|
Key term: Inductive reasoning. Drawing on information from experiences
and our environment to draw generalizations.
The ability to see the difference between these pictures based on our
knowledge of cats and dogs; this is what we know as reasoning. The goal of
artificial intelligence is to teach computers how to have similar abilities to
human-like reasoning
Computers models have been utilized to process Natural languages. Natural
language processing gives computers the ability to understand ‘natural’
languages, or what we know as our human languages. Natural language
processing relies on machine learning techniques to understand speech and
text and respond to commands and interactions.
This technology is becoming very common and accessible. GPUs (graphics
processing units) are becoming more widely available and less expensive,
which means data sets are getting bigger, and the use of machine learning is
expanding. You may have used it when you talk to Siri on your iPhone.
When you say something to Siri, your phone receives the audio. To interpret
it, it turns your audio into text. Then your phone analyzes the text to derive
meaning from the command that you gave it.
Natural language processing is one of the most common applications of
machine learning, and we use it every day. When we use a search function
on the internet, we are using natural language processing. Translation apps
must take our voice or our text and analyze the sentence structure to make
meaning. When you type up a paper or a word document, your word
processor uses natural language processing to sift for grammatical errors
and spelling mistakes.
Despite its popularity, it’s a very complex field of computer science and
artificial intelligence. Being able to interpret the meaning of the alphabet
arranged in a virtually infinite amount of combinations requires huge
amounts of data for the computer to understand what you are writing or
saying.
In addition to being able to understand what we say and write; computers
can also make strategic decisions based on what they’ve learned from data
in the 1990s. IBM created a computer called Deep Blue that defeated a
world master at chess. It was the first computer to be able to perform such a
task. Because of the simplicity of the rules in chess, computer scientists at
IBM chose to train their computer to play. But there are thousands of
potential moves and arrangements that the pieces can take once the game
has started. The computer had to learn this using data.
What makes machine learning unique to other types of computer science is
the ability of models to change their methods over time to suit new data.
What separates a machine learning model from a regular line of explicit
code is that machine learning will take in new data and improve itself. It
can also perform tasks which require planning and contain strategic
components. The Deep Blue computer had to be adept at analyzing possible
sequences of moves, rather than just one move at a time.
The same technologies that enabled a computer to beat a world chess
champion are now making it possible for self-driving vehicles to get a
passenger safely from point A to point B. Compared to the relative
simplicity of chess, self-driving cars must plan and interpret hundreds of
variables to keep its passenger safe. It goes beyond the two-dimensional
data analysis employed by machines playing chess. Self-driving cars must
master multidimensional data analyses to navigate the everchanging
environment on the road.
The machine learns through trial and error, repeating the task over and over
and learning from failures and successes. These experiences are introduced
as data, and over time, the machine will know its probability of failure or
success for every possible move.
Machine learning models interpret potential states in the environment. For
an algorithm that plays chess, this is all its possible moves and all its
competitor's possible moves. The algorithm is an amalgamation of goals
and potential actions. By using this data, it creates a plan to optimize the
likelihood of achieving the goals. It also enables computers to do self-
learning without specific directions through programming.
Trying to get a computer to do all these things sounds simpler in theory than
it is in practice. Most of the functions we’ve just mentioned; from checkers
to self-driving cars require advanced statistical techniques to optimize the
outcome and train a machine that knows how to ‘win’ with a high level of
accuracy.
Machine learning falls under the larger umbrella of artificial intelligence.
Artificial intelligence is a branch of computer science that includes
reasoning, natural language processing, planning, and machine learning.
The term was first coined by a computer scientist named John McCarthy in
1956. You’ll also hear the term data science, which encompasses artificial
intelligence and machine learning. Data science is a broader term, but it’s
often used to describe machine learning. Machine learning experts are often
referred to as data scientists, both in this book and beyond. There is overlap
between data science and machine learning, but they are not the same thing.
Data science is more of a general term, whereas machine learning is a part
of data science.
Do'stlaringiz bilan baham: |
