Marvin Minsky once said that artificial intelligence is “the science of making machines do things that would require intelligence if done by men.” Well, if artificial intelligence is the science, AI software is the means of making the science a reality. But what is it really? And where is the future of artificial intelligence headed?

Definition of Artificial Intelligence Software

Artificial intelligence software is a bit hard to define more deeply because it uses many different techniques to achieve its goal:

  • Neural networks attempt to mimic the learning potential of a human brain by mimicking a web of neurons. The largest neural network so far mimicked 2 billion neurons whereas the human brain has roughly 100 million neurons.
  • Machine learning is a method of creating artificial intelligence that enables a computer to devise complex models and develop algorithms that help it to predict patterns.
  • Deep learning uses neural networks to deeply analyze data points so that it can more quickly recognize patterns to predict behavior such as shopping patterns or search query patterns. The neural networks that deep learning software utilizes can reside on just one computer, or make use of cloud based computing to span the entire internet. Google uses deep learning for it’s AI platform and incredibly accurate image search engine. We’ll talk more about that one later.

As you can see, it’s a little complicated to define, as the science is still in it’s infancy, but it actually gets even more interesting. For instance, the software’s programming language is every bit as important as the type of system it is using to create the result. Java, Scala, AIML, IPL, Lisp, Smalltalk, Prolog, STRIPS, Python, Wolfram Language and C++ are just some of the programming languages used to create AI software and now, the first open sourced deep learning library written for Java and Scala have entered the ring.

Famous Examples of Artificial Intelligence Software

From C-3PO of Star Wars to Robby the Robot in Forbidden Planet, we don’t have to look to far to find interesting examples of artificial intelligence in fiction. But will we have artificial intelligence capable of making creatures such as these in our lifetimes? It’s quite likely that we will. The learning curve for AI software is increasing at an enormous rate – an exponential rate, in fact. We are currently at the cusp of scientific advances in this area that will likely change the face of human existence forever.

Artificial intelligence software is being used everywhere we look. The Defense Advanced Research Projects Agency (DARPA) used AI software to create a digital teacher to train Navy members faster than ever. IPSoft attempts to create the illusion of empathy in customer service chat bots by detecting anger in a customer’s tone and developing an appropriate response, and the list can go on.

graphical drawing depicting artificial intelligence software

Let’s look at some of the most impressive and lifelike AI software programs today:

Caffe

Caffe was developed by the Berkeley Vision and Learning Center. It gives access to a deep learning framework that focuses on speed and expression. It has a pretty fierce group of adherents on Github; they believe it to be the best artificial intelligence software available though opensource today.

Torch

The open sourced machine learning library called Torch uses a script language that is based in Lua. This lets it provide a huge range of algorithms that allow users to do some serious integration. The Facebook AI Research Group, Yandex and IBM all use Torch for their projects. Now, it is available for iOS and Android as well.

Deeplearning4j

DL4J is the first open sourced, commercial grade deep learning library that is written for Scala and Java languages. It can import neural net models via TensorFlow, Torch and Theano and uses the Skymind as its support arm. This artificial intelligence software download will change the AI programming world by making it more accessible to all.

TensorFlow

Google just open sourced its artificial intelligence engine, TensorFlow. Yup, that’s right. Google open sourced artificial intelligence software for PC. But before you start thinking that this is a purely altruistic move, think again. It’s more based on the old adage, “use it or lose it.” Google figures that users will help develop it further. They can then use these developments to improve the platform. It’s a great idea that could ultimately serve all of mankind, considering Google is years ahead of every other AI software developer. It may actually be one of the most exciting pieces of AI news to date.

The Human Brain Project

It’s not actually a specific software, but The Human Brain Project gets number one placing for it’s lofty goal alone. It was launched in 2013 with the goal of reaching the same processing power of the human brain. So far, the ramp up phase of this process has seen a digital reconstruction of the neocortex of a rat brain and neuromorphic computing systems.

Neuromorphic systems are capable of using models of brain cells and synapses to create run simulations of brain circuit systems that exhibit the plasticity (brain’s ability to grow with learning) of the human brain. Basically, it comes down to an advance that will make it possible to understand even the most complex neurological diseases as well as mimic human thought. A worthy goal indeed.

Current Breakthroughs in Artificial Intelligence Software

All of the examples listed above are of weak artificial intelligence, otherwise known as artificial narrow intelligence. This means that they are good at doing what they do but that’s all they do. Strong artificial intelligence is the true goal, meaning that the program will have a general intelligence capable of learning any new skill, just like a human. We haven’t done this yet but it’s the next goal and we may reach it very soon. Once scientists achieve this, the next goal will be artificial superintelligence. This will surpass human capabilities and perhaps make the need for human labors obsolete.

What do we need to achieve artificial super intelligence? Well, the human brain performs the equivalent of one billion billion operations per second, AKA calculations per second. Of course, many of these operations are focused on performing our basic autonomic functions. Take away the need for the brain to concentrate on all these things and it would need much less performance to exhibit the functions of higher intelligence. Currently, our fastest super computer, called the Sunway TaihuLight, is capable of performing 125 quadrillion calculations per second, or 125 thousand trillion. So we are starting to get there in one sense.

Tianhe-3 – Welcome to the Future

In development right now is the Tianhe-3, which will be able to handle one quintillion calculations per second. One quintillion equals one million trillions. That should do it, considering that the supercomputer won’t have to keep up with human bodily functions to boot. But the software needs to catch up, and the affordability needs to as well.

The power it takes for a computer to perform the same calculations as a human brain would light up an entire large building. On the other hand, the human brain can perform the same calculations for the amount of power it takes to light up a 20 watt light bulb. Much more efficient.

Final Thoughts

Perhaps R2D2 is out of reach at the moment. However, this is likely to change in a very short time considering the exponential growth of development in the field. In fact, Intel is boasting that they will achieve this capability by 2018. Nonetheless, most experts think that it will happen by 2020. So artificial general intelligence is not a matter of if, but how many months from now. Do you think that we will develop artificial superintelligence in our lifetimes? Share your thoughts below!

IMAGE SOURCE: 1, 2