Like everything in life, change is the only thing you can count on.

And a big change that is coming for all of us is picking up steam. In my opinion, and many others who are much smarter than me, believe it will be the most significant change in human history! 

Artificial Intelligence. 

At this point, you probably have some awareness of what’s going on. You’ve heard the news snippets, but as of May 2023, only about 14% of us have experimented with it, which is a pretty low number now six months in—especially considering that within five days of the launch of chatGPT-3 last November, 100 million users worldwide had signed up, which made it the fastest technology adoption ever!

But how big of a deal is this really? Let’s briefly ignore the scary part of how the sci-fi Skynet version could end humanity and focus on the more realistic and near-term changes impacting both increased productivity and job security—no reason to freak out yet.

In December, I started using chatGPT to help me with my podcast, saving me work hours. I gave it transcripts and had it churn out show notes and social media posts. It blew my mind, and I started to share what I had been learning with friends and family!

I’m going to do the same with you. I’m going to cut through the hype, explore these new tools, and share what I find with you. And as I’m not a computer scientist or programmer, hopefully, it’ll be easier to understand!

So let’s start with what turbocharged AI recently, OpenAI’s ChatGPT. The technology, especially since the release of chatGPT4 in March, is not far off the idea in The Matrix when Trinity downloads how to fly a helicopter in seconds or when Neo exclaims, “I know Kung Fu!” It’s kinda like that. It can unleash your productivity by a significant factor!

Let’s start with the basics. Artificial Intelligence (AI) is buzzing with all sorts of fascinating tools, and one such tool is GPT, and most likely, the one you’ve heard of. But what does GPT stand for, and what does it really do? 

First, GPT stands for “Generative Pre-training Transformer.” This might sound like a mouthful, but let’s take it step-by-step.

The “Generative” part refers to the AI’s ability to generate something new. In the case of GPT, it’s all about creating human-like text. Think of it as a digital writer, creating sentences and paragraphs that resemble something a person would write.

Next up is “Pre-training .”This stage is where the AI learns about language. It’s fed with a massive amount of text data, learning how words and sentences connect and how context changes the meaning. It’s like GPT going to school and learning English but on a massive, incredibly-fast scale.

Finally, the “Transformer” aspect of GPT. Without getting too technical, transformers are a type of model architecture in AI. They help the AI to focus on different parts of a sentence and understand how words relate to each other. 

It’s like GPT having a superpower, allowing it to understand language in a way that most AI models can’t. In the simplest form, it’s the knobs and dials to tweak the quality of the outputs from the AI. If you want to geek out, Google the paper “Attention is all you need” from 2017 and read it.

Put it all together, and you have GPT, an AI model that’s a master of language, capable of creating astonishingly human-like text. Whether you’re chatting with it, asking it questions, or even using it to write a blog post, poem, or song, GPT is transforming what’s possible with AI.

And applications such as Midjourney, DALL-E, Adobe Firefly, d-id, ElevenLabs, and others showcase the versatility and creativity of AI in multimedia. These tools can democratize content creation, enabling more people to express themselves and share their ideas with the world. In time, full-length movie creation may be possible. And this is just scratching the surface.

The race for AI is unlikely to slow down primarily because of how much money is involved. And how that’s going to affect you and me is going to be dramatic. So if you want to stay relevant, it’s time to pony up to the bar and start drinking some AI. 

So, what do you need to know?

First, understand all knowledge-based work will be affected. In the short term, tasks that are repetitive in nature are going to be handled by AI, like answering customer service calls. Later, as the technology improves, everything from law contracts to medical diagnoses to working with a therapist or your interior designer can and most likely will be impacted.

So what should you do? Learn all you can to be relevant. Sitting back and waiting to get gobbled up by AI is not a good strategy now, as at least 80% of the workforce will see at least a 10% impact on what they do in their jobs. And 19% of workers will soon see a staggering 50% of their tasks taken over by AI, according to OpenAI and the University of Pennsylvania researchers.

So how did it get so smart? The models of ChatGPT, including the latest version, ChatGPT-4, have been trained on data that spans up to November 2021​​.

The dataset used for training ChatGPT-4, the latest version from OpenAI, is estimated to consist of a staggering 100 trillion parameters, which is over five times larger than the training data for its predecessor, ChatGPT-3​​.

If you’re wondering what a parameter is, I asked chatGPT4 to explain it to a seventh-grader. Here’s what I got:

Imagine you’re a teacher, and you have a class of students. Each student has different traits, like hair color, height, favorite food, or favorite sport. As a teacher, you need to remember all these details to understand and interact better with each student. Each detail (hair color, height, favorite food, favorite sport) is like a ‘parameter’ in this example.

Now, imagine you’re not just a teacher of one class but a superteacher responsible for all the students in all the schools worldwide. You’d need to remember a colossal amount of information.

That’s what an AI model does when it has 100 trillion parameters. It’s like a super teacher trying to learn and remember a gigantic amount of details to understand and generate text. It’s learning from an immense amount of data so it can do a better job at tasks like understanding your questions and giving you useful answers. 

For ChatGPT-3, the dataset was from five primary sources. Sixty percent of the information was based on a filtered version of what is known as ‘common crawl’ data, which consists of web page data, metadata extracts, and text extracts from over eight years of web crawling.

A solid twenty-two percent of the dataset came from ‘WebText2’, Reddit posts that have garnered three or more upvotes. An additional sixteen percent of the dataset was taken from two diverse internet-based book collections, providing a diverse range of fiction, non-fiction, and a wide range of academic articles.

Three percent of the dataset was sourced from the English-language version of Wikipedia, adding a wealth of knowledge on various topics. Notably, ninety-three percent of ChatGPT-3’s dataset was in English, reflecting the language’s prevalence in the digital world​​.

As for ChatGPT-4, its training data was not only larger, but it also included feedback from users of ChatGPT-3 and feedback from over 50 AI safety and security experts.

That’s a lot of information; with it, some tests were given to GPT4 to see how well it would do.

GPT-4 scored at the 93rd percentile on an SAT reading exam. It also demonstrated its proficiency in mathematical reasoning, earning an impressive score at the 89th percentile on an SAT math test, an 88% score on the LSAT, indicating its aptitude for legal reasoning. And it’s the highest score on the GRE Verbal test, hitting 99%! This highlights chatGPT’s ability to understand and analyze complex written passages.

So it’s pretty smart and will definitely change things for you and me. Stick around my podcast to learn and explore what’s coming and how to use it to your advantage!