Happy Wednesday, Normal People!

This is Brady, but as you probably already noticed, today is a very special edition of aifnp – we have our very first Guest Writer joining us! Jacob is a close friend of mine who also happens to be in graduate school for machine learning at UT Austin (aka, way smarter than me). What he’s shared is not only genius but also hilarious – the best combo IMO. 🤝 

But before we hear from Jacob, it’s time for another AIFNP POP QUIZ featuring today’s AI vocab term: 

Reminder: We’ve got a little something set aside for anybody who gets all quiz answers right answer this week 🤓 

LIMITED LIABILITY MEGACHURCH

by Jacob Badolato

“Law degree”... “Left-leaning man”... “Limited liability megachurch”... What could I possibly be talking about? Yep, you guessed it! Large language models (LLMs). Fascinating, right!? Those opening words were some of my family and friend’s best guesses (okay, maybe not so much) as to what an LLM is. My name is Jacob Badolato. I am a software engineer and current graduate student in computer science, and I will be continuing today’s installment of AI For Normal People. These days, it is becoming more and more important to understand, at least in theory, the emerging lingo of the AI space. 

As Brady hinted at yesterday, this week we will be going over some basic terms to catch you guys up to speed on the ever-growing lexicon of the AI world. So, after no further ado, let’s dive into what exactly an LLM is. As the name implies, an LLM is a type of language model that is large. How “large”, you may be wondering. Well, GPT-4 is estimated to have been trained on over 1,700,000,000,000 (1.7 trillion) parameters. For our purposes, a parameter is a piece of numerical data that is “learned” by a model during the “training” process. Essentially, an LLM is trained on a dataset, so that it can make an educated guess based on a specific user input and the previously generated tokens (for a deep dive on this, I recommend the following paper). The reason this is significant is, as Zander mentioned this past weekend, a model can “experience” things much faster than a human can. If I were learning how to classify a 🐶from a 🐱as a 2 year old, I may only have ever seen 7 🐶and 12 🐱in my life. An LLM, however, could “see” 7 million 🐶and 12 million 🐱in two minutes. Obviously, you can see why this has benefits. We can harness this knowledge and use it to our advantage in a wide array of applications. 

Now I will discuss some of these potential applications. For one, there is ChatGPT. While ChatGPT is not the only LLM, it is just the most well-known by the general public. Some places where LLMs may impact the world around us (or already are impacting it), are as follows:
Software development, automating emails, completing this sentence for me, revolutionizing legal document review, streamlining complex financial planning, and customizing news feeds. Yes, ChatGPT really finished that sentence for me. Pretty neat, huh! 

So, there you have it! You have now learned the essence of what an LLM is, so that the next time you hear someone discussing ChatGPT or AI, you can have a little more context as to what they are discussing. If you liked this installment of AI For Normal People, be sure to keep checking back the rest of the week as we cover some more AI vocabulary. 

Stay Normal! 

Jacob

Subscriber Count 👉 314 🎉 

We did it!!! In the past few days we have BLOWN past our goal of 250 subscribers by the end of January… CAN WE DO 500?! 🚀🚀

Do you know any other normal people?!
Share this sign-up link with them today!