Understanding AI and the Foundations of Machine Learning
GitHub Copilot · AI-Assisted Software Development · Bởi Super Admin
AI is a broad term encompassing machine learning, deep learning, and generative AI, with machine learning allowing computers to learn from data through supervised, unsupervised, or reinforcement learning approaches.
Defining AI can be a challenge because it is a complicated technology that is constantly changing, and the industry is filled with hype and marketing exaggeration. A helpful way to understand it is through a diagram of four concentric circles: Artificial Intelligence is the outer, general term that includes various technologies like machine learning (ML), deep learning (DL), and generative AI (GenAI). For the most part, these technologies allow machines to engage in learning, reasoning, problem solving, and decision-making.
Machine Learning as an Early Approach to AI
Machine learning is one of the earliest approaches to AI. In 1959, IBM researcher Arthur Samuel published a paper setting forth the elements of this technology, noting that "programming computers to learn from experience should eventually eliminate the need for much of this detailed programming effort." The basic idea is that ML allows a computer to learn without being explicitly programmed. A spam filtering application illustrates this: you need large amounts of data, such as emails labeled "spam" or "not spam," and you train a model using an ML algorithm like linear regression, decision trees, or k-nearest neighbors. A popular algorithm for spam detection is the Naive Bayes Classifier, which uses probability analysis to determine the likelihood that an email is spam. Once trained, the model sorts incoming emails, sending spam to a separate folder, but an effective model still needs monitoring and updates—for example, Gmail improves its models using user feedback when people mark emails as spam or not.
Supervised, Unsupervised, and Reinforcement Learning
When an ML model uses labeled data, this approach is called supervised learning, which is used for classification (dividing data into classes or categories, as with spam filtering, image recognition, and fraud detection) as well as regression (predicting numerical values such as temperatures or home prices). Beyond supervised learning, there are two other ways ML can help a computer learn. Unsupervised learning uses unlabeled data, with an algorithm finding patterns in groups or clusters—for example, a product recommendation system might predict that a customer who buys a cooking book is likely to buy kitchen gadgets or other cookbooks. Reinforcement learning trains a model based on rewards and punishments through interaction with an environment, such as game playing, where performance improves by going through many moves and observing which ones result in the highest scores.
Key takeaways
- AI is a general term including ML, DL, and GenAI, which enable learning, reasoning, problem solving, and decision-making.
- Machine learning, dating to Arthur Samuel's 1959 paper, lets computers learn from experience rather than explicit programming.
- Supervised learning uses labeled data for classification or regression tasks.
- Unsupervised learning finds patterns in unlabeled data, while reinforcement learning uses rewards and punishments.
Thảo luận / Bình luận (0)
Chưa có thảo luận nào cho bài viết này. Hãy gửi bình luận đầu tiên nhé!
Vui lòng Đăng nhập để tham gia thảo luận.