What is artificial intelligence (AI)?
AI is a branch of computer science that deals with creating intelligence in machines via software. The field is older than one
might suspect - it was started in the 1950's.
What is the goal of AI?
The ultimate goal of AI researchers is to create human level general intelligence. However, this goal is still anywhere from 5-20 years away.
A more realistic goal today is creating specialized intelligence that performs equally or better than humans at solving certain problems.
Where can I see examples of AI today?
The use of AI technologies today is already somewhat pervasive. AI is responsible for controlling characeters in many video
games and is also used for trading applications in financial markets and medical diagnosis to name a few examples.
How does the future look for AI?
For decades now we have been witnessing Moore's law in action. This has led to faster and faster computers capable of solving
ever more complex problems in shorter amounts of time. In fact, we are already at a point where very few people actually
need the fastest desktop CPU's available. As AI technologies usually require a lot of processing power it is becoming feasible
to solve extremely complex problems on regular desktop computers.
We can not expect Moore's law to hold indefinately and barring a paradigm shift (such as nanotechnology, quantum computing or three dimensional circuits) in computer hardware it will most probably slow
down in the next decade. However, even conservative estimates predict that before the year 2020 the fastest desktop CPU's will have
the computational power to simulate a fully functioning human brain. Consider also that we are moving at brisk pace towards
understanding how the human brain works, with brain scanning technology following a sort of Moore's law of it's own.
With this in mind, the future for AI is sure to be very exciting and laced with new opportunities. For those interested in
further reading, books by Raymond Kurzweil such as The Age of Spiritual Machines and
The Singularity is Near are highly recommended.
Tools of AI
This section briefly describes the main tools and constructs used in AI research. This list is by no means exhaustive.
Neural networks are constructs inspired by the inner workings of the human brain. Each network is a collection of neurons that
are linked in a particular fashion. Input data passes into the network on one end, the data is then processed by the network and
output is returned at the other end. Neural networks are useful for pattern recognition, classification and function approximation.
Genetic algorithms / Evolutionary computing
Genetic algorithms are a technology inspired by biology and evolution. Starting with a population of randomly created individuals (GA's)
the fitness of each is determined by a fitness function. After all individuals have been evaluated the fittest ones (with the highest fitness score)
are selected for breeding where offsprings are created by mixing properties of their parents with crossover and mutation. This process is repeated
hundreds or thousands of times with results usually improving every few generations.
GA's are an excellent optimization tool for highly non-linear and/or discrete problems and can also be used as solution constructs.
Genetic programming involves automatically creating a program for solving certain tasks. The technique is very similar to genetic algorithms
except that the result is an actual program (lines of code) that can be read and understood by a human expert. Genetic programming is useful for
solving problems having properties that make neural network processing difficult or impossible.