When we think of artificial intelligence, the first thing that comes to mind for many people are sentient robots that we created that exist alongside us, sometimes peacefully and sometimes not so peacefully depending on your viewpoint on robots. But we are a far ways away from that becoming a reality. In reality, what we have accomplished so far with artificial intelligence is still nothing short of remarkable. Researchers have divided artificial intelligence into four different categories, Type 1 to Type 4. As it currently stands based on these categories we are only partially thru type 2 and we are still some ways out from reaching the criteria that defines a Type 3 Artificial Intelligence.
Type 1 artificial intelligence is defined as reactive machines. How these work is they take inputs and cues from data they are given or data that is gathered based upon what the computer can perceive. Some real-world examples of this that many of us have probably heard of are IBM’s Watson and Google’s DeepMind/AlphaGo. These operate on a form of artificial intelligence called neural networks. These are essentially computer “brains” that form from a program executing a variety of different scenarios until it eventually forms “neurons” that it remembers produces a certain result. Below is a video of one of these neural networks in action.
Type 2 artificial intelligence is defined as machines with limited memory. While Type 1 AI operates by looking at things at that current moment, Type 2 AI is capable of referencing things that it has stored as well as look at current things happen to decide what action is the best action to take. The best example of this that is out there today is self-driving cars. Tesla has been able to take this concept and figure out a way to make it work in the real world with fantastic results. However, there is still difficulty in trying to figure out to make these machines remember what they’ve learned for any prolonged period of time which is the sticking point with getting us thru Type 2 AI and into Type 3 AI.
Type 3 Artificial Intelligence
Type 3 artificial intelligence is defined as theory of mind. This type of AI deals technically with machines and technology that hasn’t been built quite yet. In psychological terms, theory of mind is how thoughts and emotions can play as a determinant as what a person, animal, or object behaves after the fact. If in the future, for example, if robots with AI are to interact and be in an environment with people, they’ll have to be able to study and understand human behavior, and be capable to adjust their thought process as to how to treat that specific behavior or action. It’s essentially like giving a brain to the machine, so it can have its own thought process.
Type 4 Artificial Intelligence
Type 4 artificial intelligence is defined as Self-awareness. This type is somewhat an extension of type 3’s “theory of mind”. It’s all about building a system that is aware of itself and has a conscious. The system is smart enough and has enough internal thought to be able to know what state it is in and being to predict the feelings and emotions of people. For example, if we see someone smile, it’s most likely because they are happy. We know this because we smile when we are happy too. A very elementary fact, but that is the basis for being self-aware. It’s like the same thing when Arizona Basketball loses, and we see people not speaking and heavy hearted because they’re sad. We know how they feel because we feel the same way. An important step towards creating systems that are self aware is being able to retain and learn from memory, essentially basing decision or action on what has happened in past experience.
Sources:
- Arend Hintze Assistant Professor of Integrative Biology & Computer Science and Engineering, Michigan State University. (2017, April 14). Understanding the four types of AI, from reactive robots to self-aware beings. Retrieved April 19, 2017, from http://theconversation.com/understanding-the-four-types-of-ai-from-reactive-robots-to-self-aware-beings-67616
- Copeland, M. (2016, July 29). The Difference Between AI, Machine Learning, and Deep Learning? | NVIDIA Blog. Retrieved April 12, 2017, from https://blogs.nvidia.com/blog/2016/07/29/whats-difference-artificial-intelligence-machine-learning-deep-learning-ai/
- Deep learning. (2016, October 10). In Wikipedia, The Free Encyclopedia. Retrieved 06:04, April 10, 2017, from https://en.wikipedia.org/w/index.php?title=Deep_learning&oldid=743563119
- (2016, October 10). In Wikipedia, The Free Encyclopedia. Retrieved 04:50, April 10, 2017, from https://en.wikipedia.org/w/index.php?title=DeepMind&oldid=743552789
- Go (game). (2016, October 12). In Wikipedia, The Free Encyclopedia. Retrieved 18:21, April 12, 2017, from https://en.wikipedia.org/w/index.php?title=Go_(game)&oldid=744033064
- Machine learning. (2016, October 10). In Wikipedia, The Free Encyclopedia. Retrieved 06:29, April 10, 2017, from https://en.wikipedia.org/w/index.php?title=Machine_learning&oldid=743566912
- Computational theory of mind. (2016, December 4). In Wikipedia, The Free Encyclopedia. Retrieved April 19, 2017, from https://en.wikipedia.org/wiki/Computational_theory_of_mind
Hey guys, great presentation!
ReplyDeleteThat video on video game learning with Mario is really amazing, and is a great introduction into how machine learning works. I wonder what other games you might be able to scale this to, as the original Mario is one of the simplest games out there.
How long until we see Tier 4 AI being commonplace in society, and are you guys worried about the movie Terminator becoming a real possibility?
The Mario example really drives home current uses of AI when it comes to processing data and executing actions based on that data- however, beat Rainbow Road on hard and then I'll truly be impressed.
ReplyDeleteI wonder what are the limits that we as humans will develop AI. Although Big Hero 6 is just a film (great one- if i may add) I can really see Type 3 AI being used in hospitals or in impoverished areas to treat patients. In addition, this form of AI can also be used as counselors for people with PTSD. However, things begin to get scary in the next level. I don't wan't my x-box to become self aware and tell me that me not scoring a goal on FIFA was due to my own lack of skill and not that it was bullshit.