AI(ICT)
Overview of AI
- AI performs tasks based on the level of human intelligence required.
Types of AI
- Rule-based AI
- Programmed AI that follows fixed instructions.
- Cannot learn independently.
- Machine Learning AI
- Learns from data and improves over time as it receives new information.
Interaction with AI
- AI improves by learning from data users provide during interactions.
- Users submit questions and AI uses this feedback to adapt its responses.
History of AI
- 1950: Alan Turing posed the question, "Can machines think?"
- This question led to the development of the Turing test, a measure of a machine's ability to mimic human conversation convincingly.
- 1956: The term Artificial Intelligence was coined at a conference at Dartmouth College.
- Early AI programs could solve mathematical problems and play basic games but were limited due to technical constraints.
- Late 1960s: AI progress stalled; known as the first AI winter due to lack of funding and technological capabilities.
- 1980s-1990s: Use of expert systems in AI began; these systems imitated human decision-making but lacked adaptability.
- 1997: Deep Blue defeated chess champion Gary Kasparov, marking a significant AI milestone.
- Early 2000s: The digital revolution transformed data availability and internet connectivity.
- 2012: AlexNet, a deep neural network, won a global image recognition contest, leading to the deep learning revolution.
- AI began to recognize faces, translate languages, and assist in various real-world tasks.
- 2016: DeepMind's AlphaGo defeated a Go champion, showcasing AI's ability to compete in complex intuition-based games.
- 2018: Introduction of Google Duplex, capable of human-like phone conversations, further humanizing AI interactions.
- 2020: Release of GPT-3 by OpenAI, a powerful language model that could create human-like text, aiding in app development and creative writing.
- 2022: AI models like DALL-E 2 and Midjourney began producing art, showcasing a new form of creativity accessible through simple text input.
- ChatGPT launched, gaining 1 million users in just five days.
- 2023: Emergence of multimodal AI with GPT-4, improved analysis of images, and solving logic problems.
Key Milestones in AI History
- 1956: Term “Artificial Intelligence” introduced.
- 1966: First chatbot conversation established.
- 1977: Development of Deep Blue for chess.
- 2020: Release of advanced AI models by OpenAI.
Machine Learning Pipeline
- The machine learning pipeline involves several stages that raw data undergoes to become useful.
- Raw Data: Unorganized and unprocessed information collected from various sources, including text, images, and numerical records.
- Preprocessing: Raw data is cleaned and transformed into structured data, which is organized and error-free.
- Training: Structured data is used to train a learning algorithm, updating the model.
- Evaluation: Models are tested using new data to assess their performance and accuracy.
- Deployment: The system is put into real-world use, becoming accessible to users.
- Monitoring: Continuous performance checks are conducted with adjustments made as needed.
Summary of Concepts to Memorize
- Key historical dates and milestones in AI development.
- Differences between raw data (unprocessed) and structured data (organized).
- Stages of the machine learning pipeline (raw data, preprocessing, training, evaluation, deployment, monitoring).
- Differences between Rule-based AI (static instructions) and Machine Learning AI (adaptive learning).