Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
If you’re new to artificial intelligence (AI) and don’t have a technical background, but want to learn the basics quickly, this post is for you.
I recently took Google’s 4-hour AI course for beginners and condensed the key takeaways into an easily digestible format.
Initially, I had some doubts about whether the course would be practical enough, considering how theoretical many online resources can be.
However, I was pleasantly surprised by how much I learned and how these concepts helped me better understand popular AI tools like ChatGPT and Google Bard.
This post covers the basics of AI, machine learning, deep learning, and generative models—all broken down into easy-to-understand chunks. So, let’s get started!
What is Artificial Intelligence?
The first important distinction I learned is that artificial intelligence is an entire field of study, just like physics or chemistry. AI encompasses many subfields, one of the most significant being machine learning (ML).
Machine learning itself is a subset of AI, much like how thermodynamics is a subfield of physics.
Within machine learning, there’s an even more specific area called deep learning, which focuses on models inspired by how the human brain works, often referred to as artificial neural networks.
To summarize:
- AI: The broad field of study.
- Machine learning (ML): A subfield of AI focused on learning from data.
- Deep learning: A specialized subset of ML using neural networks.
At the intersection of deep learning and large language models (LLMs) is where you find the technology powering tools like ChatGPT and Google Bard.
What is Machine Learning?
So, what exactly is machine learning? In simple terms, machine learning is about using data to train a model that can make predictions on new, unseen data.
For instance, if you train a model on sales data from Nike, you can use that model to predict how well a new Adidas shoe might sell based on Adidas’ sales data.
This is a powerful way to make decisions based on past performance.
Two common types of machine learning are:
- Supervised learning: This type of learning uses labeled data to train models. For example, you may have historical data where each data point is labeled (e.g., “delivered order” or “picked up order”). The model learns from these labels and can make predictions about future orders.
- Unsupervised learning: This type of learning uses unlabeled data and tries to find patterns. For instance, when looking at employee tenure vs. income data, you may not have specific labels for each employee (such as job role or gender), but the model can still group similar employees together.
One key difference is that supervised models adjust their predictions based on comparing them with known outcomes, while unsupervised models simply find patterns without any labels.
Deep Learning and Neural Networks
Moving deeper into the subject, deep learning is a type of machine learning that utilizes artificial neural networks.
These networks are modeled after the human brain and consist of layers of nodes (neurons). The more layers there are, the more powerful the model becomes.
Deep learning models can handle semi-supervised learning, which is a combination of labeled and unlabeled data.
For example, a bank might label only 5% of its transactions as fraudulent or not, leaving the remaining 95% unlabeled.
The deep learning model uses that small labeled dataset to learn the basic task and apply those learnings to the unlabeled data to make predictions. This is highly useful in scenarios where labeling every data point is time-consuming or expensive.
Discriminative vs. Generative Models
One fascinating thing I learned is that deep learning models can be categorized into two types: discriminative and generative models.
- Discriminative models: These models classify data points. For example, if you have images labeled as “cat” or “dog,” a discriminative model will predict whether a new image is a cat or a dog based on the labels it has seen during training.
- Generative models: These models don’t just classify—they create new data based on the patterns they’ve learned.
In short, if the output is a prediction or a classification (like “spam” or “not spam”), it’s a discriminative model. If the output is something new—like text, an image, or audio—it’s a generative model.
Generative AI in Action
Most of us are familiar with generative AI in the form of text-to-text models like ChatGPT and Google Bard.
These models can generate human-like responses based on the input text we provide. But there’s more—other types of generative models include:
- Text-to-image models like Midjourney and DALL·E that can create and edit images based on textual prompts.
- Text-to-video models like Google’s Imagen Video, which generate or edit video content.
- Text-to-3D models, which are used in industries like gaming to create 3D assets.
- Text-to-task models, which perform specific tasks, such as summarizing your unread emails.
Large Language Models (LLMs)
At this point, it’s essential to clarify that large language models (LLMs) are a subset of deep learning but are not the same as generative AI, though there is some overlap.
LLMs, like GPT-3 (which powers ChatGPT), are trained on massive datasets to understand and generate human language.
LLMs are usually pre-trained on large amounts of general data (think of this as learning basic commands like “sit” or “stay” for a dog).
Then, they are fine-tuned for specific tasks or industries (like training a dog to become a police dog).
For example, a large hospital might take a pre-trained LLM and fine-tune it with its medical data to improve diagnostics from X-rays or lab reports. This fine-tuning allows for domain-specific expertise without building a model from scratch.
Why AI Matters for Non-Technical People
Understanding these basic concepts is beneficial even if you’re not a tech expert. Knowing the distinction between AI, machine learning, and deep learning can help you better utilize tools like ChatGPT or Google Bard and clear up misconceptions you might have about how these technologies work.
For example, knowing that generative AI creates new content from patterns can help you see its potential for creative projects, while understanding how discriminative models classify data can be useful in more business-oriented tasks like fraud detection or sales prediction.
Final Thoughts
If you’re interested in learning more, Google’s 4-hour AI course is an excellent free resource that covers these topics in more depth.
The course includes five modules, and you can even earn a badge after completing each one.
In the meantime, I hope this post helped clarify some of the fundamentals of AI, machine learning, and generative models. Stay curious, and keep exploring the possibilities that AI can offer!
