Artificial intelligence is the future.
Artificial intelligence is science fiction. And it is already part of our everyday lives.
All these three statements are true, it all depends on what flavor of AI you are referring to.
For instance, when Google DeepMind's AlphaGo program defeated the South Korean Master, Lee Se-dol in the board game “Go” earlier in 2016, the terms AI, machine learning and deep learning were used in the media to describe how DeepMind won.
And each one of the three is part of the reason why AlphaGo beat Lee Se-dol. But they are not the same thing.
It's very easy to confuse that artificial intelligence, machine learning, and deep learning are the same thing.
The easiest way to understand the relationship between them is to visualize them as concentric circles with;
- AI – the idea that came first and is the largest
- Machine learning – which blossomed later
- Deep learning – which is driving today's AI explosion and fits inside both
From bust to boom
AI has been a huge part of our imagination in research labs ever since a handful of scientists rallied around the term at the Dartmouth Conferences way back in 1956 and “gave birth” to the field of AI.
In the decades since, the artificial intelligence has been heralded as the key to our civilization's brightest future. And it was also tossed on technology's trash heap as a harebrained notion of over-reaching propellerheads. And, to be honest, up until 2012 it kinda was a bit of both.
Over the past years, artificial intelligence exploded and especially since 2015. Most of that has to do with the wide availability of GPUs that made the parallel processing way faster, much more cheaper and tremendously powerful. Another thing that had an impact was the simultaneous one-two punch of practically infinite storage and a flood of data of every stripe. This involves the whole Big Data movement:
- Mapping Data
Artificial intelligence – human intelligence exhibited by machines
Going back to the summer of '56 on that particular conference, the dream of those artificial intelligence pioneers was simple: to construct complex machines, which was enabled by emerging computers, that possessed the very same characteristics of human intelligence.
This is the concept that we now think of as “General artificial intelligence” – fabulous machines that have our very sense, or maybe even more. It also has all our reasons and tends to think just like we do.
You've seen such machines countlessly in movies as friend — C-3PO — The terminator. General AI machines have remained in the movie industry for a good reason; we can't pull it off just yet.
What we can do falls into the concept of “Narrow AI”. Technologies that are able to perform specific tasks as well, or even better, than we humans can. A great example of Narrow AI is the image classification on services like Pinterest of face recognition.
Those are examples that are currently in practice. These technologies exhibit some facets of human intelligence. But how? And where particularly does that intelligence come from? That sends us to the next circle – Machine learning.
Machine learning – an approach to achieve AI
In essence, machine learning at its most basic is the practice of using algorithms to parse data, learn from it and make a prediction or determination about something in the world.
So, instead of hand-coding software routines with a particular set of instructions to perform a task, the machine uses a large amount of data and performs that particular task. Amazing, right?
As it turned out, one of the best app areas for machine learning for many years was computer vision. Not that now it's obsolete, but it still requires a great deal of hand-coding to get the job done.
People would generally go and write hand-coded classifiers like edge detection filters so the program would
- Identify where an object started and stopped
- Shape detection to determine if it had eight sides
- A classifier to recognize the letters “S-T-O-P”
From all those hand-coded classifiers, they would develop algorithms to make sense of the image and “learn” whether it was a stop sign or not.
It was good, but it was not yet mind-bending great. Especially on a foggy day when the sight wasn't the best, or a tree obscures part of it. There's a good reason computer vision and image detection didn't come even close to rivaling humans until in 2016, it was too brittle and too prone to error.
But, time and the right learning algorithms made all the difference.
Deep learning – A technique for implementing machine learning
Another algorithmic approach from the earlier machine-learning crows – artificial neural networks – came and mostly went over the decades.
Neural networks were inspired by human understanding of the biology of our brains – all those interconnections between neurons. But, unfortunately, unlike a biological brain where any neuron can connect with any other neuron, the neural network has discrete layers and directions.
For instance, you might take up and image, chop it up into small tiles that are inputted into the first layer of the neural network. In the first layer individual neurons, then passed the data to the second layer. The second layer does its job and so on until the final output is produced.
Every neuron assigns a weighting to its input – whether it's correct or incorrect relative to the task being performed. The final output is determined by the total of those weightings. Now, think of our Stop sign example from above. Attributes of a stop sign are chopped up and examined by the neurons:
- Its octagonal shape,
- The fire-engine red color
- Its distinctive letter
- The traffic sign size
- The motion or lack thereof
The neural network's task now is to conclude if it's a stop sign or not. It initially comes up with a “probability vector”, which is a highly educated guess based on the weightings.
In our example, the system might be somewhere along with the 85% confident that the image is a stop sign, 5% that it's a speed limit sign and 5% it's a kite stuck in a tree and so on. And the network architecture then tells the neural network if it's right or wrong.
Today, image and face recognition by machines trained to do so via deep learning is better than humans. And that ranges from cats to identifying indicators for cancer in blood and tumors in MRI scans.
Google's AlhaGo learned the game, trained for its Go match- it tuned to its neural network – by playing against itself over and over and over again.
Thanks to deep learning, artificial intelligence has a very bright future
Deep learning has provided practical applications of machine learning and by extension the overall field of artificial intelligence.
The process of deep learning breaks down tasks in such a way that makes all kinds of machine assists seem possible, even more likely.
Driverless cars, better preventive healthcare up to better movie recommendations are all here today and are used on a daily basis. The artificial intelligence is the present and the future of technology. And with deep learning's help, the artificial intelligence might even get to that science fiction state we've long imagined.
So, in essence, even though they have some similarities, deep learning, artificial intelligence, and machine learning are completely different processes.
If you have a suggestion about this article, please leave a comment below. Or, if you have an inquiry please contact us on our contact page.