An Honest Review of Google's Intro to Generative AI Courses

So Google recently just dropped a list of free Intro to Generative AI courses and the biggest question is - are they worth the hype?

I guess that’s what I’m here for. After completing Google’s Introduction to Generative AI learning path (check it out here), here are my two cents.

Table of Contents

• Course Breakdown
  • Target Audience
  • Learning Goals
  • Content Analysis
  • Summary/Cheatsheets
• Final Verdict (Is it Worth It?)
• Additional Resources

Google’s Generative AI Learning Path Course Breakdown

Target Audience

Google’s courses are catered to students without prior technical knowledge of the topic, or as explained on their blog, the path is designed for the non-technical, tech-adjacent audience with roles like sales, HR, marketing, and operations. On that note, this means that the course doesn’t provide in-depth how-to’s but does give a great introduction to fundamental concepts in generative AI.

Learning Goals

The idea is to properly answer the question “But what actually is Generative AI?”.

The emphasis of the course is on Generative AI, Large Language Models, and ethics in AI. Through the course, you’ll also be able to learn about the AI services provided by Google and Google’s role in incorporating responsible AI practices, which includes a mix of case studies, lecture-style videos, and quizzes.

Content Analysis

So let’s decipher the material in the learning path:

• Introduction to Generative AI - course
• Introduction to Large Language Models - course
• Introduction to Responsible AI - course
• Generative AI Fundamentals - course with skill badge
• Responsible AI: Applying AI Principles with Google Cloud - course

Understand this, the Generative AI Fundamentals course is basically a compilation of the first three courses (Intro to Gen AI, Intro to LLM, and Intro to Responsible AI), but with the addition of a quiz. And here’s the deal - if you were to complete the courses prior to the skill badge course, you wouldn’t need to redo them. My guess on why the learning path is formatted that way is to ensure that you’ll actually go through the course materials rather than just clicking ‘next’ when attempting the Generative AI Fundamentals skill badge.

Picking out the content from the courses, here are some cheatsheets I developed:

Summary + Cheatsheets

Generative AI

Starting with “What is Artificial Intelligence?”, it’s a field in Computer Science that mimics human cognition to perform complex tasks and learn from them. Within AI, there exists the subfield of Machine Learning which uses algorithms trained on data to produce adaptable models that can perform various complex tasks.

Within ML, there exist different types including supervised learning, unsupervised learning, reinforcement learning, and deep learning. Deep learning uses artificial neural networks, allowing them to make more complex patterns whereby its neural networks use supervised learning, unsupervised learning, and semi-supervised learning to achieve its goal. There are two types of deep learning models, namely discriminative and generative.

Generative AI is a type of Artificial Intelligence technology that can produce various types of content, including text, imagery, audio, and synthetic data. In short, GenAI is a type of AI that creates new content based on what it has learned from existing content. It uses a learning process called training which results in the creation of a statistical model, in turn, is used to predict what an expected response may be when given a prompt.

Large Language Models

Large language models are deemed large because of two factors - being trained on a large training dataset with a large number of parameters. Often called hyperparameters, parameters are essentially the memories and the knowledge that the machine learned, and are used to define the skill of the model to solve problems. LLMs are also general purpose as they strive to solve common problems using human languages.

There are two kinds of LLMs, namely pre-trained and fine-tuned whereby pre-trained models can do “everything” but have their practical limitations whereas fine-tuned models fit a specific niche or aim to solve a specific problem. It’s important to note that fine-tuning tends to be expensive, thus more efficient methods exist such as Parameter-Efficient Tuning Methods (PETM) like Prompt Tuning.

To understand prompts, they are essentially inputs that are given to an LLM to elicit a specific response. The usual misconception is between prompt design and prompt engineering. To break it down, prompt design is tailored to the specific task that the system is being asked to perform whereas prompt engineering is designed to improve the model’s performance by using domain-specific knowledge, providing examples of a desired output, and using keywords that are known to be effective for this specific system.

AI/ML Services by Google

Vertex AI offers a model garden for foundation models. Consider the use case whereby the user intends to use a model to predict customer satisfaction, they can opt-in to use the Classification Task type Sentiment Analysis Task Model.

Integrating the PaLM API with MakerSuite simplifies the Generative Development Cycle. MakerSuite includes a bunch of resources such as a model training tool to train models on the user’s data using different algorithms, a model deployment tool to allow users to deploy their models to production, and a model monitoring tool to monitor the model’s performance in production.

The GenAI Studio allows users to quickly explore and customize GenAI models with resources such as a library of pre-trained models, tools for fine-tuning models, tools for deploying models to production, and a community forum for more support.

The GenAI App Builder provides users with a drag-and-drop interface, a visual editor to edit app content, a built-in search engine, and a conversational AI engine.

Bard is a conversational AI tool that essentially is an LLM similar to ChatGPT.

Free AI Courses by Google - Are They Worth It?

Yes, in the simplest terms, they are. If you knew me, you’d know that I always stand by taking advantage of new learning opportunities.

However, it's important to note that the course is not perfect. There can be some overlap between certain topics, and it only focuses on Google's contributions to the AI industry. Additionally, the quizzes may not be challenging enough, especially considering the fact that each module usually consists of only 3-5 questions per quiz.

However, it's worth considering that the course is completely free and allows you to showcase your achievements on social media platforms and your resume. Furthermore, the course is concise and straightforward, so it won't consume too much of your time. Take it from me, I was able to finish the path in less than a day 👀

How to Continue Your Generative AI Learning Journey Beyond Google Courses

I’d like to think that our learning journey should never depend on only one source, here are some other courses you can take a look at:

• Generative AI with Large Language Models by DeepLearning.AI
• Introduction to Generative AI by Microsoft
• Generative AI Foundations on AWS Technical Deep Dive Series by AWS
• Do You Speak Vector? Understanding the Language of LLMs and Generative AI by Datastax
• A Beginner’s Guide to Using Large Language Models (LLMs) with the PaLM API

Others:

• Hugging Face
• Tensorflow
• LangChain