Explaining AI at the family Christmas lunch

As a developer at DataStax, I often find myself trying to explain what I do for a living. For people working in tech and using AI tools or coding copilots, telling them I work with generative AI is often well understood. Sitting around the table at Christmas lunch with the family can be a different story.
 
If you're not prepared when someone innocently asks, "What exactly is this AI thing everyone keeps talking about?" you might find yourself floundering. So, here's an easy way to explain AI—and related concepts like RAG—to your parents and grandparents while they enjoy the Christmas ham.

What is AI?
You can start by explaining artificial intelligence. Since the 1950s, artificial intelligence has been a field in computer science in which developers have tried to create programs that simulate human thinking or learning. Most programs act like humans learning to cook by following recipes exactly as written. However, programs that exhibit artificial intelligence might act more like confident chefs, making decisions and altering the recipe based on what they see or experience.

It's worth pointing out that you constantly interact with AI, whether you realise it or not. Whenever you search with Google, watch a movie recommended by Netflix, or search your phone for pictures of your dog (this can't just be me, right?), you use artificial intelligence.

But we're not suddenly excited about some algorithm invented in the 50s. The recent wave of activity in AI is due to generative AI.

So, what is Generative AI?
Whereas traditional AI (and yes, it does feel weird to call it that) is good at things like finding hidden patterns or predicting results, the magic in generative AI (GenAI) is that it can do two things:

• Understand inputs like natural language or media, like images
• Generate new content, like human-sounding text or even images, audio, and video

You might have seen tools like ChatGPT, Google Search AI Overview, or Amazon's user review summaries. These are all powered by generative AI, and the common feature is that a human didn't write the text in these examples. That simply wouldn't scale; an artificial intelligence wrote it.

That might set off some conversation around the dinner table, leading to follow-up questions. One such question might be, "Oh yeah, didn't Google's AI suggest people eat rocks and glue cheese to pizza?" 

This is the perfect opportunity to describe retrieval-augmented generation (RAG).

What is RAG (Retrieval-Augmented Generation)?
These GenAI programs seem to capture a lot of human knowledge and the ability to generate meaningful text. However, they are not always right and certainly don't know everything, even if they often sound very confident. So, we need to help them out with the right information.

For example, if Santa brought you a new pair of Bluetooth headphones, you might want to ask, "How do I connect my new headphones to my phone?" A generative AI might suggest you put the headphones in pairing mode and then open your phone's Bluetooth settings. But it could be a lot better.

Here's how RAG helps:

1. Retrieval (getting information): RAG lets AI search for extra information when it doesn't know the answer off the top of its head. It fetches the correct information from places like the internet, just like how you might look up how your headphones work in a manual.
2. Augmentation (adding extra help): Once the AI has the additional information, it can offer more specific details to ensure it answers your question correctly. Providing the headphones' manual, particularly the pages about Bluetooth connections, to the AI sets it up for success.
3. Generation (giving you the answer): The AI can generate a clear, easy-to-understand response with the correct contextual information. Your AI assistant is not just memorising answers but combining information to create something helpful. It's like telling you which tiny button to hold to put those headphones in pairing mode.

RAG gives AI an extra boost by allowing it to research, add context to the information, and give thoughtful and informed answers.

Real-Life Example
Say you want to buy the perfect gift for someone this Christmas, but you have no idea what they'd like. AI can help by searching through millions of gift ideas and generating suggestions based on what it knows about your friend or family member. If AI is powered by RAG, it will search for the best ideas from the internet and bring back more personalised and up-to-date recommendations. So, not only does it find gift ideas, but it also pulls in information from the best sources to make the suggestions more accurate.

Wrapping It Up
As a developer or anyone working in the AI space, it can be challenging for family members to understand what you are discussing regarding work. This guide should help you explain the details in an easy-to-understand way. Even if, after explaining all of this, they still ask you to fix the printer while you're visiting.