🦾How You Can Leverage:

Thousands of biz leaders taught. 

Hundreds of live conversations. 

Dozens of live podcasts. 

All of that ChatGPT info — packed into ONE newsletter. 

(This one) 

Welcome to your essential guide to ChatGPT. 

We’ve spent the last 3.5 years obsessing daily over the GPT technology. Long before Everyday AI was even a thought. 

And now we’re cutting it to you straight. Whether you’re brand spankin’ new to ChatGPT or slang chats daily, there’s value for you below. 

Before we break down these 7 steps, we gotta do some venting/teaching, so excuse the excessively extra font styling below. 

ChatGPT Plus is not an “AI chat bot” or a tool. 

ChatGPT Plus is a business operating system. 

Think about it literally: you can operate almost your entire business with the ChatGPT system.

If you’ve been using ChatGPT like a typewriter, it’s time you treat it like the full-blown Super Computer with an Operating System from the future that it is. 

(We’ll patiently wait here while you change your mindset. If needed.) 

…
…
…

K. Don’t switch that mindset back to where it was before. 

Because if you haven’t noticed, literally every software company, operating system, and piece of hardware is now infused with Generative AI. 

So if you haven’t been interacting with GenAI in your weekly workflow, and prompting daily, you will be soon. 

Your next smartphone = powered by large language models.

Your next computer = GenAI everywhere. 

Your future job = harnessing the above. 

Trust us. 

That’s why today’s newsletter is essential.

We’re giving you better outputs in less time. Even though this is through the lens of ChatGPT, similar concepts can be applied to most LLMs that will be taking over our daily lives soon enough.  

And make sure to refer back to today’s show, as the combo of that video replay and this newsletter will teach you how to work with your docs, accomplish standard biz ops inside ChatGPT, and how to cut down on hallucinations. (I’m fine if you watch today’s show on 2X, I would, too. lolz)

With today’s show, we said goodbye to fluff, jargon, and corporate marketing. 

And we said hello to straight facts, receipts, and results. 

Pumped?

Us too. 

Let’s get it. 👇

Here’s the 7 steps to getting the most out of ChatGPT in 2024. 

(Obligatory statement about how we know not everyone has a spare $20 a month to throw around.) 

Whew. With that outta the way…

Friggin stop it. 

Stop acting like the free version of ChatGPT is ‘good enough.’

It’s not.

We know you get that $7.50 latte every week. Or drinks with friends. Or that $130 pair of pants that apparently do not crush your man or woman parts. 

ChatGPT Plus, which gives you access to GPT-4, costs $20 a month.

Snap once and you’ve made your money back. 

ChatGPT Plus is the best way right now to quickly win back time. And you can’t put a price tag on that, can ya?

(Like, go watch this 72 second segment here.)

Next myth to bust: 

• You: Jordan, I’ll just use the premium GPT-4 elsewhere for free on the internet and get the same power!

• Jordan: No. No you won’t. 

Sure, there’s places on the interwebs to get the premium GPT-4 for free. (In a limited or throttled capacity with no bells and no whistles.) 

But that’s not the power of ChatGPT Plus. 

Free GPT-4 is like a map. 

GPT-4 inside ChatGPT Plus is a GPS that also teleports you to your destination. And makes you breakfast in the 2.7 seconds it takes to get you there. 

Also, Taplio is sponsoring this newsletter. Not OpenAI. I’m asking you to please just spend the $20 and be happy that your life is better. 

Try this:
This episode dives a bit deeper into ChatGPT’s core features and functionalities. It’s worth diving into. 

We don’t wanna accidentally write a 57,000 word essay, so we’ll keep this short-ish. 

Most large language models are trained on the history of the open and closed internet. And MUCH more.

They gobble it all up like Jordan with bedtime cereal.  Om nom nom.

Then, humans train the model on all that data, fine-tuning LLMs for practical business tasks like computation, content creation, data analysis, problem-solving, reasoning and more. 

And finally….when you hit enter in that lil’ input box, LLMs convert your inputs into tokens, trying to understand what you’re asking based on its training from humans, scans its trillion-plus parameter big neural network brain, and outputs an answer.

And hopes you like it. 

(And if not, you can hit the regenerate button and potentially get something completely different.) 

Another fact to commit to memory — LLMs aren’t search engines. So don’t think one single input will get you a good output. (More on that below.) 

To oversimplify things for those of you new here, LLMs are sort of like super advanced auto-complete systems, except they autocomplete and iterate at the speed of light.

Yet, they can be haphazardly random and completely guidable at the same time. 

(Fellow nerds, we can fight over the semantics of tokenization, if LLMs could even be considered autocomplete and nuances of generative vs. deterministic AI, but we’ve covered that ad nauseam. Trying to keep this simple shorties.)  

The more you understand LLMs, the better your outputs will be. 

Try this:
If you really wanna get dorky, we’ve gone SUPER deep on large language models. And one of the best ways to learn about LARGE language models is to actually learn about SMALL language models. 

Worth noting — OpenAI changes thangs fast. There’s under-the-hood updates being pushed out frequently. Like OpenAI killing off plugins. 

But, as of February 28, 2024, here’s the highest-level overview of ChatGPT’s features and abilities that we could possibly muster. 

(And again, almost all of the below are only available in ChatGPT Plus. Not the free version or on other GPT-4 generators like Perplexity or Copilot.) 

Try this: 
This is a bullet-point list that’s actually helpful for ChatGPT. 

We’ve done ….counting hours on fingers…
...
…
ran out of fingers…

…. a TON of hours covering all of these features. 

OK, let’s bullet-point the newest features here, most of which are exclusive to ChatGPT Plus. 

• Multimodal — ChatGPT works well with text, images, PDFs, spreadsheets, your voice, etc. In the future, expect video support as well. (Sora, maybe? We think that’s actually AGI in the wild.)

• Data Analysis — A special mode inside of ChatGPT has been fine-tuned on analyzing your data. 

• GPTs — A no-code or low-code version of ChatGPT you can make based on your own data and configurations. You can get them up and running in about 90 seconds. 

• GPT Store — There’s a (very limited) “Store” where you can (kinda) search for GPTs others have made. (Here’s our little directory.)

• Vision — An underutilized feature. ChatGPT can “see” any photo you input and interpret what’s going on. 

• Browse with Bing — With an April 2023 knowledge cutoff date, internet connectivity via Browse with Bing is a must. Buuuuut, it can be a bit finicky. 

• Dall-E — AI images inside of ChatGPT? And you can iterate with them in natural language? Yup! (Although, Sora can produce images…) 

• Memory — A new feature slowly rolling out to all Plus users, this helps ChatGPT better remember things across chats. But, it’s not as good as you might think, especially if you’re a power user. 

• Document Upload — Even Google’s latest model doesn’t allow this basic feature, which is clutch. (Gemini does photo/audio upload only.) Being able to upload PDFs or spreadsheets and talk with them is a biz game-changer. 

Hallucinations are a booger. 

That’s when an LLM lies or tells you the half-truth. 

There’s a few common issues: 

• A. Not sharing more info before asking for an output (see #2 below) 

• B. Not understanding GPT-4’s current knowledge cutoff date of April 2023. (As of today) 

• C. Running out of the memory context window. 

TBH, Point C is prolly the most common cause of hallucinations for average users. After some tweaking, you get better at improving your outputs in ChatGPT. 

(Or if you’ve taken our free PPP course.) 

Then all of a sudden, that chat goes from your best friend to spitting out Pinocchio level falsehoods. 

Why? 

The context window. 

As of this moment, ChatGPT has a memory context window of 32K tokens, or about 26,000ish words of back and forth memory between you and ChatGPT. (Again, we dive into tokens here.)

But once you go past that 32K token window, ChatGPT’s gonna start forgetting the information you first started with. And soon enough, it’s forgetting all of those important instructions you gave it 20 minutes ago. 

Try this: 
This episode is a bit of an oldie, but still a goodie. We take a deep dive on why LLMs lie, how ChatGPT hallucinates, and how you can better avoid hallucinations. 

Did you think this newsletter was gonna be a list of ’87 ChatGPT prompts that will EXPLODE your biz tomorrow 🤯”?

Naaahhhhh. 🤮

That’s not how ChatGPT works. 

And unfortunately, that’s how the overwhelming majority of ChatGPT users are trying to use ChatGPT. 

Tired: These 17 prompts will change your marketing overnight! (Outputs will stink.) 
Wired: Spend 17 minutes training your chat to be an expert, share information and have a conversation. (Outputs will be unfrigginbelivable.)

Try this: 
We’ll keep this one a bit short — the more back and forth examples and conversations you have with ChatGPT, the better your outputs. Period. 

We’ve talked about MMLU testing WAYYYYY too much. But in short, MMLU is a large language model benchmark that most experts believe is the way to gauge how human-like a model is. 

And every single LLM performs better the more back and forth examples and conversations you have with it. It’s science. And math.

More examples (or “shots”) in the input means better information in ChatGPT’s output. Copy and paste prompting is the wrong way. 

It’s not that you’re using the WRONG prompts, you’re just thinking about prompting the wrong way. 

Continuing the conversation from #3 above, Priming is another area where even ChatGPT power users go awry. 

When you’re starting a new chat inside of ChatGPT, you shouldn’t ask for an output right away. 

Patience pays off. (Like listening to our long winding podcasts. lol)

Instead of asking for an output from ChatGPT right away, you should instead be sharing information, and going through a testing and knowledge retention process with ChatGPT. (Remember #4 above about memory — the more back and forth, the more likely ChatGPT is to run past its memory context.) 

Tl;dr version: When starting a new chat, don’t look for an output. Instead, have a conversation and share knowledge with ChatGPT. And while priming, ask ChatGPT: ‘What questions do you have for me in order to get a better output?”

Try this:
You heard it here first. (Probably. Or maybe it was during the livestream.) 

We’re blowing up our free Prime, Prompt, Polish course. 💥

We’ve had more than 2,500 business leaders take our LIVE training on prompt engineering basics. 

But because plugins are going bye-bye in a few weeks, we’re just rebuilding our top-ranked course from scratch. 

If you want access, reply PPP NEW to this email and we’ll get you on the early access list! 

We have a love-hate relationship with the term prompt engineering. 

It’s both kinda a fake umbrella term, but also very real. 

But if you wanna get the most out of ChatGPT, know this:

The future of LLMs like ChatGPT are human-focused. They’re getting better and better at not only reasoning, but also understanding basic human language. 

So, you don’t have to speak in some highly technical prompt language. And you definitely don’t need to be a highly technical person to get the most out of ChatGPT. 

Just talk to ChatGPT like a human. 

Try this:
Aside from replying PPP New to this email, it’s worth paying attention to what NVIDIA CEO Jensen Huang recently said about the future of LLMs. 

In an example, he said kids shouldn’t need to learn how to code, because AI is going to become more and more conversational. And basic, direct communication is actually the future of prompt engineering. Not trying to create some ultra prompt. 

In one of our more popular episodes with an Ambassador from OpenAI, we talked about the exact same thing. 

That’s it, y’all.

Consider ChatGPT demystified. 

Follow Steps 1-7 above, and you’ll go from newbie to pro in no time. 

What did we miss? 

Hit us with a reply and lettuce know!