What is AI and What is not?
What I learned
I learned about the core components of AI; dataset, algorithm and prediction. It turned out that AI is widely used than I’ve been expected. By focusing these three elements, you can classify “What is AI and What is not”.
Dataset
Dataset is a collection of curated data, used for AI. AI is supposed to learn something from data and carry out an intelligent task. However, AI cannot learn directly from a large amount of miscellaneous data. It might be able to do, but it takes quite long time at least. Therefore, when designing AI, data should be curated so that machine can learn from it efficiently. For instance, let us think about an AI which classify the kind of dogs from dogs pocture. For this case, the dataset should be consists of a set of dogs picture, but not pictures of other animals nor objects. Furthermroe, it’s ideal if the pictures highlight the unique characteristics of each dog breed.
Algorithm
Algorithm is a set of rules of how to treat the data and learn from it. Algorithm decides the learning process of AI or machine learning. There are a lot of algorithm in today’s world. One of the most famous algorithms is what used in Google Search, called “PageRank”. PageRank is a method for rating Web pages objectively and mechanically, efectively measuring the human interest and attention devoted to them [1]. At PageRank, each websites are ranked in that those websites has been hyperlinked from reliable or important website, such as government websites or websites has big traffic. Search engine program can search effectivly by focusing high-ranked websites at the first and expand scope when its user ask for other search results. For this case, the set of websites and their links the dataset and the rules of ranking system is the algorithm.
Fig 2: PageRank Calculation, retrieved from PageRank Paper
Prediction
Finally, AI predict the result about given problem using what I learned from dataset, in a way, their “knowledge”. At Google Search, the system predict user input or the website they want to visit using its knowledge, such as the score of website retrievd by Page Rank algorithm. How often the result of AI’s prediction agree with the answer is an important parameter on comparing AI models.
My interest
Nowadays, AI is a kind of buzz phrase, and it is very often to hear or see in daily life. Besides AI, I also hear resemblant terms such as Machine Learning and Deep Learning. My interest lies in undesrstanding what each words mean and how they relate to each other.
AI, machine learning, Deep Learning
According to IBM web articles, these three words are defined as below [2],
- AI … technology that enables computers and machines to simulate human learning, comprehension, problem solving, decision making, creativity and autonomy
- Machine Learning … Directly underneath AI, creating models by training an algorithm to make predictions or decisions based on data.
- Deep Learning … a subset of machine learning that uses multilayered neural networks, called deep neural networks, that more closely simulate the complex decision-making power of the human brain.
AI refers to a broad varity of applications or technology for problem solving based on data, and machine learninig is a concrete technique that allow systems to learn from data utilizing Algorithm. Deep learning is a subset of machine learning, focusing on neural networks, and can be said a more specialized area of Machine Learning.
Fig 3: Abstract Venn diagram of AI, Machine Learning and Deep Learning
References
[1] L. Page, S. Brin, R. Motwani, and T. Winograd, “The PageRank citation ranking: Bringing order to the web,” Stanford InfoLab, 1998. [Online]. Available:http://ilpubs.stanford.edu:8090/422/
[2] IBM, “What is artificial intelligence (AI)?,” IBM Topics. https://www.ibm.com/topics/artificial-intelligence (accessed: Oct. 17, 2024).