Why Define Artificial Intelligence Spectrum?
It’s not easy to do for the field is fluid and the interpretations many. I define the Artificial Intelligence spectrum so that future references are readily understood as this site progresses. As references are made I will keep that and the following in mind.In Artificial Intelligence (AI) – 6 Things To Know For Now I wrote that Artificial Intelligence (AI) is a branch of computer science. The aim of which is to create hardware and software capable of thinking intelligently. Artificial Intelligence is the science of making machines do things that normally require human intelligence. My goal for this post was to describe a coherent understanding of how Artificial Intelligence. It will be a useful reference as the site identifies the leaders in Artificial Intelligence.
The AI Spectrum
AI is being used prolifically in various media and the trend is to classify anything that has some smartness involved as AI. But AI is more than just a catch phrase when interpreted correctly. Is it simply smart technology? Think smart phones, smart cars, smart Fitbits. The answer is yes and not really. To understand that response it is necessary to understand the AI spectrum.
Applied Artificial Intelligence
Applied Artificial Intelligence is at the left of the spectrum. This is the AI we live with today in many applications. Not just on the phone but on any device that can benefit from the integration of data.The apps on your phone such as Siri, Google Maps, security recognition, etc. are examples of Applied AI in everyday use. Applied AI is all around us. For example, auto pilot on planes which control most except for 7 – 20 minutes of flight.
Applied AI are machines equipped with the programs and algorithms that conduct high repetition high frequency tasks better than humans. Driving a car, reading a map, flying a plane, sailing a boat, cutting grass, etc. are examples of how we use Applied AI. They require human interaction but appear to do complicated tasks on their own. This is where the action is happening today and one day may lead to breakthroughs that allow for machines to manage humanity and all of earth’s resources. Advances in Applied AI will provide clues as to if and when Strong AI will emerge.
General Artificial Intelligence
Beyond our knowledge of existence today there is an elusive future known as General AI. These are machines that are indistinguishable from human intellectual capability. Examples of strong AI are not available at the moment and won’t be for a long time, if ever. Science fiction and unproven theories offer glimpses into what general AI may look like or be capable of doing. Imagine all past human knowledge combined with all present human experience embedded in the Cloud to provide answers. Possibly answers to everything. General AI is elusive and not universally agreed upon on as likely to occur or even a goal that humans should endeavor to pursue.
Four Types of Artificial Intelligence
Spread across the Artificial Intelligence Spectrum are four types of Artificial Intelligence that will help serve the purpose of this website. As I proceed with this site I will refer to companies and leaders in the field and attempt to map them along the spectrum. That way I may be able to compare definitions of Artificial Intelligence in a coherent manner.
I. Reactive machines
These are the most basic type of Artificial Intelligence. Deep Blue is one example of this type of machine at the far left of the Artificial Intelligence Spectrum. This was the machine that beat Chess Champion Gary Kasporov. It was only to do so because the programmers learned to limit the possibilities that Deep Blue could review. Without this limit the games would have turned out differently and most likely in Kasporov’s favor.
Reactive Machines seem to be remarkable in their capacity to execute actions based on limited predictions. However, they cannot call on past experiences to make decisions. From a choice of reviewed possibilities reactive machines can play games well but they cannot call on past moves to decide future moves.
II. Limited memory
Still at the far left of the Artificial Intelligence spectrum are smart machines that can draw on the past for actions in the present. Ever wonder how spam mail is sorted to the spam folder? Think programmed email filters. They can learn with the use of artificial neural networks to filter email and make our lives a little less time constrained. It saves us from the rudimentary work that would be required of us to send spam to the trash. In much the same way Google’s driverless cars, and speech recognition machines like Siri, OK Google present some of the characteristics of limited memory.
However clever they may seem at the moment there are limitations for they cannot store experiences as humans do and draw on them later for action. Driving down the road a driverless car may recognize and reactive to the myriad of traffic signals and signs and avoid accidents. They cannot accurately predict what other cars may do in the way that experienced have learned to drive defensively with distracted drivers.
III. Theory of mind
This type of Artificial Intelligence lies at a point on the Artificial Intelligence Spectrum that has not yet been achieved. It is not currently available but it is the bridge between the smart technologies of today and self awareness machines. Theory of mind refers to understanding the motivations and resulting behaviors of humans, animals and objects in the world. We humans know we have this capacity to discern the world around us based on experience and understanding how people and things work and most important why they work in the manner they do. We tend to not grant this ability to animals because that would imply self awareness.
As humans it is natural to want to keep this a human trait. If however neural networks can envelope greater amounts of data than currently possible the day may come when machines can form societies among themselves. This may begin as a survival mechanism similar to how humans learned to live together to fend off the elements and common enemies. I hope that if this theory of mind develops in machines they would also have an understanding of how and why humans are to be a mutual part of their existence.
IV. Self Awareness
As the name implies self awareness is the ability to recognize how one recognizes their place in the world and the Artificial Intelligence Spectrum. Many of things we do on a daily are performed without consideration for why they are done. We all have particular habits, routines and schedules that are pursued relentlessly and most often subconsciously. To be able to achieve self-awareness one must be open to review and at times improve our daily lives. Our daily decision making processes are influenced by culture and background. These influences often explain why an activity is performed and what we achieve from it. In many instances the reasoning for our actions may elude logic but that is the evolution of human nature.
Self awareness on the Artificial Intelligence Spectrum is not necessarily limited to what we have learned to survive as long as we have. Self awareness and explaining why we think the way we do on certain issues is not the same. Although we may think that is the case. There are myriad reasons for explaining human behavior versus understanding how one perceive their place in the world.
Perfect self awareness would allow for us to pursue what is in the best interest of the world. Self awareness informs us of where we come from and where we are going. Lots of people possess this capability but not all. The elusive future that could be General Artificial Intelligence would consist of machines with consciousness. This would lead to the creation of rights bestowed on machines. Sound familiar? It is what humans have been striving for since civilization began.
Notes and References
- Arend Hintze, Michigan State University, The Conversation
- 33 Types Of Artificial Intelligence
- 3 Types of Artificial Intelligence Everyone Knows About
- Jeff Smith, Reactive Machine Learning Systems