The Grumpy Guide to AI, ML and Deep Learning
Oh no. Not another piece from someone banging on about how robots will take over the world, we will all be on the dole and that their new business has some AI stuff or product roadmap with machine learning on it?
So here is the thing. I get a lot of decks and a lot of pitches weekly. In 2012 I made the comment that everyone's business had something to do with Cloud. Even if you woke up and saw a cloud it gave you rights to whack it into a business plan and raise your valuation by at least 50%.
Today I get a similar vein, although the valuation is at least 200% currently and most people who are pitching to me seem a little confused by the difference in ML and AI. Just to be clear there is a difference.
Birthplace: Dartmouth Conference
Birth Year: 1956
Artificial Intelligence (AI) Human Intelligence exhibited by machines.
In 1956 those super smart people, pioneers, wanted to create complex machines that had the same characteristics as Human Intelligence.
Today, general AI as we know it (and if you have ever watched the Terminator movie, then that's it in a nutshell) is still a thing of science fiction but may not be one day.
Today we can do the “Narrow AI.” concept of Weak AI that means a machine can perform specific tasks as well as, or better than, a human can. Apple’s Siri, Amazon’s Alexa, Face recognition on Facebook etc are great examples. Others exists in all parts of our daily life.
Machine Learning — An Approach to Achieve Artificial Intelligence
Machine Learning is the practice of using algorithms (people love using this in 121 pitches) to parse data, learn from it and make a prediction about something in the world.
No more hand coding lots of software routines to do specific instructions to achieve something the machine is instead trained using the large amounts and data and algorithms to give it them ability to learn how to perform a task. Cool hey? And from now on if you have read this and are pitching to me, you better be able to give me an example if ML is in your pitch.
So this all came about from different algorithmic approaches, clustering, inductive logic programming, decision tree learning and the list goes on. This was proper early AI crowds and none achieved the goal of general AI and even Weak/Narrow was pretty hard to achieve.
But the early adopters, those crossing the chasm, were not entirely wrong and it was just a case of time and the right learning algorithms.
Deep Learning - A way to implement Machine Learning
Yep more learning.
Deep Learning is feeding a computer system a lot of data, which it can use to make decisions about other data. Data is fed through neural networks.
Neural networks are logical constructions which ask a number of binary questions of every bit of data which passes through them and then classifies the answers received.
Focus is on developing these Deep Neural Networks which are logic networks of a complexity to classify gigantic datasets for companies such as Google and Twitter.
Deep Learning what can it do?
Let's do another favorite of 2017 buzz word. Autonomous vehicles.
Navigation in self driving cars. Utilising the sensors and onboard analytics cars, yes cars, are learning to recognize obstacles and deal with them appropriately using deep learning.
My favourite one is being able to predict legal proceedings. Imagine all that money you could save on lawyers.
The list goes on and it's not all hype.
I still refer people to the Terminator movie from 1984, yes 1984. I dont think its far wrong on some fronts. In 1993 I was seriously interested in AI and no one had a clue what I was talking about. I wrote a white paper titled “Will Computers ever Love” on the basis that I felt by 2025-2030 we might have examples of AI over taking Human Intelligence, people maybe even forging relationships with AI let alone singularity.
I often say the World Wide Web/Internet is and has been the Wild Wild West, but what about AI? How wild will that get?
The Grumpy Entrepreneur @TheGrumpyE