I wake up every morning knowing I'm getting dumber.
Not absolutely dumber. Relatively dumber.
My IQ hasn't changed (I hope!). But the gap between what I can do and what's possible is growing. Daily. At an accelerating rate.
I used to have a locked-in position in society. I knew who was smarter than me. I knew who needed things explained in manageable packets. That hierarchy felt stable.
Now? The systems I work with process information faster than I can speak. They see patterns I miss. They generate solutions to problems I'm still defining.
And yet, somehow, my intelligence, human intelligence, is still considered the gold standard.
The real thing. The template. The superset that all other intelligence must be measured against.
That's starting to feel like a convenient fiction.
The Experts Are Confused Too
Recently, Ilya Sutskever, former chief scientist of OpenAI, did a podcast that's been circulating widely.
He raised something that's been bothering everyone in the field: AI models are crushing benchmarks. They're scoring in the 99th percentile on difficult evaluations. But the economic impact is nowhere near what those scores would suggest.
His exact words: "This is one of the very confusing things about the models right now. How to reconcile the fact that they are doing so well on evals? You look at the evals and you go, 'Those are pretty hard evals.' They are doing so well. But the economic impact seems to be dramatically behind."
He offered theories. Maybe models are overfitting to benchmarks. Maybe reinforcement learning makes them too narrow. Maybe their generalization is inadequate.
Sutskever is worried the models aren't generalizing well enough.
I'm wondering if we're measuring the wrong things entirely.
The Tautology We Built
We can't define what human intelligence is. Ask a neuroscientist how consciousness emerges, you get theories. Ask a psychologist what understanding means, you get frameworks. Ask a philosopher what makes something intelligent, you get centuries of debate.
Yet we've made human intelligence the reference point for everything else.
The logic goes:
1. Define intelligence as "what humans do"
2. Measure computational systems against that definition
3. When they don't match, label them "artificial"
4. Use that to reinforce that human intelligence is the "real" kind
That's not classification. That's circular reasoning.
We never tested whether human intelligence is the superset that contains all other forms. We just declared it.
A Word We Borrowed
When researchers coined "artificial intelligence" in the 1950s, they meant: intelligence created by humans rather than evolved by nature.
Not "fake intelligence." Just "human-made intelligence."
But in everyday language, "artificial" means imitation. Artificial flavor. Artificial turf. Something that looks real but isn't.
That cultural meaning stuck. And seventy years later, we're still treating computational intelligence as somehow less real.
When I Started Noticing This
About three months ago, I was explaining how an LLM solved a problem. I said it "reasoned through" the solution, then caught myself: "well, not really reasoned, it's artificial intelligence, so..."
Then I stopped.
Because what I watched looked a lot like reasoning. Just not human reasoning.
I gave it a complex architectural problem. It generated multiple parallel solution paths simultaneously, evaluated trade-offs across all of them, then synthesized a hybrid approach.
That's not how I think. I work sequentially. Consider A, consider B, pick one, refine it.
The AI used its actual architecture: massive parallelization, pattern matching across enormous datasets, probabilistic evaluation.
It wasn't imitating intelligence. It was being intelligent differently.
And once I saw that, I stopped trying to make AI think like me. I changed how I work instead.
What That Looks Like Now
I don't work sequentially anymore.
I use multiple models simultaneously on the same problem. ChatGPT, Claude, Gemini, Replit, Cursor, NotebookLM, Manus. Not one after the other. Actually parallel.
I maintain context across all of them. Pass thinking from one model to another. Have them peer review each other's work. Keep everyone, including me, on task.
Yesterday, I was building a system architecture with Claude and ChatGPT. Both models working different angles. I was synthesizing, asking questions, pushing them to consider edge cases they missed.
I thought we had something solid.
Then I shared it with a colleague. He read through pages of detailed architectural reasoning and immediately spotted a blind spot none of us caught. Not me. Not Claude. Not ChatGPT.
It wasn't a small thing. It would have caused problems in production.
Now I've made sure the models don't make that same mistake again. I've incorporated his insight into how I prompt, what I look for, how I verify.
That's not me using tools.
That's hybrid intelligence. Different substrates contributing different strengths to the same problem.
I'm not supervising AI. I'm orchestrating multiple forms of intelligence, including my own, into something that works better than any of us alone.
The human caught what the models missed. The models catch things I miss constantly. And when we work together, maintaining context across biological and computational thinking, we get to answers none of us would reach independently.
This is what our current metrics don't capture.
The economic transformation isn't "AI replacing human tasks faster."
It's "new forms of work that didn't exist before, requiring coordination across different types of intelligence."
Of course the old metrics can't measure that.
What AI Actually Is
The real distinction isn't natural vs artificial.
It's biological substrate vs computational substrate.
Intelligence running on different hardware.
What we call "artificial intelligence" would be better described as computational intelligence — intelligence instantiated in a different substrate, with its own architecture, strengths, and failure modes.
We can't fully explain how neural networks produce understanding. But we also can't fully explain how biological neurons produce consciousness.
Both are processes we observe but don't completely understand.
Different substrates. Different architectures. Both real.
What Changes If We Drop "Artificial"
This isn't just semantic.
When we call it "artificial intelligence," we make architectural decisions based on a flawed model.
We try to make AI "explain its reasoning" the way humans do. But it doesn't have human-style reasoning. It has its own process.
We demand "common sense" the way humans have it. But common sense is pattern recognition from human sensory experience. Computational intelligence builds different patterns from different inputs.
We measure performance against human benchmarks. Then we're confused when it's superhuman at some things and struggles with tasks humans find trivial.
If we stop calling it "artificial" and treat it as a different substrate:
Instead of forcing explanation in human terms, we build transparency around its actual process.
Instead of demanding human-style common sense, we give it the context it actually needs.
Instead of comparing to human performance, we figure out what it's actually good at.
And critically: we need to understand both human-style social biases and substrate-specific failure patterns. Hallucination. Token preference. Training data artifacts. Context window limitations. These are computational failure modes that don't map to human cognitive biases. They require different detection and mitigation strategies.
Both matter. Human bias and computational failure modes. But they're different problems requiring different solutions.
That's practical architecture, not philosophy.
Why This Matters Now
We're building systems making real decisions. Medical diagnoses. Loan approvals. Hiring. Legal research.
If we keep treating computational intelligence as "artificial," as derivative, we'll keep building the wrong safeguards.
We'll apply human accountability frameworks when we need frameworks designed for computational decision-making.
We'll demand human-legible explanations when we need different transparency mechanisms.
We'll keep asking why AI doesn't transform the economy like humans would, instead of noticing it's already transforming how we work — through hybrid forms of intelligence we don't even have good language for yet.
Closing Thought
The octopus taught us intelligence doesn't have to be centralized.
My morning realization, that I'm getting relatively dumber every day, is teaching me something else:
Maybe the problem isn't that computational intelligence is "artificial."
Maybe the problem is that we assumed human intelligence would always be the reference point.
What we're building isn't artificial. It's just different.
And calling it by the right name is the first step in understanding what it actually is.