AI State Management (AISM)

One of the biggest pushbacks against AI is best articulated as a single question.

So what?

The argument goes something like this:

Cool. So we can summarize things, write stories, make cool images, and even do some business stuff using agents. That's impressive, but is it really worth all the hype?

Like, why do people think this is worth billions or trillions of dollars?

It's a fair point, and I would say—if that's all it were—the hype would be unjustified. But I don't think the list above gets even close to what AI is actually going to do.

Rather than just tell you, "Trust me it's going to change the world!", which sounds like what many said about crypto and a dozen other technologies, I'm going to tell you how and why AI is different.

The Delta Between Current State and Desired State ​

Let's back all the way up to what humans do all day every day, and at every scale.

1. We have a world we live in, full of problems
2. We try to understand that world
3. We try to articulate the world we want to live in
4. We try to figure out how to make the former match the latter

This is true at any scope and scale of human endeavor.

• A college student is unemployed and wants a job, and then a great salary, and then a family
• A city wants to reduce crime and grow and have better student outcomes in its schools
• A general wants to capture X territory and destroy Y enemy capabilities, while minimizing losses
• A country wants a higher GDP, less income inequality, and to eradicate child hunger

All of these require that we can assess reality accurately, see what's causing us to not have what we want, figure out the steps to take to get there, and then take action.

This is a universal problem—and arguably the universal problem. And it's a human thing—not a technology thing.

Stated simply, the central problem is that our current state doesn't match our desired state.

AI's role in this ​

Ok, interesting idea, but what does that have to do with AI?

Everything, actually. But let's forget the word AI for a second. It's a confusing term that means different things to different people.

Let's instead ask a question:

What would a technology look like that could constantly assess our current state, help us figure out and articulate our desired state, and then help us take the actions to make them match?

What would that tech look like? And how much would that tech be worth?

It would be worth trillions, and here's what it would look like.

AISM (AI State Management) ​

Again, keeping AI out of the specifics here, here are the steps:

1. CURRENT STATE UPDATES: Use hundreds/thousands/millions of intelligent entities to constantly gather state on everything you care about. That's: Creating, gathering, and analyzing all available data about a thing, and/or creating new signal about the thing, so that you understand its current state at an acceptable level of detail.
2. CAPTURING DESIRED STATE: Use another set of intelligent entities to help the leader(s) of this entity articulate what they want, and/or infer their desires from their statements and actions. This is a hard problem, because it's not just about what they say y they want, but what they actually want. This is a challenge of psychology, philosophy, and ethics, not just technology. But once we have that, we can articulate it in extreme detail in precise language.
3. CONTINUOUSLY TAKE ACTION TOWARDS DESIRED STATE: Now that we have our current state and desired state, we can then spin up another army—or multiple armies—of hundreds/thousands/millions of intelligent agents to figure out and execute the actions required to affect the change.

If we had trillions of people on Earth that were smart, and trainable, and didn't have to rest or sleep, that we could deploy anywhere with minimal friction, we would use them to do all these tasks for every individual, business, and organization on the planet. It would be great!

But we don't have those people. We have very few, actually, and the ones we have take forever to train, get bored, get tired, seek different roles, move, get sick, die, and are otherwise very transient.

So if we were to put a percentage on this, what percentage of the world's people, teams, companies, organizations, cities, countries, etc., are able to:

1. Assess their current state anywhere near to real time
2. Describe their ideal state in significant detail
3. Take actions to get from 1 to 2

???

This is highly speculative, obviously, but it has to be a tiny fraction of a percent. Essentially zero.

We need trillions of intelligences to pull this off, and we are starting with a rounding error on nothing.

Well...that's literally the human and market opportunity for AI.

Summary ​

1. AI's final form isn't about tech, it's about humans.
2. The best way to think about it—and especially agents—is as a way to manage the transition from current state to desired state for anything we care about.
3. If you're an analyst, or policymaker, or just someone who thinks deeply about this stuff, stop getting hung up on models, and tokens, RAG, etc.

• Think instead about human problems—what humans actually struggle with every day in their lives and at work. People. Growth. Improvement. Business. Relationships. Family. Thriving. Society.

I hope this perspective gives you clarity—as it has for me—on questions related to not just AI, but tech, work, career, and society in general.

Questions like:

• What to work on
• What to build
• What to tell your kids to study
• What to invest in
• Etc.

And ultimately—how to prepare for a world where this type of desire-to-reality transition is possible.