AI

• ANI (Artificial Narrow Intelligence) or Waek AI - e.g., smart speaker, self driving car, web search, …

• AGI (Artificial General Intelligence) or Strong AI - do anything a human can do

1. Machine Learning: A field focused on developing algorithms that enable computers to learn and make decisions from data, without being explicitly programmed for specific tasks.

2. Computer Vision: Concerned with enabling computers to interpret and process visual information from the world, such as image and video analysis.

3. Natural Language Processing (NLP): Focuses on the interaction between computers and human language, enabling computers to understand, interpret, and generate human language.

4. Robotics: Integrates AI with mechanical engineering, electrical engineering, and computer science to create robots that can perform tasks autonomously or assist humans.

5. Data Science: Involves extracting insights and knowledge from data through various techniques, combining aspects of statistics, machine learning, and data analysis.

6. Expert Systems: A branch of AI that deals with systems that emulate the decision-making ability of a human expert, using a set of rules and knowledge base to solve specific problems within a domain.

Machine Learning:

• Supervised Learning: Learning from labeled data (classification, regression).

• Unsupervised Learning: Finding patterns in unlabeled data (clustering, association).

• Reinforcement Learning: Learning based on feedback from interactions with an environment.

• Neural Networks: Systems inspired by the human brain, used to recognize patterns and solve various types of problems.

Computer Vision:

• Image Recognition: Identifying objects, people, places, and actions in images.

• Object Detection: Locating objects within an image.

• Image Segmentation: Dividing an image into multiple segments or pixels.

• Motion Analysis: Analyzing and understanding movement within videos.

Natural Language Processing (NLP):

• Text Classification: Categorizing text into predefined categories.

• Sentiment Analysis: Identifying and categorizing opinions in text.

• Machine Translation: Automatic translation of text or speech from one language to another.

• Speech Recognition: Transcribing and understanding spoken language.

Robotics:

• Autonomous Vehicles: Robots capable of navigating without human intervention.

• Humanoid Robotics: Robots designed to resemble and interact like humans.

• Industrial Robotics: Automated, programmable robots used in manufacturing and production.

• Swarm Robotics: Systems involving multiple robots working collaboratively.

Data Science:

• Predictive Analytics: Using statistical algorithms and machine learning techniques to identify the likelihood of future outcomes.

• Data Mining: Extracting patterns from large datasets.

• Big Data Analytics: Techniques to analyze extremely large datasets.

• Data Visualization: Representing data in graphical or visual format.

Expert Systems:

• Rule-Based Systems: Systems that use predefined rules to make decisions or solve problems.

• Decision Support Systems: Providing support for decision-making activities.

• Diagnostic Systems: Systems used for diagnosing problems, especially in complex scenarios like medical diagnosis.

• Knowledge Engineering: The process of creating rules and knowledge bases for expert systems.

1. Basic Concept:

   • Mimic the Brain: Neural networks are computing systems vaguely inspired by the biological neural networks in human brains.

   • Neurons: They consist of interconnected units or nodes (analogous to biological neurons) that process information using a connectionist approach.

2. Structure:

   • Layers: Typically organized in layers (input layer, one or more hidden layers, and an output layer).

   • Connections: Each connection, like the synapses in a biological brain, can transmit a signal from one neuron to another.

   • Weights: Each neuron's output is computed using a weighted sum of its inputs, followed by a non-linear function.

3. Types of Neural Networks:

   • Feedforward Neural Networks: The simplest type where the connections between the nodes do not form a cycle.

   • Convolutional Neural Networks (CNNs): Primarily used in image recognition and processing, these networks employ a mathematical operation called convolution.

   • Recurrent Neural Networks (RNNs): Suitable for sequence data like time series or natural language, with connections forming cycles.

   • Deep Neural Networks: Neural networks with multiple hidden layers, capable of learning complex patterns.

4. Learning Process:

   • Training: Involves adjusting the weights of the connections based on the error of the output compared to the expected result.

   • Backpropagation: A common method used for training, especially in deep learning, where the error is propagated back through the network to adjust the weights.

   • Activation Functions: Functions like Sigmoid, ReLU (Rectified Linear Unit), and Tanh that help the network learn complex patterns.

5. Applications:

   • Image and Speech Recognition: Exceptional performance in recognizing patterns in images and audio.

   • Natural Language Processing: Used in understanding and generating human language.

   • Predictive Analytics: Forecasting trends and patterns in data.

   • Autonomous Systems: Empowering self-driving cars, drones, and other autonomous systems.

6. Challenges:

   • Requires Large Datasets: Generally, the more data available, the better a neural network can learn.

   • Computationally Intensive: Training can require significant computational resources, especially for deep learning models.

   • Interpretability: Often referred to as "black boxes" because understanding the internal workings and decision-making can be challenging.

1. Basic Concept:

   • Mimic the Brain: Neural networks are computing systems vaguely inspired by the biological neural networks in human brains.

   • Neurons: They consist of interconnected units or nodes (analogous to biological neurons) that process information using a connectionist approach.

2. Structure:

   • Layers: Typically organized in layers (input layer, one or more hidden layers, and an output layer).

   • Connections: Each connection, like the synapses in a biological brain, can transmit a signal from one neuron to another.

   • Weights: Each neuron's output is computed using a weighted sum of its inputs, followed by a non-linear function.

3. Types of Neural Networks:

   • Feedforward Neural Networks: The simplest type where the connections between the nodes do not form a cycle.

   • Convolutional Neural Networks (CNNs): Primarily used in image recognition and processing, these networks employ a mathematical operation called convolution.

   • Recurrent Neural Networks (RNNs): Suitable for sequence data like time series or natural language, with connections forming cycles.

   • Deep Neural Networks: Neural networks with multiple hidden layers, capable of learning complex patterns.

4. Learning Process:

   • Training: Involves adjusting the weights of the connections based on the error of the output compared to the expected result.

   • Backpropagation: A common method used for training, especially in deep learning, where the error is propagated back through the network to adjust the weights.

   • Activation Functions: Functions like Sigmoid, ReLU (Rectified Linear Unit), and Tanh that help the network learn complex patterns.

5. Applications:

   • Image and Speech Recognition: Exceptional performance in recognizing patterns in images and audio.

   • Natural Language Processing: Used in understanding and generating human language.

   • Predictive Analytics: Forecasting trends and patterns in data.

   • Autonomous Systems: Empowering self-driving cars, drones, and other autonomous systems.

6. Challenges:

   • Requires Large Datasets: Generally, the more data available, the better a neural network can learn.

   • Computationally Intensive: Training can require significant computational resources, especially for deep learning models.

   • Interpretability: Often referred to as "black boxes" because understanding the internal workings and decision-making can be challenging.

1. Understanding context: While generative AI models like GPT-3 can create grammatically correct and contextually relevant responses, they don't truly "understand" the text in the way humans do. They are essentially pattern-matching algorithms hat have learned to predict, based on the data they were trained on, what comes next in a sequence of text.

2. Lack of common sense: Because generative AI models learn from data, they don't possess innate human knowledge or common sense unless it was present in those training data. For example, AI models might not inherently understand that an elephant cannot fit inside a car, unless they've seen similar information in the data they were trained on.

3. Dependence on training data: The quality and scope of the training data greatly affect the performance of generative AI models. If the training data are biased, the model's output will likely also be biased. Similarly, if the training data lack certain information, the model won't be able to generate that information accurately.

4. Control and safety: It can be challenging to control the output of generative models. They might create content that is inappropriate, offensive, or misleading. This is a significant area of ongoing research in AI safety.

5. Resource intensive: Training generative AI models typically requires a lot of computational resources and data, making it inaccessible for individual researchers or small organizations.

6. Inability to verify facts: Generative models like GPT-3 don't have the ability to access real-time or current information and can't verify the truth of the information they generate; they can only draw on the knowledge that was available up until the point they were last trained. Applications on top of the models are being developed to perform web searches to look up facts.

7. Hallucination: The term comes from the idea that the model is "imagining" or "making up" details that were not in the input and do not accurately reflect reality. Hallucination can be a major issue in tasks where factual accuracy is important, such as news generation or question answering.