AI Secrets Revealed

In this memo, we break down perspectives, definitions, and guardrails for artificial intelligence that you may not have heard before.

Perspectives

Everyone believes Artificial Intelligence will change the world, but the question of how it will do that is divided into three camps: It's going to take our jobs, it's going to kill us, or it's going to make our lives better.

The real question is not which one will happen, but rather how do we incentivize the positive ones and disincentive the negative ones? Thus, the work is to ensure it makes our lives better without taking our jobs or killing us.

But first, we need to describe AI appropriately because its PR has been abysmal at best, which resulted in this issue in the first place.

Definitions

AI is math. It's as simple as y = mx + b, the equation for a line, if you remember your middle school algebra. The simplest "artificial intelligence" is:

1. A spreadsheet graph with a scatter plot of dots on it. You can call that AI Training Data.

2. You then right-click and choose Add Trendline, which creates a straight or curved line based on what fits those dots the best. Now you have an equation that represents that line of best fit. If it's a straight line, it will be in the format of y = mx + b. This represents an AI Model.

3. Next, you put new dots (training data) into that line (AI model) and out pops an answer. Congratulations, you are a god. You just created artificial intelligence.

Remember that old refrain?

What am I going to use this math for in the real world?

Now you know. It's one of the most valuable technologies ever created by humankind, and it's sucking up all venture capital investment, consumer attention, data center build-outs, and big tech product development time.

Guardrail 1

Hollywood needs to create compelling stories to keep people engaged. One way to do that is with movies like Terminator, which is more memorable and profitable than Bicentennial Man. As a result, generations of humans have subconsciously believed that AI is inherently evil. But math isn't evil. Humans are.

If I give a murderer a pencil, he will likely stab you with it. If I give Walt Disney a pencil, he will probably entertain your family and make you feel good.

So, it matters who we give this new pencil called AI to and how they use it. Thus, we need appropriate guardrails to give more responsibility to trustworthy people with empathy and strong values. Similarly, we keep AI out of the hands of sociopaths.

But how many companies have a hiring process that tests for empathy over raw skills? Very few.

And how many political elections vote empathetic leaders into office? It's not part of anyone's campaign platform, so there are few, but in time perhaps there will be.

As a result, you shouldn't be scared of math or AI but rather of who you elect to build it, control it, and give your data to. That's the first guardrail.

Guardrail 2

The other primary issue with AI is that it gives fuzzy answers. Some call them hallucinations. Who are the ad wizards who name things these days? Can we please get some values-based product marketers, with an advertising background, who know how to tell the truth about the essence of a thing in a friendly, human way? It seems rare.

As a result, we need to establish protective guardrails on the output of AI models, what those models are connected to, and which things they can do. These balances and checks will help control second-order consequences that we can't yet see. Some organizations call this "AI Safety," which I suppose is a response to poorly executed PR, but it doesn't tell you much about what the definition of safety is in this new world.

That's why we like words such as Guardrails and Permissions better.

Give permission to a trustworthy few individuals and organizations to develop the math that turns into products, and only give certain permissions to these mathematical products that have proven they can control themselves. These can be written into software, much like certain users have admin permissions and other users have read-only permissions.

We like the golden rule: do unto others as you would have done unto you.

If you or your product hurts people, animals, or our environment, it may end up circling back around and hurting you too. Like a boomerang of mathematical karmic energy.

Evergence™

As we move through the convergence of technologies and into the emergence of new business models, it's helpful to stay aligned on which layer artificial intelligence operates in the internet tech stack of today:

• Network layer: people and machines (communication, transaction)

• User interface layer: augmented reality and robotics (sight, sound, touch)

• Application logic layer: artificial intelligence and creation tools (math)

• Database incentives layer: cryptocurrency and blockchain (computers)

Once we do that, we can begin to focus on how these things come together and their impact.

For the AI layer, it's likely the intersection of multiple technologies will create the end-game automation layer: if-then statements, machine and deep learning, large language models, and biologic intelligence in the real world. Each are good for certain things, but taken together, we can unlock Voltron.

Remember, the people and organizations you give capital and power to matter a great deal. Profit potential must be balanced with core values. How do you determine someone's core values? Look at what they stood up for and what they suffered as a result of standing up. Those are their strongest values.

Stay tuned.

--Sean