2. Our Vision. What Everyone is not Telling You about Artificial Intelligence

No comments:

Post a Comment