exam 2 AI

Artificial Intelligence

The simulation of human intelligence processes by machines, particularly computer systems, encompassing tasks like learning, reasoning, and self-correction.

Human Intelligence

The mental quality that consists of the abilities to learn from experience, adapt to new situations, understand and handle abstract concepts, and use knowledge to manipulate one's environment.

Turing Test

A method proposed by Alan Turing in 1950 to evaluate a machine's ability to exhibit intelligent behavior indistinguishable from a human, where a human evaluator engages in text-based conversation with both a machine and a human, without knowing which is which; if the evaluator cannot reliably distinguish them after sufficient interaction, the machine passes.

Imitation Game

The original name for the Turing Test, involving a man, a woman, and an interrogator separated by a screen, where the interrogator tries to identify the woman based on text responses, adapted to test machine vs. human indistinguishability.

System Capability Limitations

AI systems excel at narrow, well-defined tasks (e.g., pattern recognition) but struggle with general intelligence, common-sense reasoning, and adapting to novel situations without extensive retraining, as highlighted in key takeaways on AI's bounded capabilities.

AI Rights Debate

Arguments for granting AI rights center on potential sentience and moral agency if AI achieves human-like consciousness (e.g., rights to non-exploitation), while against emphasize current AI as non-sentient tools lacking qualia, emotions, or true understanding, making rights anthropomorphic and impractical.

Sentience in AI

The capacity for subjective experience, feeling, and consciousness; pursuing it in AI raises ethical dilemmas like moral status and suffering, but current AI lacks it, as it's a philosophical threshold beyond computational mimicry.

Weak AI

AI designed for specific tasks (e.g., chess-playing programs like Deep Blue), realistic because it leverages narrow expertise without needing general intelligence; example: voice assistants like Siri for query handling.

Strong AI

Hypothetical AI with general intelligence matching or surpassing humans across all cognitive tasks; not realistic currently due to unresolved challenges in common-sense reasoning and consciousness simulation.

AI Evolution Stages

Progress from rule-based systems (1950s-1970s, symbolic AI) to machine learning (1980s-2000s, data-driven) to deep learning era (2010s+, neural nets with big data), marked by AI winters from overhype and funding cuts.

Multidisciplinarity of AI

AI draws from computer science (algorithms), ethics (moral implications), psychology (human cognition modeling), philosophy (intelligence definitions), and neuroscience (brain-inspired architectures), requiring integrated approaches for holistic development.

Supervised Learning

A machine learning paradigm where models are trained on labeled data (input-output pairs) to predict outcomes, using algorithms like regression or classification to minimize prediction errors via loss functions.

Loss Function

A mathematical measure in supervised learning quantifying the difference between predicted and actual outputs, guiding model optimization (e.g., mean squared error for regression).

Unsupervised Learning

Machine learning on unlabeled data to discover hidden patterns, such as clustering (grouping similar items) or dimensionality reduction (simplifying data while retaining structure).

Clustering

An unsupervised technique grouping data points based on similarity (e.g., k-means algorithm partitioning into k clusters), useful for exploratory analysis without predefined categories.

Reinforcement Learning

A learning method where an agent interacts with an environment, receiving rewards or penalties for actions to maximize cumulative reward over time, often modeled as Markov Decision Processes.

Markov Decision Process

A framework for reinforcement learning defining states, actions, transition probabilities, and rewards, enabling agents to learn optimal policies through trial and error.

Artificial Intelligence

Narrow focus on creating systems that perform tasks requiring human-like intelligence in specific domains, often rule-based or data-driven.

Augmented Intelligence

Human-AI collaboration enhancing human decision-making (e.g., AI suggesting options in medical diagnosis), emphasizing symbiosis over replacement.

Prescriptive AI

AI that recommends actions or decisions based on rules or predictions (e.g., navigation apps suggesting routes), focusing on "what to do."

Descriptive AI

AI that analyzes and summarizes data patterns without prescribing actions (e.g., dashboards showing sales trends), focusing on "what is happening."

Deep Learning

A subset of machine learning using multi-layered neural networks to automatically learn hierarchical feature representations from raw data, excelling in tasks like image recognition.

Neural Networks

Computational models inspired by biological neurons, consisting of interconnected layers (input, hidden, output) that process data through weighted connections and activation functions.

Backpropagation

A training algorithm for neural networks that computes gradients of the loss function with respect to weights, propagating errors backward to update parameters via gradient descent.

Generative AI

AI systems that create new content (e.g., text, images) by learning data distributions, such as GANs generating realistic images or LLMs producing coherent text.

Generative Adversarial Networks (GANs)

A generative AI architecture with a generator creating data and a discriminator evaluating realism, trained adversarially to improve output quality.

Traditional AI

Rule-based or symbolic systems relying on explicit programming and logic for decision-making, limited in handling ambiguity or large-scale data.

Generative AI Differences

Unlike traditional AI's deterministic outputs from fixed rules, generative AI produces novel, probabilistic content from learned patterns, enabling creativity but risking hallucinations or biases.

Agentic AI

AI systems acting autonomously as agents in environments, pursuing goals through planning, decision-making, and adaptation (e.g., robotic assistants), building on reinforcement learning for multi-step interactions.