Demystifying LLMs for the Business Thinker

Everyone’s heard of ChatGPT, but fewer of us (millennials and up) have used it, and even fewer of us are familiar with the term “LLM” or Large Language Models. In my discussions with business leaders, I find myself asking the same few questions to level-set their understanding of Generative AI. It goes something like this:

1. Have you used ChatGPT?

2. Do you know what an LLM is?

3. Do you know what LangChain is?

I decided to conduct a poll of my LinkedIn network (a highly scientific study) and found that 63% of respondents did not know what an LLM is.

If you are a non-technical business leader trying to understand Generative AI, use this as an introduction to LLMs so you are better positioned to capitalize on their immense power.

What is an LLM?

Large Language Models (LLMs) function as software that internalizes tons of information like books, websites, transcripts, research papers, etc. The model uses all this knowledge to understand language and generate content that aligns with our expectations of meaningful communication. Most impressively, the LLM is not just copying and pasting; it's actually creating new sentences based on patterns and structures it's learned. This capability moves them beyond static repositories and transforms them into dynamic creators of content. As you engage with an LLM, you're not limited to just retrieving existing information; you can actually generate fresh insights and perspectives.

How do LLMs Work?

LLMs simply look at a piece of text and guess what word comes next. For example, if the text is "I enjoy eating...", the LLM might guess "pizza." It does this not just for one word, but for whole sentences and paragraphs. Over time, and with a lot of expensive practice or “training” (we're talking billions of examples of natural language, like the whole internet), the LLM gets better at guessing, and the result is coherent and meaningful sentences, paragraphs, or even entire essays!  The process of generating language as a response based on context is called inference.  

Understanding ChatGPT

ChatGPT is an application on top of an LLM called GPT (Generative Pre-training Transformer).  ChatGPT gives users an easy interface to interact (or chat) with GPT, the LLM.  GPT was built by OpenAI, which is heavily backed by Microsoft.  Check out our Guide for using ChatGPT in the business setting: Yeager.ai's ChatGPT Guide.  

Here are a few other ChatGPT-like applications built on top of different LLMs: https://claude.ai/login by Anthropic.ai, Bard by Google https://bard.google.com/

So What?

LLMs are changing the way we interact with computers. Instead of writing commands or precise formulas, now we can just chat with computers as we'd talk to a person, aka in “natural language.” This saves us time and enables non-technical people to build powerful applications seamlessly.  

Also, LLMs can serve as our brainstorming partner, tackling the daunting “blank page” problem.  Let's say you're tasked with writing a product launch announcement or a press release, and you're finding it hard to begin, or you're stuck on a section or simply not convinced by your current draft. You can give the LLM context about your project, ask for tailored customer benefits or messaging specific to your market, and it will churn out valuable suggestions. You’ll often find that these AI-powered insights can be the catalyst for new ideas.

LLMs can be used for far more than content creation.  With their ability to retain massive amounts of information and process natural language (“natural language processing” or “NLP”), LLMs can be used to write code, provide education and learning tools, data summarization, customer support, and much more. 

In this post (https://medium.com/yeagerai/is-your-cto-using-vector-databases-94809147c029), we discuss how businesses are using LLMs to capitalize on their unstructured data using something called a Vector Database.

LLMs the Mystery - Be Careful!

There are still several areas of uncertainty with LLMs. One key mystery is how exactly LLMs generate certain outputs - their internal operations can often be a "black box" where inputs go in, and outputs come out, but the exact internal processing is unclear. 

While powerful, this technology is still early, and it is not 100% reliable.  LLMs make mistakes called hallucinations.  We strongly recommend that you do not simply copy and paste results from an LLM as a final work product.  Instead, use the LLM as a tool to unblock your mind, followed by a human process of internalizing, verifying, and refining the generated content.

Recently, two lawyers faced punishment and possible disbarment after blindly using ChatGPT in a court case.  The LLM “hallucinated” fake court cases which the lawyers cited in court (https://apnews.com/article/artificial-intelligence-chatgpt-courts-e15023d7e6fdf4f099aa122437dbb59b).  The lesson here is not that we shouldn't use Generative AI, it’s that we should use it the right way.  Had these lawyers taken our advice and tried to internalize the results before using them, they would have quickly found these cases didn't exist.

Also, LLMs are not good at everything.  They generally don't perform well with calculations or anything that requires deterministic answers.  LLMs provide probabilistic answers based on their training.  If you want to learn more on this, check out our post on how we use LLMs as a subconscious mind: https://medium.com/yeagerai/leveraging-llms-as-a-subconscious-mind-c57ee97e7bcd

Lastly, when you use closed-source LLMs, the LLM provider has access to all of the data.  In other words, if you are using ChatGPT or GPT via an API, you are sending your information to OpenAI.  Because of this fact, we see open-source LLMs playing a major role as companies will want LLMs hosted on-prem for their sensitive data.

Conclusion

We are at the very beginning of a major technological transformation.  In a very short period of time, most of our professional and personal productivity will be powered by Generative AI-based applications.  LLMs are the engine of Generative AI, so understanding them is essential for business leaders.  These models are transforming how we interact with technology, shaping future business operations, and opening up new possibilities for innovation. 

Leaders who understand this technology will have a strong competitive advantage, better equipped to identify opportunities for its application.  

Learn More About 🧬🌍GenWorlds

Demo: https://youtu.be/INsNTN4S680

GitHub: https://github.com/yeagerai/genworlds

Docs: https://genworlds.com/docs/intro

Discord: https://discord.gg/wKds24jdAX

Blog: https://medium.com/yeagerai

About Yeager

At Yeager.ai, we are on a mission to enhance the quality of life through the power of Generative AI. Our goal is to eliminate the burdensome aspects of work by making GenAI reliable and easily accessible. By doing so, we foster a conducive environment for learning, innovation, and decision-making, propelling technological advancement.