1. What is generative AI?
Hello! I'm Daniel and I'll be your instructor for this course, where we'll learn about generative AI.
2. We've long dreamed of tools that can create
For centuries, humans have imagined powerful tools that can take a creative idea and bring it to life right before our eyes, whether that's in ancient stories of base metals transforming into gold, genies turning wishes into reality,
3. We've long dreamed of tools that can create
or in modern imaginings of virtual worlds we call to life at a whim.
While these have remained far-off fantasies, many artificial intelligence technologies are bringing powerful creative processes like this closer to reality.
4. Unparalleled creative tools
A new domain called generative AI is producing technologies that can similarly bring our ideas to life using various models. From ones that create magazine covers to others that appear to converse as a human would, these generative AI tools are rapidly bringing new creative capabilities to the mainstream.
5. What is generative AI?
Generative AI, or generative artificial intelligence, refers to machine learning models that can generate new content based on data they have already seen, when prompted.
One model might generate text in response to a user's questions.
6. What is generative AI?
Another can create videos for filmmakers to visualize their ideas using descriptive prompts.
7. What is generative AI?
And another can even write code to help developers work faster.
8. How does it work?
So how does it work?
A generative AI model takes a prompt as input. A prompt is a piece of data that guides the model to complete its task.
The model then processes the prompt to output, or generate, a response similar to data it has already seen.
9. How does it work?
A prompt is a piece of data that guides the model to complete its task.
The model then processes the prompt to output, or generate,
10. How does it work?
a response similar to data it has already seen.
11. Create images
Through this flow, one generative AI model can take a prompt describing a picture of a female astronaut as input, process it by running it through the model, and finally produce the image in the magazine cover we saw earlier.
12. Hold conversations
Similarly, another model can receive a prompt asking it to write an imagined conversation between the Sun and Pluto, process that prompt, and produce such a conversation as its response.
13. Input more than text
But prompts (our inputs) aren't limited to text. Depending on the model, any kind of data can be an input and any kind of data can be an output.
For instance, this prompt is not text, but rather an image of a person walking along a beach. The person is highlighted.
The model then processes that image, and its response is the same beach image but with the highlighted person removed.
14. Real-world applications
As these models are increasingly capable, they will appear more frequently in daily life and work. For instance,
they can be used to help draft sales outreach emails,
analyze financial data,
generate marketing ads to test,
help legal professionals sift through a regulatory landscape,
customize education experiences,
support medical practitioners in analyzing medical data,
automate repetitive tasks for industrial engineering and design,
and generate 3D objects for games,
among many more applications.
15. The end of work?
So, will the robots be taking all our jobs in short order?
Not exactly, there are still limitations and risks to this technology, many of which we'll explore throughout the course.
For now, a greater understanding of these technologies will open up new ways to utilize these tools in our daily and professional lives.
16. Course goals
We'll learn all about how these models
are trained to create new content,
bring up new legal and ethical considerations,
and will significantly impact many aspects of human society in the coming years.
17. Let's practice!
Time for some exercises to test our understanding.