What is artificial intelligence or AI?

What is artificial intelligence or AI?

Artificial intelligence (AI) is a set of technologies that enable computers to perform a variety of advanced functions, including the ability to see, understand, and translate spoken and written language, analyze data, make recommendations, and much more.

AI is the backbone of innovation in modern computing, helping to automate processes and analyze large data sets, creating value for people and businesses. A wide variety of AI use cases are emerging, from robots that can navigate a warehouse on their own, to cybersecurity systems that are continually analyzed and improved, and virtual assistants that can understand what people are saying and act on it. on that information. Machine learning (ML) is a particularly important subset of AI in which machines build models based on training data, typically to generate more accurate predictions.

To learn more about machine learning, read about Google Cloud AI products , including Vertex AI, Google’s unified platform for training, deploying, and managing machine learning models; APIs like Speech-to-Text ; CCAI for the automation of call centers and virtual agents; and solutions for industries such as healthcare, lending, and other verticals.

Everything ready to start? New customers get $300 in free credits to use on Google Cloud.

Definition of artificial intelligence

Artificial intelligence is a field of science related to the creation of computers and machines that can reason, learn, and act in a way that would normally require human intelligence or that involves data whose scale exceeds what humans can analyze.

AI is a broad field that encompasses many different disciplines, including computer science, data analysis and statistics, hardware and software engineering, linguistics, neuroscience, and even philosophy and psychology.

At the operational level for business use, AI is a set of technologies that are primarily based on machine learning and deep learning, used for data analysis, generation of predictions and forecasts, object categorization, natural language processing, recommendations, intelligent data retrieval and much more.

Types of artificial intelligence

Artificial intelligence can be organized in various ways, depending on the stages of development or the actions being performed.

For example, four stages of AI development are generally recognized.

  1. Reactive Machines: Limited AI that only reacts to different types of stimuli based on pre-programmed rules. It does not use memory and therefore cannot learn on new data. IBM’s Deep Blue, which beat chess champion Garry Kasparov in 1997, was an example of a reactive machine.
  2. Limited Memory: Most modern AI is considered to be memory limited. You can use memory to improve over time by training on new data, typically through an artificial neural network or some other training model. Deep learning, a subset of machine learning, is considered artificial intelligence with limited memory.
  3. Theory of Mind: There is currently no AI with theory of mind, but different possibilities are being investigated. The term refers to AI that can emulate the human mind and has human-like decision-making capabilities, including recognizing and remembering emotions, and reacting in social situations as a human would.
  4. Self-Awareness: A step beyond theory-of-mind AI, the concept of self-awareness AI describes a mythical machine that is aware of its own existence and has the intellectual and emotional capabilities of a human being. Like AI with theory of mind, AI with self-awareness does not currently exist.
    A more useful way to broadly categorize AI types is by what the machine can do. Everything we call artificial intelligence today is considered “narrow” intelligence because it can only perform a narrow set of actions based on its programming and training. For example, an AI algorithm used for object classification will not be able to perform natural language processing. Google Search is a form of narrow AI, just like predictive analytics or virtual assistants.

Artificial general intelligence (AGI) would be the ability of a machine to “feel, think and act” as a person would. The AGI does not currently exist. The next level would be artificial super intelligence (ASI), in which the machine could function superior to the human in every way.

Artificial intelligence training models

When companies talk about AI, they often talk about “training data.” But what does that mean? Remember that AI with limited memory is an AI that improves over time as it is trained on new data. Machine learning is a subset of artificial intelligence that uses algorithms to train data and get results.

Broadly speaking, three types of learning models are commonly used in machine learning:

Supervised learning is a machine learning model that maps a specific input to a result using labeled training data (structured data). In simple terms, to train an algorithm to recognize images of cats, you feed it images tagged as cats.

Unsupervised learning is a machine learning model that learns patterns based on unlabeled data (unstructured data). Unlike supervised learning, the end result is not known in advance. Instead, the algorithm learns from the data and classifies it into groups based on various attributes. For example, unsupervised learning is good at identifying patterns and doing descriptive modeling.

In addition to supervised and unsupervised learning, a mixed approach called semi-supervised learning is often used, in which only some of the data is labeled. In semi-supervised learning, an end result is known, but the algorithm must determine how to organize and structure the data to achieve the desired results.

Reinforcement learning is a machine learning model that can be loosely described as “learning by doing”. An “agent” learns to perform a defined task through trial and error (a reaction cycle) until its performance is within a desired range. The agent receives positive reinforcement when he performs the task correctly and negative reinforcement when he performs poorly. An example of reinforcement learning would be teaching a robotic hand to pick up a ball.

Common Types of Artificial Neural Networks

A common type of AI training model is an artificial neural network, which is roughly based on the human brain.

A neural network is a system of artificial neurons (sometimes called perceptrons), which are processing nodes used to classify and analyze data. Data is fed into the first layer of a neural network, and each perceptron makes a decision, then passes that information to various nodes in the next layer. Training models with more than three layers are called “deep neural networks” or “deep learning”. Some modern neural networks have hundreds or thousands of layers. The output of the final perceptrons allows the task imposed on the neural network to be carried out, such as classifying an object or finding patterns in the data.

These are some of the most common types of artificial neural networks you can find:

Feed – forward (FF) neural networks are one of the oldest forms of neural networks, as data flows in one direction through layers of artificial neurons until the result is obtained. Today, most feedforward neural networks are considered “deep feedforward” with multiple layers (and more than one “hidden” layer). Feedforward neural networks are often tied to an error correcting algorithm called “back propagation” which, in simple terms, starts with the output of the neural network and works backwards to the beginning, detecting errors to improve accuracy. of the neural network. Many simple but powerful neural networks are deep feedforward.

Recurrent neural networks (RNNs) differ from feedforward neural networks in that they typically use time-series data or data that involves sequences. Unlike feedforward neural networks, which use weights at each node in the network, recurrent neural networks have “memory” of what happened in the previous layer as contingent on the output of the current layer. For example, when performing natural language processing, RNNs can “take into account” other words used in a sentence. RNNs are often used for speech recognition, translation, and image description generation.

Long/Short Term Memory (LSTM) RNNs are an advanced form of RNN that can use memory to “remember” what happened in previous layers. The difference between RNNs and LTSMs is that they can remember what happened several layers ago by using “memory cells”. The LSTM is often used for speech recognition and predictions.

Convolutional Neural Networks (CNN) include some of the most common neural networks in modern artificial intelligence. CNNs are often used in image recognition and use several different layers (a convolutional layer and then a pooling layer) that filter different parts of an image before putting it back together (in the fully connected layer). It is possible that the convolutional layers above look for simple features of an image, such as colors and edges, before looking for more complex features in additional layers.

In generative adversarial networks (GANs) , two neural networks are used to compete with each other in a game that ultimately improves the accuracy of the result. One network (the generator) creates examples that the other network (the discriminant) judges to be true or false. GANs have been used to create realistic images and even make art.