Artificial Intelligence is a massive area which is currently growing at an incredible rate. How and why is this happening?
This article attempts in plain English to explain the major concepts of the area of AI and it’s associated fields.
I’m going to do my best to explain this without using maths or using programming… good luck me!
I will be putting together more detailed area-specific articles but let’s start with an introduction to the main areas.
- Learning
- Artificial Intelligence
- How
- Next steps
[read more=”Read more” less=”Read less”]
Learning
Standing on the shoulders of giants
Ok, firstly you’re carrying around one of the most powerful computers in existence. The human brain.
Like DNA, there remains mysteries in how it actually works. Yes we know some of the general ways it works but there are a lot of specifics we don’t.
Every generation adds a bit more to the knowledge of the world. So what can you add?
Bernard of Chartres from around 1000AD is accredited as the source of “If I have seen further it is by standing on the shoulders of Giants.” The idea being, if you can use what someone else has already accomplished you can do a bit more on top.
Even around 1000 AD appreciation of others work has helped our species go further.
Artificial Intelligence (AI) represents an era of growth in technology and understanding which will be stood on by future generations.
Nature as a guide
Nature has a very good system for finding the best way that works. Make lots with variations.
The variations that don’t work die away and the ones that succeed continue. Did you use a motorola phone or a zune? I had both and I loved them, but they weren’t to be.
People use nature as a template system for “how does that work.” We have flight from birds. Submarines from whales.
A massive current focus for people and as a result science is: “how does the brain work?”
In a computer a Central Processing Unit (CPU) is the main brain of the computer. Some things in technology are far better than a human brain. Doing maths for example, a computer is far faster.
When you consider learning and reasoning the human brain wins out massively as the technology is not as advanced yet.
There are major areas within this area of study of the brain, Artificial Intelligence, where it is very new and we’re attempting to match how the human brain works.
Adapting
Teach a human or a computer to do one job with similar inputs and both can do the same job.
The computer will probably do a better job as it’s a machine, where a person isn’t.
Now, ask the human and the computer to adapt to a situation with statistically varying inputs.
At our current technology level, the human will adapt and the computer will most likely fail.
Flying uses planes. There are many ways of flying or falling with style. One branch of flying is called aeronautics and that is a science.
Yet rocketry is also a science. Both of these live with aerodynamics which explains the motion of air and the forces on moving bodies moving through the air.
So you can have many sciences within one area.
Artificial Intelligence is the collective term for the area of getting computers to act and adapt like a human brain.
The study of Data Science is the branch of science which delivers the area.
Wikipedia defines data science as an interdisciplinary field that uses scientific methods, processes, algorithms and systems to extract knowledge and insights from data in various forms, both structured and unstructured … Nate Silver referred to data science as a sexed up term for statistics.
Artificial Intelligence
Working a definition
The definition of AI depends on who you ask
- Britannica: Artificial intelligence (AI), the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings.
- Wikipedia: Artificial intelligence (AI), sometimes called machine intelligence, is intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and other animals.
- Techopedia: Artificial intelligence (AI) is an area of computer science that emphasizes the creation of intelligent machines that work and react like humans.
- Merriam Webster: 1: a branch of computer science dealing with the simulation of intelligent behavior in computers. 2 : the capability of a machine to imitate intelligent human behavior
- NewGenApps: Artificial intelligence refers to the simulation of a human brain function by machines. This is achieved by creating an artificial neural network that can show human intelligence.
Commonalities
Because the area is growing the definitions change but here are the major considerations.
- Firstly we have mathematics and a massive sub branch of that statistics, to guide decision making.
- Artificial Intelligence combines computer science, electronic engineering and mechanical engineering.
- Needing something to learn from / model upon is the study of Brain Science, which involves biology
- A massive area is communication and interaction with the current primary way through speech which involves Linguistics and Visual Recognition
- Also it asks computers to make decisions and as a result brings in work from philosophy and psychology.
- When you add this intelligence to machines that manipulate the physical world you have the area of Robotics.
So this is very much a multi-discipline area which includes many sciences.
Rather than have all this information in one person, teams depend on each other to work out their areas. To see further, stand on the shoulders of a giant.
Ethics
When making decisions ethics are the set of rules you as a person have to dictate if you should or shouldn’t.
Rather than get stuck into the moral and ethical debates here, I’ll link this section to a future article.
How
Statistics
I’m going to grossly oversimplify the area but in short it all boils down to one word. Statistics.
Following the basics of any system, Input > Process > Output.
- Calculator: Numbers > Addition > Result
- Patient system: Symptoms & Details > Reasoning > Diagnosis and treatment plan
So the key part for all of this is the process.
- The human body takes in a lot of inputs through the five senses.
- We learn and have experience to help us develop processes and our decision making.
- The result is always the product of getting the right inputs and having the process / experience to do something with it.
Rules and experience can break down into very clear processes and computer programming can instantly help here.
Now when you get to 50:50 decisions or “grey areas” is when things get a lot more complicated so we need some science to refine them.
Computer programming: Level 1
Firstly think of a cooking recipe. The statistics in this are almost 100%. There is very little wiggle room.
Someone can articulate how the process works and what you need.
It’s turned into a set of rules, an algorithm which the computer can follow.
This is the calculator, a tool.
Everything that follows is a subset of computer programming and artificial intelligence.
As a result of everything being interdisciplinary it’s easy to get lost. Consequently all that follow are subsets of each other.
Data Mining: Level 2
Ok, so we have all the absolute rules worked out, now we get to the grey questions.
What will the weather be like today? We always have weather but predicting it is still difficult.
“Red sky at night, shepherd’s delight. Red sky in the morning,shepherd’s warning” first appears in the Bible in the book of Matthew.
An old weather saying often used at sunrise and sunset to signify the changing sky and originally known to help the shepherds prepare for the next day’s weather
This is one of the earliest examples of how to weather forecast was defined from a data mining rule.
In modern terms, you don’t have to know how the tools work, you can just apply them to your data and they will spit out results.
People’s knowledge and experience are still applied to the outputs as to how useful / not useful the result is.
The human skill is turning those rules into something useful. So a human is required to be present to react to the outputs.
Machine Learning: Level 3
We have data but we don’t want to spend all day babysitting the process. There are many users with many choices.
More powerful computers enable these capabilities.
From the late 80s, early 90s this area emerged. Three different science branches are involved. Statistics, Neuroscience and Computer Science.
- Statistics needed algorithms, sets of rules on how to deal with data on an ongoing basis.
- Neuroscience are trying to come up with operational models that explain how the brain work.
- Computer science needed more and more powerful machines to apply all the rules.
What it does is adapt the models it creates when you give it more data.
A simple example is Netflix. You start watching a type of program, magically Netflix can recommend tv and programs similar to the one’s you’ve been watching.
Consequently using your choices and the choices of all others using the system, the statistics work out what you’re likely to watch.
It doesn’t always get it perfect and when it starts it could be wildly wrong. However as people make choices, you’re giving more data and teaching the algorithm.
This is why for Netflix, things like age, location, language and a history of the types of you program you like all factor into its algorithms. The rules are constantly evolving around different scenarios.
Deep Learning: Level 4
How
This area is an emerging area because this takes the statistics to whole new level, doing many calculations simultaneously.
Firstly very very powerful computing makes it far more possible to do what was impossible. A return to RISC based computing has enabled this area.
So you have lots of rules you’ve learnt in your life and all these rules are instantly available to your decision making.
Attempting to mimic how the brain works, deep learning involves building structures like your brain. These structures in your brain are called neurons.
Neurons have multiple inputs and resulting outputs. Connecting lots of neurons together is called a neural network.
Let’s take the following picture, how do you work out what the thing on the left is?
- Firstly do they have pointed ears?
- Are their pupils rounded?
- Have they a long nose?
- Also are they big or small?
There are a host of rules. Firstly your brain worked out the black shape and brown shapes are different things. Next you worked out general features which you know as eyes, ears, noses and fur.
Your brain did that almost instantaneously and you didn’t feel a thing!
Are the rules you’ve learnt complete in helping to identify between dogs and cats? None of them are perfect i.e. statistically 100%. You can hazard a good guess.
You run your data through all the rules separately you’ve learnt in your life to come up with an overall statistic. It’s likely its a cat on the left and a dog on the right.
Consequently as you grow up and learn, you create more and more filtering rules. This is how your brain works.
What rules led you to believe those are animals? Why aren’t those animals a horse? What rules would you need to add?
Adaptation
So you have filters in your brain. One rule might be that it doesn’t “have long legs” whilst another might be “not 6 foot’ish tall”. Ok, they’re probably not horses.
This ability to pass through multiple filters in real time, with lots of inputs is in part due to computing power. It is also, in part, possible due to improved programming languages that allow the addition of new rules (neurons).
One definition of deep learning (also known as deep structured learning or hierarchical learning) is part of a broader family of machine learning methods based on learning data representations, as opposed to task-specific algorithms.
If you see a hippopotamus, how do your statistics rules react?
Your first reaction might be a type of fat horse if you’d never seen a hippo before. That’s where the name came from.
The Latin word “hippopotamus” is derived from the ancient Greek ἱπποπόταμος, hippopotamos, from ἵππος, hippos, “horse”, and ποταμός, potamos, “river”,meaning “horse of the river”.
This ability to at least try guess an answer means computers evolve and improve over time.
Next steps
Ok, the sheer scale and areas of discipline required to understand the area are wide and varied. This puts people off.
Break it up. You just need to understand the various parts in their own context.
Science Fiction loves to make things appear unreal or just that, fiction. Artificial Intelligence is not a sci-fi word, it’s just a branch of science that uses a lot of statistics.
You don’t need to understand it all yet understanding which giants shoulders you’re on does help.
Words like statistics and algorithms scare people and they shouldn’t. When Netflix recommends your next watch it’s helpful, not scary, right?
You worked out the cat and dog I hope and that was your brain using neurons to perform statistics.
The mathematical process of Division scares some people.
If you have a calculator, you don’t need to know how to perform division. You just need to know how to use the tool and press some buttons.
I’ll link a follow up article to go into the areas in a bit more detail when written.
If there’s anything I can do to help, or explain, please ask.
[/read]
