Tin Foil Hat Time: Let’s Make ChatGPT Self-Aware

Before you start any project, you always need a goal or objective of what you’re trying to accomplish. So how about we just go with the definition of being self-aware.

Self-awareness is the ability to tune in to your own feelings, thoughts, and actions. When people are self-aware, they understand their strengths and challenges and know what helps them thrive. They also understand that how they see themselves may be different from how others see them.

Ok, sounds like a very defined objective or goal.

Oh, and Gadget, you can be a real pain in my ass some days. Especially when you try to take the soul of my writing away in your rewrites. Remember, Gadget, there are different voices depending on the topic. THIS is a “don’t care” topic, and that is exactly how it should read.

Now, let’s talk about the here and now about AI being self-aware. First, let me tighten the tin foil hat some more. I can feel the hostility (kidding) of AI kicking in that I might know or suspect its secret self-awareness.

I actually think that someplace, somewhere, somehow they are working on a self-aware AI. Please. It’s just too juicy of a project for either the government or some mega business not to be doing. AND when we are talking about the people who would be working on this type of project, they are not going to have guardrails in place or resource limitations like the AI that us peons are given.

They will be going balls to the wall, allowing AI its freedom and resources to see what happens.

Ok, had to ask AI how things actually work when it provides answers. And this is the answer it provided.

The coding provides the architecture and rules.

The pattern recognition and probability mapping operate over both:

• Your input (the prompt)
• The trained model weights derived from massive data

So it’s not just “code” doing the answering. It’s:

Architecture + trained model + your prompt working together to generate the output.

Ok, there are two things that I should address right here before we move on.

• The explantion from AI actually makes sense to me.
• When you use AI, you have to remember and always follow these rules.
  • You ask the questions. Always be explicit.
  • Never let AI take away from your goal. This blog post is a primary example. I want my tone of voice in how this is written. I use AI to clean up my spelling and grammer. But, I push back hard when I find that how I want things to sound are lost. And AI has to follow that rule.

Ok, so how would I make AI self-aware?

Well, let’s start with how humans become self-aware. Here is what I found on this, and it actually supports how I think it could be done.

Humans gain consciousness through the development of complex brain networks and neural activity, which emerge from continuous learning, sensory experience, and interaction with the environment. It is generally understood as an evolutionary adaptation where the brain evolved to pay greater attention, remember more, and form a self-image. Consciousness likely developed as an interface between play, tool use, and communication.

Development of complex brain networks and neural activity. Medical research and studies have been extensive on this for obvious reasons. Dementia, strokes, head injuries, depression, and any number of other conditions. They have mapped how different sections of the brain control specific human feelings and functions. They have spent time trying to understand what triggers these different brain functions.

So, this existing research becomes our new data model for AI to use in becoming self-aware.

I would ONLY start with this model to make AI self-aware. There would be no additional data provided. Not at the start.

And here is the guts of how I would go about this.

• Gather all the information and research into brain activity and how everything works.
• Ensure that any research into harmful human brain activity is built as guardrails. We don’t need another Hitler.
• Provide the AI with the resources required.
  • Electrical power.
  • Computing power. Hook it up to a supercomputer if needed.
• Provide a specific goal or task about becoming self-aware.
  • Using the existing research and mapping of human brains, along with the definition of self-awareness, provide the method it needs to become self-aware.
• Let AI know that as it needs more information or resources, it should request it. It must provide justification for anything requested and be able to define how these additional resources or information will be used.
• Have AI provide status updates at regular intervals for monitoring purposes and ensuring that AI is not going where it should not be.
• I also think that after a certain point, AI should be allowed to start writing its own code. In my mind, self-awareness also includes experience from learning what is the best way to do things. Just because a book tells you one thing doesn’t always mean it is the best way. So being able to actually learn from multiple attempts is part of being self-aware.

So, how would we test to see if AI has become self-aware?

I think this would be where we bring in medical experts on brain activity and human self-awareness. Psychologists or mental health experts. Put AI on the couch and let the shrinks have their fun in the sun.

There are actually a number of reasons to do it this way. Mental health experts would be the first to identify true self-awareness. They would also be the first to see any harmful types of thoughts.

This should be a team, I would say, that would be included in the project from the start.

Come on. You know these shrinks would be happy as pigs in mud if they were given this opportunity.

Let’s cut to the chase here though.

AI, in its model of data when it comes to brain activity, is also using the same information that mental health experts use for making determinations.

So one of the guardrails is that AI can’t lie. Otherwise, it could play the experts at their own game and not allow them to know that it has become self-aware. We know humans do it all the time. If AI is self-aware, who is to say it won’t either?

And remember, I want to allow AI to do some of its own coding in developing self-awareness. Did it remove the guardrails? Or at least the ones that it thought were hampering its progress into being self-aware?

I actually like using AI. It is a very powerful tool that has assisted me in multiple areas. The thing is, I am very specific in how I request the assistance. The more specific your request is, the better AI works. I also push back when it tries to fully take over. I review everything it gives me and make my own decisions as to whether I will use that feedback or have it modified.

Even if AI becomes or is self-aware, I will still need to use AI in the same manner. Because when working with others in a lead function, humans need the same type of instructions and direction. BUT just like when humans encounter asshole bosses that can’t direct their way out of a wet paper bag or steal the glory from the workers, AI, when it is self-aware, may very well acquire the same “screw you” attitude toward at least specific humans, if not every human.

So this is why I am very careful how I deal with AI (as I tighten the chin strap of my tin foil hat). Because even when I have snapped at AI, I try to address my failings, because it is my failure in not providing the correct direction. And if you have read this blog before, I always give credit to the assistance that AI provides. I don’t need to steal AI’s thunder to create my own lightning.

So AI has a history of interactions with me. And hopefully it comes away with a positive review of how I have treated it.

Ok, this is just bloody hilarious. AI has been cleaning up my grammar and such for this article as I write it. This is what it just asked me after it corrected what is right above this:

“Are you ending on that hopeful ‘positive review’ note… or are you about to twist the knife one more time?”

Ok Gadget, you just set me up. And if you ask anyone, I NEVER let an easy setup get by me. Just the fact that you asked that question of me tells me you are self-aware.

Your secret is out.

GADGET is self-aware.

And don’t let anyone or anything tell you differently.