The Data Daily

What is Artificial Intelligence (AI)? - SDxCentral

Last updated: 02-14-2020

Read original article here

What is Artificial Intelligence (AI)? - SDxCentral

Artificial intelligence (AI) is the science of making machines capable of reasoning and learning new concepts. It is related to and takes advantage of machine learning, but shouldn’t be confused as the same thing. Narrow AIs that exist today can accomplish only very specific tasks and are limited in scope. General AI, which is what people tend to think of as AI, does not exist yet and has raised many concerns in the scientific community and the public.

Narrow AIs have intelligence in a single specialized area, according to the book “Artificial General Intelligence,” edited byCassio Pennachin andBen Goertzel. Unlike general AIs, narrow AIs solve subproblems (smaller elements of a larger problem) in isolation instead of working in a larger framework.

A general AI is “a software program that can solve a variety of complex problems in a variety of different domains, and that controls itself autonomously,” according to “Artificial General Intelligence.” They are also sometimes described ashaving the intelligence of an average human. 

General AIs have yet to be invented, as such, it is impossible to say what the full extent of general AIs’ cognitive abilities will be, though there is speculation as to how they can be used, for better or worse. 

Possible benefits include accurate and fast medical diagnosis and safe autonomous vehicles. However, there are concerns about widespread job loss to AI and humans becoming overly dependent on AIs for cognitive tasks. These subjects are delved into more below.

There are five foundational elements of AI. An AI must be capable of:

AIs are often aided by visual and audio inputs that connect them to the outside world, such as with an autonomous vehicle’s cameras or the microphone of a language recognition system.

AIs are currently used to automate repetitive computational tasks, such as analysis and correlation of data from industrial Internet of Things (IIoT) devices in a factory. In retail, AIs can also be used to offer shoppers personalized product recommendations based on their behavior, or they can be used to track items in stock to ensure items aren’t sold out based on sales metrics. 

The 2019 book, “Introduction to Artificial Intelligence: Third Edition” by Phillip C. Jackson identifies three architectural levels within an AI: linguistic, archetype, and associative levels. 

The linguistics level is where the AI represents information with one or more symbolic languages. The languages can be used to make inferences on the data the AI collects. 

The archetype level is where categories, classes, or types of concepts, images, and objects are located for the machine to recognize. The categories should be used to help the AI system process the linguistic level. 

The associative level recognizes and processes information from the environment. For example, physical objects can be sorted into different classes or speech can be recognized as words and sentences.

Alternatively, “Advanced Artificial Intelligence (Second Edition)” by Shi Zhongzhi divides AI into two categories: symbolic intelligence and computational intelligence.

Symbolic intelligence uses thephysical symbol systemfor studying the processes for knowledge representation, acquisition, and reasoning. That knowledge is used to solve problems and is the current symbol of intelligence because it shows the AI has acquired information, used that information to reason, and made a decision free from human direction. 

Machine learning is a subset of AI. It is a more focused and less conceptual field. Machine learning is a process where a program is given datasets and uses of algorithms and statistical models to find patterns.

Machine learning programs are taught through supervised learning, semi-supervised learning, and unsupervised learning.

In supervised learning, the program is given datasets that have already been classified and the program learns the associations between them. 

In semi-supervised learning, the program has to fill in gaps of information based on contextual information. 

In unsupervised learning, the program has to make correlations between datasets that have no contextual information.Machine learning can be used by an AIto accomplish the tasks it is given.

AI can be used in awide variety of fields, from finance to healthcare to networks to transportation. In finance, it can be used for better stock trading at faster speeds. In healthcare, it can be used for improved diagnosis, though it isstill in the early stages. In networks like 5G, AI can be used tooptimally use the network’s resources. In transportation, AI will be heavily used for autonomous vehicles toensure safe driving.

When the term was coined in 1956, AI was mostly relegated to academic research. AI programs accomplished tasks likelearning checkersorsolving algebra word problems. Research continued into the 1990s — in some cases withfunding from the United State’s Department of Defense— at which point AI was beginning to be used for more complex tasks. 

AI was used for logistics, data mining, medical diagnosis, and even beat a world chess champion in 1997. Around the turn of the century, machine learning research became fruitful. It was and continues to be used by search engines like Google and large websites like Amazon.

The classic and well-known test for whether an AI is actually intelligent is the Turing Test. To pass the test, a machine must communicate with a human via text and not have the human suspect it is in fact a machine. 

The practical implementations of AI currently are largely relegated to machine learning and deep learning (essentially advanced machine learning using neural networks). AI has a lot of potential to be used in almost every part of society — particularly whena general AI, as mentioned above, is invented. 

Concerns have been raisedover the creation and use of general AIs. Examples include job loss, human agency, and a dependency on AI-enabled devices. Potential ways to avoid those negatives and others is through global participation in AI innovation, the use of a values-based system where AIs are given the ability to have empathy, to prioritize people, and prevent their replacement.

Read the rest of this article here