Machine Learning for All Blog
What is AI? - An Introduction to Machine Learning
Artificial intelligence is one of the most rapidly growing industries in our current society, and many new tools and platforms have been unveilved recently by large companies that are making impactful strides in this field. However, many are unaware of AI and why this field is so big right now. They feel that this is a really new and innovative concept that can seem hard to udnerstand, a field that seems valuable but difficult to grasp. If that sounds like you, you are in the right place. In this article, we will dive into the basics of machine learning and AI, what the difference is between those two terms, and how AI is currently impacting our society. By the end of this tutorial, you should have a basic understanding of what machine learning is, the different types of machine learning, and how it is used in our society. You can thank me later. Machine Learning and AI - A Definition We will start by defining these prevalent terms first - machine learning and AI. The definition of artifical intelligence, as the name suggests, is how we use machines (artifical units) to imitate human brain function (intelligence). Here is a more detailed definition. Artifical Intelligence - Computer systems capable of performing tasks that only humans could do and thus imitate human intelligence Examples of these tasks include reasoning, making decision, and solving problems So we have our first definition. Now what is machine learning? As the name suggests, machine learning is using machines to learn things. In case you want the definition in more professional language: Machine Learning - A branch of artifical intelligence that focuses on utilizing algorithms and data analysis to enable AI systems to imitate human intelligence and how humans learn. An easy way to differentiate between these two terms: AI is the what, and machine learning is the how. We want machines to perform tasks that only humans could do, tasks that could not be replaced by simple 1s and 0s, and we do this through computer algorithms and training data, which is the process of machine learning. Machine learning, at its core, is the process of using past data (training data) to predict future outcomes. The machine learning model, which is an algorithm programmed using a computer language, will train itself on past data and learn from it to predict what will occur in the future. This is similar to how us humans detect patterns. When we watch a video and an ad pops up once in the first 60 seconds, then another time in the next 60 seconds, we see the pattern that the ad occurs about once every minute. From this training data, we learn that the ad will likely show up in the next minute. How is Machine Learning Affecting Our Society? Machine learning and AI is prevalent in our everyday lives, and to drive this point home, I will give you some examples of how machine learning affects you in your normal routine. A streaming service (Netflix, Prime Video, etc.) recommending a new TV show for you to watch A website displaying ads on its home page for you to click on A video streaming channels like Youtube recommending a video Your virtual assistant, whether it is Siri or Google Assistant Maps and navigation apps rerouting you from your current path to a quicker way to your destination Online chatbots Google Translate or other translation services Search engines recommending articles or news for you based on your previous consumption (and maybe even recommending this article to you!) And this is just scratching the surface. Artifical intelligence is becoming widespread in industries such as medicine and finance where AI is projected to have enormous impacts in the near future. Just imagine an AI model being able to detect the likelihood of cancer in a patient, and at what stage that the cancer is currently in. Imagine the AI model then recommending or even carrying out a treatment for it. Beyond diagnosis, AI can even sequence the human genome to determine the causes and locations of certain conditions and diseases in the DNA, thus providng invaluable information as humanity searches to eradicated diseases in the future. Machine learning models can even predict house prices, stock market increases, business revenues and sales in the future, and market simulations. With the current influence and future impact AI has on our society, it is clear that understanding AI and how it works is essential to thriving in the job market. As large technology corporations like Meta and Google ramp up their AI industries, jobs in the machine learning sphere are becoming invaluable. Types of Machine Learning Now that we have familiarized outselves with what AI and machine learning are in general terms, let us examine different types of machine learning models and how they are used. The first and most commonly used type of machine learning is called supervised learning. In supervised learning, a machine learning model is fed data, often called training data, and uses this data to learn and train itself. The training data is usually labelled, and this means that the "coprrect answer" or "right answer" is already known. For example, if we have a predictive model that we want to predict future house prices, we could feed it data from 2015 house prices and then house prices in 2016, then 2017, and 2018, thus showing the relationship between 2015 house prices and subsequent prices. After the training process, the model is ready to predict future outcomes, and it usually does this on a new set of data, called the test data. Using our house prices example, we could give it 2022 data on house prices and ask it to predict the house prices in 2023. Using the trends shown in the training data, like increasing or decreasing trends, our model can predict what will happen next year. Here is the flow of a supervised machine learning model. Feed initial data that is labelled (has the correct answer) Train the model on this data Feed the model new data to test itself on (test data) Evaluate how the model did (how accurate was the model in predicting this?) The second type of machine learning is unsupervised learning. Here, the ML model is not fed training data initially. Instead, we just throw a whole bunmch of data into the model and ask it to find patterns in the data on its own. Due to this strategy of relying on the computer algorithm to find patterns, unsupervised learning models can sometimes be less accurate than supervised learning models. Examples of unsupervised learning are finding patterns in a song (the first 30 seconds is a chorus, followed by a riff, then another chorus, then a bridge, and an outro.) Here are the specifics on types of supervised machine learning and unsupervised machine learning. These examples will provised concrete examples of both types of ML and how they are used. Supervised Learning Regression - Predicting a numerical quantity (house prices, percentages, number of future employees) based on past data (training data) that the model uses and learns from. Predicting house prices in a certain neighborhood from the current and past home prices and sales Estimating employee productivity based on hours worked Estimated delivery time of a shipment based on past delivery rates Classification - Predicting the label, or category, that data falls into Image Classification - Detecting if images are dogs or cats, or if images contain phones or cars Sentiment Analysis - Determining the connotation, or attitude, of a text or audio (whether the overall tone of the data is positive and happy, or negative and sad) Categorizing books or television shows into genres Unsupervised Learning Clustering - Taking in data and sorting the data into groups, or clusters, based on certain attributes When processing the contents of a shopping cart full of fruits, the clustering algorithm may place them into categories of shape, thus placing apples and oranges (roundness) into one category and bananas and cucumbers into another (elongated) The algorithm may also choose to sort the contents into categories based on color - strawberries and apples in one category and kiwis and watermelon in another How Do We Code ML Algorithms? One of the most common programming languages used in AI is Python. This powerful and versatile programming languages is easy to learn and use, but it also has specific libraries like Tensorflow and Pytorch, both of which are extremely widespread today in the field of AI. These libraries are vital to machine learning and have several methods, functions, and workflows that aid in developing ML algorithms, which is why Python is the most widely used AI programming language today. Another programming language used for AI is R, which is a language focused on data analysis and processing. This is less widely known and used, but it is a language that aids in developing AI programs and algorithms in both supervised and unsupervised learning. However, I recommend using Python for AI because it is the most widespread language used today and it is easy to learn. Conclusion Now you know what machine learning and artifical intelligence are, how they are used in our society today, the impacts it can have, and why it is important for you to learn it. We have also covered the different types of ML and their implementations, providing concrete understanding of AI and how it affects our daily lives. In our next articles, we will examine these various implementations and algorithms in detail, such as Linear Regression, Neurwal Networks, and K-Means Clustering (do not worry if you have never heard these before!). These algorithms are key to implementing what we learned about supervised and unsupervised learning. Until then, see you later and keep exploring the diverse and fascinating field of AI!
OpenAI Unveils Sora: Advanced Text-to-Video Model
OpenAI has unveiled Sora, a cutting-edge generative video model that transforms concise text descriptions into detailed, high-definition film clips lasting up to a minute. This development represents a notable leap in text-to-video generation, showcased through sample videos that demonstrate Sora's impressive ability to understand complex 3D interactions and effectively handle occlusion. Despite strict secrecy conditions during the preview, OpenAI has not released a technical report or a demonstration of Sora in action, and there is no immediate plan for public release. The company is currently sharing Sora with a select group of safety testers and creative professionals to gather feedback and address potential misuse concerns. Built upon technology from DALL-E 3, OpenAI's flagship text-to-image model, Sora combines a diffusion model with a transformer, allowing it to process video data across both space and time. The transformer's capability to handle long sequences of data, similar to its application in language models like GPT-4, enables Sora to be trained on diverse video types in terms of resolution, duration, aspect ratio, and orientation. While the showcased videos highlight Sora's strengths, including high-definition output and effective occlusion handling, OpenAI acknowledges the need for further refinement, particularly in ensuring long-term coherence. Video generated with Sora Ai OpenAI is attentive to potential risks associated with generative video models, including misinformation and deepfake misuse. To address these concerns, Sora includes filters blocking requests for violent, sexual, or hateful content, and a fake-image detector developed for DALL-E 3 is adapted for use with Sora. Industry-standard metadata tags are embedded in Sora's output to indicate how the video was generated. Overall, Sora showcases significant advancements in generative video models, but OpenAI remains cautious about deployment, emphasizing the importance of gathering feedback and ensuring safety before any public release.