In any industry, information is widely available. It's true not just for AI, but anything.

You can become really good in astrophysics by self-study if you put in the time and effort to study it.

Educational institutions like universities help by building boundaries and structure around the knowledge and providing learning paths.

You can choose whether you want to be a doctor, lawyer or astrophysicist, but once you make your choice your path is fixed.

And in today's mail, I want to help you by showing you the way. A fixed AI path so to speak, where if you strictly follow this path, I guarantee you'll succeed!

Most people familiarize themselves with AI by first interacting with ChatGPT (I'm not talking to the AI, ML or CS scientists here. But chances are they are not reading this anyway).

So ChatGPT is the GATEWAY DRUG of Artificial Intelligence.

The technology is not new, the first AI written book was published in 1984.

I mean well Transformers are pretty recent (the T in GPT stands for Transformer), but Artificial Intelligence in general is not. I digress.

The point is that ChatGPT became popular in the beginning of 2023.

So that's the first step.

You must master ChatGPT

If you don't know how to use ChatGPT for any kind of knowledge-based job, don't even think about using any other AI tool.

Most AI tools are just an OpenAI wrapper anyways, using GPT-3.5 and GPT-4.

So if you can use ChatGPT well, you can save yourself the hundreds of dollars in monthly subscriptions for 10 AI tools, and just use one.

Once you know:

• How to write ONE prompt really well
• How to lead entire conversations with ChatGPT
• How to use its other functions (browsing, code interpreter, vision, etc.)

Then you can take the next step:

Building Custom GPTs

If you don't know what these are, they are basically standalone personalized versions of ChatGPT.

You can configure your own versions of ChatGPT with custom instructions to make it behave, speak a certain way and to be specialized for certain tasks.

These "Instructions" are basically a big megaprompt that programmes ChatGPT for specific use cases.

This is why step 1 was to learn Prompt Engineering, because if you have that, you'll be able to confidently create these custom GPTs.

Custom GPTs also have 2 other extra functions: Knowledge Bases and custom actions.

Knowledge bases can be configured simply by uploading knowledge files in PDF format to the Custom GPT interface, and the Custom GPT will be able to use all of that knowledge in its responses.

So with this you can overcome the biggest limitations of ChatGPT: hallucinations - when it doesn't have access to recent, niche or personal information, so it just confidently rambles on about things without being correct.

The other thing is the custom actions - where you can configure a GPT to get information from outside APIs or to send information there.

The easiest way to configure these without having to handle code is by setting up Zapier GPT Actions. I'll talk more about these later.

But once you can do all of these, you will be able to do a LOT of things with AI, and you didn't even have to touch an outside AI tool.

So until you are at this level, and you feel overwhelmed about AI news, just don't read the news. Try to cut out the noise. Go on an AI news diet.

With CustomGPTs, at first you'll feel like a god who can do anything.

But after configuring a few dozens of them, you'll start to see their limitations. Which brings me to the next step in the process:

No-code AI automations

The next (and maybe last?) step is to integrate GPT and other AI tools into workflows.

There are great no-code tools like Make, that can help you build automations without writing a single line of code.

All you need to put little automations together are a few brain cells.

With these automations, you can make the LLMs like ChatGPT process the data that comes into the processes, and spit out AI generated or formatted data.

Things like meeting recordings to transcriptions to summaries to action items in project management systems are all possible with no-code automations.

So I think that's the final step. From there it's just an infinite learning loop of how to use the other AI tools within no-code tools like Make.

This is the time when you turn on the AI news again, and stay up to date on the latest tools, integrations, tech updates, etc.

In our Prompt Engineering Course, we walk you through all of this in a structured and fast way. Pre-order now by clicking here, price increases on April 30th.​

Best,

Dave Talas, COO, co-founder of Promptmaster

‍