If you’ve tried to go past the APIs like the ones OpenAI offers and learn how they work “under the hood” by trying to build your own neural network, you might find yourself hitting a wall when the material opens with equations like this:
How can you learn how neural networks — or more accurately, artificial neural networks — do what they do without a degree in math, computer science, or engineering?
There are a couple of ways:
Follow this blog. Over the next few months, I’ll cover this topic, complete with getting you up to speed on the required math. Of course, if you’re feeling impatient…
Read Tariq Rashid’s book, Make Your Own Neural Network. Written for people who aren’t math, computer science, or engineering experts, it first shows you the principles behind neural networks and then leaps from the theoretical to the practical by taking those principles and turning them into working Python code.
Along the way, both I (in this blog) and Tariq (in his book) will trick you into learning a little science, a little math, and a little Python programming. In the end, you’ll understand the diagram above!
One more thing: if you prefer your learning via video…
Masterminds Tampa Bay is holding their AI Superpowers Unlocked panel on Wednesday, May 15th with the following panelists:
Ken Pomella, CEO of RevStar, known for leveraging AI to enhance business growth and scalability.
Lenar Mukhamadiev, from IdelSoft, focusing on GenAI solutions for organizations and developing an AI-powered startup.
Sat Ramphal, CEO of Maya AI, a serial entrepreneur with deep expertise in AI applications in regulated industries.
Yours Truly, Joey de Villa, Supreme Developer Advocate for Unified.to, AI enthusiast, Python instructor, and general computational person about town.
Here’s Tampa Bay Masterminds’ description of the event:
Unlock the future of Artificial Intelligence at “AI Superpowers Unlocked: An Expert Panel,” an event meticulously crafted for entrepreneurs, tech enthusiasts, and forward-thinkers ready to explore AI’s transformative potential. Join us to gain practical insights on becoming a leader in AI application and connect with industry pioneers.
AGENDA
6:00 PM – 6:30 PM: Socializing Time
6:30 PM – 7:15 PM: Expert Panel Discussion
7:15 PM – 8:00 PM: Audience Q&A / Conclusion
Main Takeaways:
🧠 Understand the crucial role AI plays and why mastering it is essential.
🧠 Learn strategies to best leverage AI for 2024 and beyond.
🧠 Discover essential AI tools beyond ChatGPT.
🧠 Explore best practices, ethics, and more through interactive FAQs.
Why You Should Attend:
🚀 Tailored for Forward-Thinkers: Designed for those poised to disrupt markets and lead innovations, this panel will help you stay ahead in the AI curve.
🚀 Unparalleled Insights: Spend an hour with AI luminaries discussing strategies and visionary applications to outpace competitors and drive success.
🚀 Networking Opportunity: Connect with like-minded professionals and innovators, and perhaps discover your next great collaboration.
This is a paid event — attendance is $35 and supports Tampa Bay Masterminds’ mission of fostering innovation and education in technology, with all ticket sales considered donations.
By popular demand, here are the “slides” from my presentation this morning at Civo Navigate Local Tampa, Make Smarter AI Apps with RAG!
Retrieval-Augmented Generation, also known as RAG for short, is an AI technique that combines…
A machine learning model with
A mechnanism for retrieving additional information that the model doesn’t have
…to enhance or improve the responses generated by the model.
At this point, you’re probably thinking this:
This talk runs from 11:15 to 11:30 a.m., which is just before lunch, and I’m not at my cognitive best. Can you explain RAG in an easy-to-digest way, possibly using Star Wars characters?
I’m only too happy to oblige!
Consider the case where you ask an LLM a question that it doesn’t “know” the answer for. The exchange ends up something like this:
Tap to view at full size.
With retrieval-augmented generation, you improve the response by augmenting the prompt you send to the LLM with data or computation from an external source:
Tap to view at full size.
Because RAG provides additional information to the LLM, it solves two key problems:
Tap to view at full size.
Here’s a lower-level view of RAG — it starts with the cleaning and conversion of the supplementary data:
Tap to view at full size.
Once that supplemetary data has been cleaned and converted, the next step is to convert it into small chunks of equal size:
Tap to view at full size.
Those chunks are then converted into vectors. If you’re not really into math but into programming, think of vectors as arrays of numbers. Each of the numbers in the vector is a value between 0.0 and 1.0, and each vector typically has hundreds of elements. In a diagram below, I’ve greatly simplified the vectors so that they’re made up of only three elements:
Tap to view at full size.
The whole process of cleaning/converting, then chunking, then embedding is called indexing:
Tap to view at full size.
Now that you know what’s happening “under the hood,” let revisit the RAG diagram, but with more detail:
Tap to view at full size.
Here’s what’s happening:
Luke asks the question: “Who built you, Threepio?” That’s the query.
The query is converted into vectors.
The “vectorized” query is compared against the vectors that make up the supplementary information — the vectorstore — and the system retrieves a small set of the vectors that are most similar to the query vector.
The query vector and the supplmentary vectors from the vectorstore are combined into a prompt.
The prompt is then sent to the LLM.
The LLM responds to the prompt.
That was the “hand-wavey” part of my lightning talk. The rest of the talk was demonstrating a simple RAG system written in Python and running in a Jupyter Notebook. If you’re really curious and want to see the code, you can download the Jupyter Notebook here.
My session was part of Austin Forum on Technology and Society’s third annual AI April, a month of presentations, events, and podcasts dedicated to AI capabilities, applications, future impacts, challenges, and more.
Here are links to the video and supplementary material for the presentation:
At the upcoming Civo Navigate Local Tampa (taking place Tuesday, April 16 at Armature Works), I will be giving a 15-minute lightning talk on RAG — Retrieval-Augmented Generation — and how you can use it to make your AI apps produce better results.
And by “better,” I mean:
Able to use data that’s newer than their last update
Incorporating information that they wouldn’t otherwise have
Using the content of a document that you provide
Able to incorporate data from a database
And yes, there will be Python code, which I’ll run live for your viewing pleasure, explain for your understanding, and give to you for your own use!
My presentation will be one of several that you’ll be able to catch at Civo Navigate Local Tampa, and best of all, you can register for the conference for the low, low, low price of…
Next Tuesday, April 2nd at 6:15 p.m. Central / 7:15 p.m. Eastern / 23:15 UTC, I’ll lead an online introductory session for people who to dive into AI titled AI: How to Jump In Right Away.
My session is part of Austin Forum on Technology and Society’s third annual AI April, a month of presentations, events, and podcasts dedicated to AI capabilities, applications, future impacts, challenges, and more.
My presentation will start with a brief history of AI, as well as the general principles of how “old school” AI works versus “new school” AI…
…but we’ll quickly dive into building Sweater or No, a quick little AI application that tells you if you should wear a sweater, based on your current location. Here’s a screenshot of some of the code we’ll build:
This is a FREE online session, so you don’t have to be in Austin to participate. I’m not in Austin, but Tampa Bay, and you can join in from anywhere!
