1. 10,000 ft view: how does AI work?
Let’s understand how AI works.
2. Finding patterns in big data
To put it simply, AI is able to efficiently find
3. Finding patterns in big data
patterns within massive datasets. It is then able to use those patterns to make decisions. For example modeling the relationship of colored squares in this picture, or predicting what the picture is about.
4. High-level AI model process
Like most systems - technical or otherwise - there is a three-step process. First, an input, in this case, data.
5. High-level AI model process
An action is done with the input. In an AI solution, this is a model which will identify the patterns.
6. High-level AI model process
Finally, there is an output. Typical output falls into the categories of detecting, predicting, or classifying something. For example, predict a sales outcome, classify a piece of marketing content, generate new data, or automate a task.
7. How is generative AI different?
Generative AI is a sub-category of AI. These types of models generate new outputs based on the data they have been trained on. This new output is in the form of human-like content such as images, text, audio, and more.
In most cases, this is done through an interaction using natural language that is more human-like than ever before.
Most models are also built in such a way to have multiple functions such as text creation and summarization.
Be careful though, the output is still not 100% correct but pretty amazing none-the-less.
8. Create new content with generative AI
Here are a couple of examples on how generative AI can be used.
A media team using a text-to-image generator to create images and videos for a new mobile app launch.
A legal team using generative AI to summarize documents and highlight keywords or topics.
9. Basic AI solution
Let's extend on the basic concept of an input, action, and output, to understand the basic components of an AI solution.
First, there is a lot of data! We will dive deeper into the specifics and best practices for data later on in the course.
10. Basic AI solution
This data is then used to build the AI model, which will identify the patterns and return an output.
11. Basic AI solution
The AI output will then be integrated into some existing or new system.
12. Basic AI solution
This system typically will have a communication setup to the AI model for real-time output.
13. Basic AI solution
Finally, there is an interaction with the AI solution through the "user experience or user interface", aka UX/UI. This is the medium through which the user will receive output from the AI solution and give new inputs, for example a chat experience.
14. Basic AI solution
This interaction is important as it, along with new data, will help the AI model develop an even deeper understanding of the patterns. Therefore, a strong feedback and communication system must exist within the solution.
15. Basic AI solution
Because of this complexity, the amount of data required, and the type of actions being done to the inputs, AI solutions require specialized hardware built to handle high volume and intensive processes.
16. AI solutions vs. non-AI data solutions
So how do AI solutions stack up against data solutions? An AI solution differs from a non-AI data solution in a couple of ways.
It requires more data.
There are more intensive processes.
There are two sub-systems in the architecture - one for the AI model and one for the system it will be integrated into.
AI can have a more interactive user interface, for example chat bots, prompts to generate images, and inputs to evaluate different "What-if" scenarios.
Specialized hardware which may increase costs.
Because the AI solution is continually learning, a monitoring component is also required.
Finally, risk, privacy, and compliance need to be thought about more.
17. Let's practice!
This is the 10,000 foot view of how AI works. It’s helpful to have a basic understanding as you pursue the goal of using it within your business.