Introduction

A Test developed in the 1950s to determine the intelligence of a Computer/AI. It involves a computer and two humans. One human is a judge asking questions to both the secoond human and the computer without seeing them (-> through terminal). The judge has to determine wich answer is Computer generated and which comes from a human. If the judge is wrong in half or more of the cases it would mean that the computer can effectifly mimic a human and therfore can be considered intelligent.

CAPTCHA -> reverse turing-Test

Traditional: By used algorithmn

Mashine Learning: By lerning to recognize patterns

Similiar as humans:

1. through Examples -> supervised learning

2. through Assumptions & Observations -> unsupervised learning

3. through Rewarding and Punishment -> reinforcement learning

1. When Expert knowledge is missing -> making Predictions

2. When humans can only do it “intuitively” -> natural language processing

3. When something changes frequently -> Weather forcasting, Stock market analysis

4. Costomized leraning -> e.g. Spam filters

1. Natural language Processing

2. Medical Diagnosis

3. Software engineering

4. Robotics

5. Autonomous driving

6. Fraud Detection

1. Artificial Intelligence (AI): A collective term that summarizes techniques and methods to enable computers to mimic human intelligence.

2. Machine Learning (ML): A subset of AI that can parse large amounts of data, finding patterns and learning from these patterns to make predictions or conclusions. (actually applied Staatictsics)

3. Deep Learning (DL): A subset of machine learning that uses multilayer neural networks to learn from vast amounts of data.

1. discrete Data is used for classification problems to determine which category or class an instance belongs to (spam or not spam email)

2. continuous data is used for regression problems to determine/ predict a continuous output value (prediction of house price based on certain features)

1. Supervised Learning: Involves labeled data with known input-to-output relationships. The machine learns from these known relationships to make predictions on new data (classify).

2. Unsupervised Learning: Involves unlabeled data where the relationship between input and output is unknown. The machine tries to find patterns by clustering the data.

1. Automate manual tests and deploy testing techniques.

2. Object recognition can help find changes in GUI appearance, such as a changed appearance of a button.

3. Based on data, AI can identify potential weaknesses and test for them more effectively.

4. AI can be used to mimic customer behavior and test the system like a real human would interact with it.

1. Results might not always be trustworthy.

2. Potential for overfitting, where the AI model performs well on training data but poorly on new data.

3. Generating the right results might be time-consuming.