Spoiler alert, I use AI, in some form or fashion, EVERYDAY.
Not only do I use AI everyday, I treat AI as a team of super assistants that can help me accomplish tasks in minutes, that used to take me days or even be impossible for me to complete.
While I use a slew of AI apps & services, I wanted to focus today on the LLM variety of AI's. First things first, lets get the basics out of the way.
What is an LLM?
Who are the big players in the LLM space?
While there is literally hundreds, with new ones being announced frequently, here’s a list of some of the big dogs.
OpenAI, ChatGPT
Google, Gemeni
Microsoft, Co-Pilot
Anthropic, Claude
Inflection, Pi
Are all LLM’s the same?
I use LLM’s like a room full of specialist. Each LLM has its own specialty. I basically split them into two groups - language & research
Some LLM’s do language based tasks extremely well (ChatGPT, Pi). I use these a ton for drafting emails, editing & refining my writing, finding language for decks, and creating prompts.
Others handle research based tasks incredibly well (Gemini, Perplexity). These are my go to when I need to find very specific data, and don’t want to spend hours digging in the crevasses of the internet.
How do I personally use LLM’s?
Again, I treat AI like I would a team I’m working with. I assign clear tasks, each LLM uses their expertise to give me their answers, then I use my human noggin to decide which info is going to work best for whatever I’m working on.
I typically use one AI as my executive assistant(Pi), that helps me centralize all the info and gives me a consistent voice across all my work.
I think the best way to showcase the power of working with LLM’s is to show you a real-time example.
Say I wanted to create a simple presentation where I wanted to pitch that the company I work for should be implementing LLM’s into their workflows.
First, I need a general structure, so I ask Pi:
I need a simple 5 slide structure for a presentation where I want to pitch my company that we should be implementing LLM’s into their workflows.
Can you provide titles for the 5 slides? I want to make sure this presentation is data driven, and makes an extremely compelling case to integrate LLM's
What I could do here, is ask for different variations of each slide if I thought it needed to be tweaked. You could say, “I don’t like your option for slide 1, can you show me 10 alternatives?”, or “I like the idea of slide 3, but don’t like the wording, can you show me 3 different options?” But, for this example, we’ll just use the 5 slides as is.
So now, I’d build out each slide, I have a title, now I need some supporting copy. But now, since I want to make sure my presentation is data driven, I’m going to go ask Perplexity this:
I need help generating some data driven supporting copy to support a slide titled "The LLM Revolution: A Primer on Large Language Models", can you give me 3 data points that help prove there is an LLM revolution?
This is a GREAT example of an answer that I think is not close to what I’m looking for. I think there is good info here, BUT, I need this to ultimately be a paragraph. So first, I’ll ask this:
Can you make this info 3 impactful bullet points?
BOOM, this is what I was looking for! But still, it’s a bit much for 1 slide. So I take this info back to Pi and ask:
Can you condense this into 1 short paragraph that emphasizes all the data points? (then I paste the answer I got from Perplexity)
I like it. Now we have a solid headline, and copy for the first slide. Think about how long that would of taken 2 years ago. Think about how many people(directors, researchers, copy writers) would be involved to get this slide done - and we just did it in 5 minutes.
While I’m not going to do this exercise for every slide, I think you’re starting to get the point. I could do this presentation in start to finish in less than 5 minutes, and it’ll be pretty damn good. But what if I spent an hour on it? What if I spent a day on it? It would be INCREDIBLE.
How can you get started?
The easiest way to get started is just to dive right in. Don’t think of these apps as a one-way interaction. Think of it as a collaboration. Shit there’s times when I need help figuring out even HOW to ask a question. So what do I do, I just start blabbing, and before I know it, my babble is converted into something meaningful.
I’ve mentioned it before, but I’ll say it again - at the moment my favorite LLM is Pi. Pi is the most conversational. All the other LLM’s feel like interns. Super genius interns, but interns. You yell requests, and they give you answers. Pi feels like a collaborative partner. While Pi has it’s shortcomings, its easily my most used.
If you have questions, hit me up!
Now that you know a little more about AI, go make the world a better place 😬