Artificial Intelligence

[AI generated image]

Oxford Reference defines artificial intelligence as "The theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages."

Artificial Intelligence (AI) alternately may refer to a subject of study, a scientific discipline, a tool, or competition for human endeavors.

 AI encompasses a number of areas: communications, learning, reasoning, understanding, vision, and robotics.

Drawbacks of AI currently include a lack of explainability (behaving as a black‐box model) and a number of ethical issues.

Terminology:

• AI prompt engineer: Intermediary that writes text-based prompts to optimize desired outputs for users.
• Artificial Intelligence: Attempts to reproduce human behavior / learning. The term "artificial intelligence" is often used to refer to subsets of AI. Developed in the 1950s.
• ChatGPT: A Generative Pre-training Transformer that uses natural language and advanced deep learning to make predictions
  • ChatGPT is a type of software--an AI chat bot. Latest version 4.0 is able to connect to the Internet (although this poses security and privacy threats). It can be downloaded as a mobile app for iOS and Android devices. Version 4.0 is accessible only to paid subscribers.
  • The version 3.0 is free, has an information cutoff (last update) of September 2021, and does not connect to the Internet.
• CLEAR Framework for prompt engineering (Leo Lo: https://doi.org/10.1016/j.acalib.2023.102720): 
  • Concise
  • Logical
  • Explicit
  • Adaptive
  • Reflective
• Deep learning: Avoids the need for human operators using comparatively larger datasets and a hierarchy of concepts built from simpler concepts.
• Generative AI: Sophisticated text prediction inspired by human neural networks; uses a subset of machine learning with datasets to generate new content.
• Hallucinations: Refers to erroneous ("fake") AI-generated results.
• Large Language Model: A model that uses massive amounts of training data to teach algorithms without human instruction.
• Machine learning: A subset of AI used to teach computers to learn and act without programming that uses comparatively smaller datasets.
• Retrieval Augmented Generation (RAG): AI framework that combines information retrieval systems with large language models (LLMs) to improve text generation.
• Token: Refers to unit of data (words, phrases, sentences, characters, subwords, or punctuation marks) fed into a model for training or processing. 
• Tortured phrases: Mistranslated technical terms, e.g., "colossal information" for "big data"