We are further down the road to A.I. As we grow in understanding, so, too, do we grow in understanding its differences.
In 2020, we can classify artificial intelligence into 4 different categories. The categories are loosely similar to Maslow's hierarchy of needs, where the simplest level requires only basic functioning and the highest level is Muhammad, Buddha, Christian saint, omniscient, all-seeing, Self-awareness is consciousness.
Four A.I. There are types
1 Reactive machines
2 Limited memory
3 Theory of Mind
4 Self-aware
Reactive machines
Reactive machines perform basic functions. This level of A.I is the easiest. These types react to certain inputs with certain outputs. There is no learning that happens. This is the first step in any A.I. A machine learning system that takes a human face as input and outputs a box around the face to identify it as a face is a simple, reactive machine. A model does not store any input, it does not do any learning.
Static machine learning models are reactive
machines. Their architecture is the simplest and can be found in GitHub repos
all over the web. These models can be easily downloaded, traded, passed around
and loaded into a developer's toolkit.
Limited memory
Limited types of memory refer to the AI's ability to store previous data and/or predictions, using that data to make better predictions. With limited memory, the machine learning architecture becomes a bit more complex. Each machine learning model requires limited memory to build, but the model can be deployed as a type of reactive machine.
There are three major categories of machine learning models that capture this type of limited memory:
Reinforcement learning
These models learn to make better predictions through several cycles of trial and error. This type of model is used to teach computers how to play games like chess, Go, and DOTA2.
Long Short Term Memory (LSTMs)
The researchers reasoned that past data would
help predict the sequence of next items, especially in language, so they
developed a model called long-short-term memory. To predict the next elements
in a sequence, LSTM tags recent information as more important and past items as
less important.
Evolutionary Generative Adversarial Networks (E-GAN)
The memory of E-GAN is such that it evolves on every evolution. The model produces a type of growing object. Growing things don't follow the same path every time, the paths are slightly modified because statistics is the math of chance, not the math of accuracy. In modification, the model can find a better path, the path of least resistance. The next generation of models changes and evolves along the path that its ancestors found in error.
In a way, E-GAN creates a simulation of the
evolution of humans on this planet. Each child, in perfect, successful
reproduction, is better equipped to lead an extraordinary life than its
parents.
Practically limited types of memory
Although every machine learning model is created using limited memory, they are not always built that way when deployed.
Limited memory A.I. Works in two ways:
A team continuously trains a model on new data.
A.I. The environment is designed in such a way
that models are automatically trained and updated based on model usage and
behavior.
For a machine learning infrastructure to
maintain a limited memory type, the infrastructure needs to incorporate machine
learning into its architecture.
Theory of Mind
We have yet to get to Theory of Mind types of artificial intelligence. These are only in their early stages and can be seen in things like self-driving cars. In this type of A.I., A.I. Interactions begin with human thoughts and emotions.
Currently, machine learning models do much more
than guide a person to achieve a task. Current models have a one-way
relationship with A.I. Alexa and Siri bow to every command. If you angrily yell
at Google Maps to take you in another direction, it doesn't provide emotional support
and say, "This is the fastest direction. Who do I call to tell you that
you're going to be late?" ?" Google Maps, instead, continues to
return the same traffic reports and ETAs
Self-aware
Eventually, in some distant future, perhaps
A.I. attains Nirvana. It becomes self-aware. This type of A.I. exists only in
the story, and as stories often do, creates immense hope and fear in the
audience. A superhuman self-aware intelligence has an independent intelligence,
and it is likely that humans will have to negotiate terms with the entity it
created. What happens, good or bad, is anyone's guess.
0 Comments