Artificial Intelligence (AI) Technology
Artificial Intelligence (AI) is the simulation of human intelligence processes by machines, especially computer systems. Here is a simple overview of how AI functions:
1. Data Collection and Input
Data is the foundation of AI. When we start to teach an AI system, ideally with as much data as possible (images, text or sound). The performance of the system depends on the quality and quantity of data.
Unsupervised Learning: The system identify hidden pattern in the data which is without explicit labels.
Reinforcement Learning: Ai learns through the iterations, by trying again and again, based on feedback — rewards or penalties.
2. Modeling and Training
In the world of AI, some algorithms for developing a model are defined in training data. A model is, at its core, a way to capture the mathematical relationships between all potential inputs and outputs. This is best illustrated using a neural network, which is one of the most popular AI models and leverages layers of interlinked nodes (akin to neurons in the brain) that receive and send messages.
While training, the model performs predictions and compares them to real outcomes — it also adjusts itself to explore errors. This continues till the model becomes accurate enough.
3. Inference and Prediction
We use the AI model to make predictions or decisions on new, unseen data that wasnt used during training. Inference is the process.
An AI system could be trained on images of cats and dogs that then would categoris new images as either a cat or a dog.
4. Optimization and Improvement
Feedback mechanisms, commonly used to facilitate AI systems learning over time. For example, in reinforcement learning the system upgrades its actions according to rewards or punishments. Supervised learning is when models are trained with fresh data to keep their accuracy up.
5. Types of AI Systems
Artificial Specific Intelligence (or Narrow AI): The vast majority of investments made in this field are still for very specific tasks, eg. language translation, image recognition or how to win games like Chess. All of these are narrow AIs; they operate within specific domains.
General AI — Theoretical future artificial intelligence systems, in which the intellectual capacity of a machine is equivalent to (or higher than) human beings. We haven’t achieved this yet.
Key Components:
Neural Networks: Inspired by the human brain, this is a form of deep learning. Layers are groups of nodes where data is preloaded and each layer processes the date then passes that to the next, hence the model makes decisions.
NLP: Natural Language Processing, a subfield of Artificial Intelligence covered by the alt FINS Gazette that deals with enabling machines to understand, interpret and generate human language.
Computer Vision: Common application of AI, it allows machines to interpret and understand visual information ex recognizing faces or objects in images.
The behaviour of AI systems largely depend upon the data they are trained on, architecture (design of the algorithm) and support for training/inference available by computational power.
Types Differently According To Create This Reality Types this according to how it acts and what it does. Capability and functionality (most widely used) — this is the most common classification.
1. Based on Capabilities
This categorization looks at the sophistication or strength of the AI.
a. Narrow AI (Weak AI)
Narrow AI: Narrow AI refers to the type of machine intelligence that plays Go, beats Grandmasters at chess, retrieves mail from a cubby in an office building or read pages out loud. Rather, it has specialized intelligence that can operate within predefined parameters.
Examples:
AI virtual Assistants (Siri, Alexa, Google Assistant)
Netflix or Spotify like Recommendation Systems
Image recognition software
Drawback: It is able to do tasks that are out of his defined scope.
b. General AI (Strong AI)
It means: General AI = From the term, we can say that general intelligence is intelligent in more than one task.
Status Quo: There is no known general AI This a theoretical possibility that researchers are aiming for.
Modified Example: An imagined AI, which has artificial general intelligence (the ability to perform any intellectual task a human can) and is capable of reasoning, learning from experience and understanding emotions.
c. Superintelligence
Meaning: By achieving AI that is greater than human intelligence in virtually every area such as creativity, general knowledge and understanding or social intelligence.
Example: Commonly shown in science fiction( e.g” The Terminator”, ” Ex Machina”), this type of AI would be able to invent for itself without any information beyond what it is given and will then execute the decision using pleanty of other interactions.
When can this be expected: This is still some time in the Artificial Minds future.
2. Based on Functionality
This taxonomy looks at how AI systems are working and learning with respect to human intelligence.
a. Reactive Machines
Single Machine- These are AI systems that can solve tasks based on the current input only and are not capable of storing past information or experiences for use in future decisions.
Features: No memory and no learning.
Examples:
Deep Blue by IBM, the chess-playing machine that beat Garry Kasparov
Artificial intelligence in computer games such as NPCs (characters on the screen) and their responses to user inputs
b. Limited Memory
What its means: These are AI systems that can make decisions not just based on current inputs but also some previously stored information. They are able to learn, through historical data, how they can perform better in itself.
Features: Simple generalization on past data, short-term memory.
Examples:
Self-driving cars: They take a peek into how the other moving objects including humans are on the road, and respond based on past data.
Empirically improving image recognition systems.
c. Theory of Mind AI
What It Is: AI systems able to grasp emotions, beliefs, intentions and thoughts — the way humans sense or predict others’ behaviors.
Objective: Practice Social Interaction and Emotional/Social Context
Current Status: RND stage.
Case in point: a realization of an AI which can empathize with humans, or an expression of emotions through words at the very moment when a human is feeling them.
d. Self-Aware AI
What it means: The most complex AI, one which is yet to be truly realised — capable of consciousness and self-awareness. This would be the ability of AI to act based on its identity and knowledge of the world.
Current Condition: It is completely theoretical; no AI that can realize or train itself in this way exists as of date.
For example: all time favorite Sci-fi stuff like HAL 9000 from “2001 A Space Odyssey” or the AI in movie “Her”.
3. Technique / Approchment Based
This examines the different directions in which AI is developed, and how they perform tasks.
a. Machine Learning (ML)
Machine Learning: The field of study that gives computers the ability to learn without being explicitly programmed.
Techniques:
Supervised Learning: Based on training with labeled data.
Unsupervised Learning: In this type the model tries to learn patterns in the data.
Type 3 Reinforcement Learning-The model interacts with the environment and receives feedback (in terms of rewards or penalties)
Examples:
Spam filters in email
Credit card fraud detection
b. Deep Learning
Definition: A machine learning framework employing artificial neural networks of interpolative layers (hence “deep”) to model high-level abstractions in data. It will work very well for things like image and speech recognition.
Examples:
Similarly, Image classifiers such as Google Photos
Speech recognition which uses deep learning with voice assistants like Google Assistant
c: Natural Language Processing (NLP)
Natural Language Processing definition: A branch of AI that helps computers understand, interpret and communicate human language.
Examples:
Chatbots/Virtual Assistants
E.g., Google translate
d. Robotics
Meaning: Applying AI to both the design and functions of the robot itself. AI also learns human-like behavior with robots who are able to perform every real world task.
Examples:
Autonomous drones
E.g. Roomba, Robotic Vacuum cleaners
e. Expert Systems
Definition: AI which incorporates a human expert knowledge base and an inference engine to solve problems based on these bits of information.
Examples:
Medical diagnosis systems
Stock Trading Investment Financial Systems
Summary:
Today, the type of AI that is commonly and successfully used is called narrow AI which only does one thing.
Human-like — has all the corresponding AGI features, but note that this one does not exist yet.
A superintelligent AI is a future in which the artificial intelligence i.e. machines exceed all human intelligence or equivalent to mankind thinking ability.
This means any AI from your washing machine to sentient robots (which are still only theoretical).
