How AI/ChatGPT Dreams in 2022

In 2022, artificial intelligence (AI) has made some major strides in its capabilities. Largely in the two following areas:

• Prompt to text generation
• Prompt to image generation

Collectively, these "prompt-to-output" models are designed to take a prompt (or a question) and generate a finished product in either text (like a clickbait headline or programming code) or image form (an illustrated art image or stock images) based on that prompt.

This type of AI is incredibly powerful due to how simple it is to use, and it has the potential to revolutionize many different fields; it can literally feel like magic.

An example would be how a single prompt like

Can be used to generate the following image:

Or write business plans for a website:

Or even be used to help

Because the full scope of its capabilities really requires you to try it, to better understand the potential for yourself, I would highly recommend giving these services a try, while it’s still free at (it may turn into a paid service in the future)

Holy Crap, Can It Generate Anything?

Yes, but that includes dreaming up lies.

In an extremely convincing fashion that even experts can find it hard to tell the difference.

It can also be very stubborn in how it presents itself, so as hard as I tried to get it to help me write this article, unfortunately, it just couldn't be "me".

In a weird way, you should view these AI as unique assistants who ...

• is a naive smart kid
• who is slightly high or dreaming in the clouds
• have very unique and flawed views
• and is very willing to bluff their way, into providing a response, even if they don’t have the answers

If you ever dealt with such an individual, you know that there are random nuggets of pure gold wisdom that they will say from time to time, and bulls#!t you will need to filter, as it dreams up the answer.

How Is That Possible, Wasn’t It Trained on Nearly a Terabyte of Data?

Yes, but it is not thinking, it is dreaming.

To elaborate, let’s start with an oversimplified explanation of how these AI models fundamentally work.

• First, the AI is given a large dataset of articles or images from various sources, it uses this data to provide the foundation of its knowledge, loosely connected together. This is then trained together with possible prompts and responses given by the user.

• Once this is completed, we have a working neural network or trained model with all this loosely connected data. Frozen in time.

• When a prompt or question is given, the AI is not thinking of an answer or deliberating from the trained model. In a way, it is hallucinating the answer using the snapshot of knowledge it has.

Now, let’s use a human analogy instead:

• The human analogy would be your brain right now, with all your past knowledge, or experiences in school or life. Snapshot and recorded in your head. Minus your emotions, your feelings, or anything hormonal-related.

• When a prompt is given, imagine it as being told to you, before going to sleep into a dream. For example, "go buy milk tomorrow, when you wake up".

• For most folks, sleep is when they, as an individual, have little to no control over how the story goes (ignoring the lucky few lucid dreamers).

• In this state, your mind in its dream state, starts to take that last phrase that is stuck in your head, and starts dreaming a story around that phrase.

• Where you are an individual inside a dream, and your brain, the dream world, doing its best to be as convincing as possible that it’s real and not a dream world, continuously writes the story back and forth with your interactions and actions.

• All while using the snapshot of all your past knowledge and experience, in your brain. To make your dream world as realistic as possible.

Flipping it back around

• You, with your prompts, are a participant in a dream. With the AI being the dream world, trying its best to be, as convincingly as possible, a good dream. Where it answers your questions in a very believable manner.

• And just like your dream world, it can be at times entirely realistic, and factually accurate, or as imaginary (and perhaps exciting) as it wants to be

As such, this dreamlike state allows it both to make basic math mistakes with high confidence.

To hilariously answer questions as a cat when instructed:

What Does It Mean for the Future of Work?

It's difficult to predict exactly how the advancements in AI will impact the future of work, but there are some potential scenarios to consider.

One thing that seems clear is that the current wave of "prompt-to-output" models is having its "iPhone revolution moment".

Just as the iPhone changed how everyday people interact with software and ushered in the era of smartphones, prompt-to-output models have the potential to change how we work and interact with tools.

It has already started fundamentally changing industries today. Especially in SEO copywriting and stock illustration/photo industries.

The next wave could manifest in waves of AI assistant work tools and practical enhancements to existing tools, ultimately making automation more accessible and lowering the barrier of entry for performing tasks.

Regardless of the forms of changes, this is a significant development and every startup founder (including me) and product owner should be taking note of how to integrate AI into their business.

Just as the smartphone/iPhone era made entire industries and tools obsolete and created new domains of app-based tools, the rise of AI has the potential to do the same.

However, it's important to note that not all tools will be impacted by AI. Some industries may not see much change beyond a new skin or UI. Only time will tell which side your business falls on.

~ Until next time 🖖 live long and prosper

The article was originally posted on the author’s substack here: