Quick note before today's email: NEO Alpha is almost done and I thank all the applicants for your patience.
Beta will start before the end of this week. I will get in touch with you soon. 🙏
---
I've been quite inconsistent with my emails lately. But it's not because I wasn't writing anything. I've been writing more than ever before.
Manually. With AI.
There has been much to think about and I've been busy putting thoughts into words.
I've seen much lately, and many thoughts and feelings arose, and are swirling inside.
There's a part of me that, for a time, felt guilt for not being consistent with my writing.
"Wasn't I the Consistency Guy?"
Yes, yes I am.
But it's now I've realize exactly what I've been doing.
I'm building an ideology.
Even though many of the things that I've been working on lately doesn't contribute directly to my business, it feels deeply important in my gut.
Something bigger than my own business.
We need to have a cure to these problems plaguing humanity.
From the threat of AI misalignment, AI brain rot and psychosis...
...to political unrest culminating in the recent assassination of Charlie Kirk.
As well as what's been happening globally.
Conflicts have been happening - and escalating - all around us.
Russia and Ukraine.
Gaza.
Thailand and Cambodia.
India and Pakistan.
Riots in Indonesia, Nepal, and Philippines.
I've been trying to make sense of it all, wondering what does this mean to our future, near and far.
Where are things leading us? Are we heading to World War 3?
And what can I personally do to contribute, or at least, prepare.
I don't know what exactly the solution is, systemically, that can permanently solve these issues.
But from my perspective, all these problems are one problem only: The Human Problem.
We are both the problem and the solution.
Whatever the answer ultimately is, the elevation of human consciousness must be a part of it.
Everything is tied together.
Over the past few days I've been working on yet another important thought piece for the Augmented Intelligence Community.
It doesn't directly address the conflicts that I just mentioned, but it is an important part of the idea and message of Augmented Intelligence.
We, humans, have to get smarter.
In today's article, I make the case for why AI, AGI, or even ASI (artificial super intelligence), is not the answer humanity needs.
If you want to read it in Facebook, click here.
Otherwise, feel free to read it here:
Article: AI Will Never Be Safe, And The Reason Is Not What You Think.
AI will never be safe. At least, not in a way everyone accepts.
Not because we’ll fail technically, but because of human nature itself.
In this article, I will explain why I said this, and what we can do about it.
The First Observation
To understand what I mean by “AI will never be safe”, we need to first understand what “safety” even means in the first place.
Let’s be precise about the word we keep throwing around.
People call something “safe” when, relative to what they value, its behavior is predictable.
It means losses stay within acceptable bounds, and any leftover risk is placed where they consider fair.
Ideally with some control or recourse.
We don’t demand zero risk. We demand bounded, predictable, fairly placed risk to the things we care about.
If you value life, “safe” means threats to life are rare, limited, and not dumped on you without consent.
If you value family, “safe” means your family’s downside is contained and you have ways to intervene.
Move up and down Maslow's hierarchy of needs, the definition shifts with context, but the structure holds: predictability, bounds, and fairness.
If you can live with this definition, keep reading.
If you don’t, this is the fork in the road.
Push back here before we build on it.
The Second Observation
The problem with us humans is that we can't seem to agree fully on what matters.
We keep disagreeing on methods, politics, lifestyles, perspectives, and so on…
At times we can’t even agree on fundamental truths.
Majority of the world says: “Earth is a sphere.”
Some people continue to say: “That’s a conspiracy. Earth is flat.”
Worse, humans perceive existential threats to each other’s values even when no conflict actually exists.
The reason why this happens is because nobody can see “everything” all at once.
We are not omniscient, nor omnipresent.
We are not gods.
Take a flat-earther into orbit and spend a day there, or even easier, just up in the air and they will see the curve over the horizon.
Or take them past the North/South Pole, show them the compass, and they will see that too.
Until then, they will remain fervent in their conviction that the Earth is flat.
But in the first place, why does it even matter to their lives that the Earth is flat or not?
How does that change their lives, or what they need to do on a day-to-day basis?
If you think about it, they still need to get up, work, take care of their health, their relationships, their families, their money, and so on.
Does it even make them happier, truly, to know whether they are right or wrong?
It doesn’t really change people’s lives, yet so much effort, time and resources have been spent on this disagreement.
Are you starting to see the problem?
Coming back to AI and why AI will never be safe:
A farmer fears AI will automate away traditional farming. But an agtech CEO fears AI won't modernize agriculture fast enough to feed 10 billion people.
A Bangladeshi garment worker needs AI to stay out of textile manufacturing. But a fashion industry CEO needs AI to automate every garment factory to stay competitive.
A privacy advocate needs AI to delete all personal data. But a surveillance state director needs AI to retain all personal data.
A single mother in Detroit needs AI to help without moral judgment. But an Iranian cleric needs AI to assist her with moral judgment in every interaction.
A climate activist needs AI to halt economic growth. But a developing nation economist needs AI to maximize economic growth.
An Amazonian elder needs AI to stay out of indigenous lands. But a tech evangelist needs AI to reach every corner of Earth.
The reason why these people have diametrically opposed definitions of what “safety” means is because they all have different values, inspired by different upbringing, and where they are in life leading to different scale and scope of perspective.
They can’t see each other’s perspective because they are not living each other’s lives.
These aren't just different preferences, but completely different realities entirely.
Third Observation
This brings me to the third observation.
Reality doesn't care about your preferences, beliefs, values, or worldviews.
Politicians understand this better than anyone.
They promise lower taxes to win votes, better services to keep people happy, and reduced debt to sound responsible.
Then they get into office and face reality: You can't do all three simultaneously.
The math doesn't work, and someone always gets hurt.
Many people say that’s because politicians are evil.
But I think that reality has constraints that campaign promises don't acknowledge.
Because if they do, they’d lose.
Now with AI, the intention is now to create this entity that the developers imagine will be able to solve everything.
And by ‘everything’, it means ‘everything’ from the perspective of the developers.
This naturally makes it vulnerable to the problems of the Second Observation.
An AI making optimal decisions would face the same constraints, but without the need to win elections.
It would have to make trade-offs based on whatever it's optimizing for - and those trade-offs would likely violate billions of human preferences.
Not necessarily because it judges us or our values, but simply because the math of optimization under real-world constraints rarely produces outcomes that everyone likes.
When resources are finite, time is limited, physics has laws, and actions have consequences - something has to give.
The politician chooses based on political survival.
We don't know what an ASI would base its choice on.
We probably wouldn’t even understand the constraints it's operating under, or why certain trade-offs are necessary.
Fourth Observation
Cognitive ability itself seems to create different values about what matters in life.
Map "safety" and its various across cognitive abilities.
Someone with low IQ would want protection from complexity and uncertainty.
Someone with moderate IQ understands that complexity and uncertainty is necessary.
They don't love complexity for its own sake. But they recognize it as a tool to achieve goals - whether that's learning something new, getting a better job or solving meaningful problems.
What about someone with high IQ? I honestly don't know what they value.
Their values might be as foreign to moderate IQ individuals as moderate IQ values are to low IQ individuals.
But IQ is just one dimension. Consider emotional intelligence.
Someone with high EQ fears AI that makes us more isolated, even if it's more efficient.
They want AI that enhances human connection instead of replacing it.
Their "safe AI" would be something that preserves empathy, relationships, the messy beauty of human interaction.
But someone with low EQ might not even notice if AI eliminated most human contact.
They might actually prefer it.
Efficiency without the complications of emotions? Perfect.
Then there's practical intelligence - street smarts versus book smarts.
Someone with high practical intelligence only trusts what they can see and touch.
They want AI they can fix with a wrench, not some abstract system in the cloud.
Their "safe AI" has an off switch they control.
Someone who lives in abstractions however might trust complex AI systems more than physical reality.
They're comfortable with algorithms making decisions because they understand the logic, even if they can't touch it.
But even though these may not be mutually exclusive visions…
Even though an ASI (artificial super intelligence) may be able to come up with a solution that is good for everyone…
Because of the Second Observation, people won’t be able to see or understand it.
Fifth Observation
Now we reach the final and most uncomfortable observation.
The smarter something is, the less we can understand its reasoning.
Think about humans separated by significant differences in intelligence.
(By ‘intelligence’, I mean the full spectrum of intelligence that I pointed out in my Fourth Observation)
Someone with high IQ trying to explain their thinking to someone with moderate IQ faces a translation problem.
The concepts themselves resist translation, and simplifications are necessary, which robs the explanation many of its nuances.
It's almost like trying to explain color to someone born color blind.
Right now, the average human’s IQ ranges between 85–115.
Even with this narrow range of between 85 and 115 (30 points), we find ourselves unable to understand each other well - the point of reaching violence among the fringe.
Most times, those of us in the average IQ range can’t even understand the geniuses or their motivations.
Now imagine an AGI with capabilities far beyond human intelligence.
Or an ASI that surpasses us by orders of magnitude.
The difference of one hundred. One thousand. Ten thousand IQs?
We wouldn't even recognize what we're missing.
Whatever it proposes as the ‘way’ to solve the big problems: Curing aging, redistributing resources, merging consciousness, optimizing biology - humanity will likely be unable to comprehend it.
And because we are unable to comprehend it, we will either:
A) Surrender power to it fully and completely; or
B) Resist and rally/work against it.
‘B’ will lead to conflict with entities vastly superior to us in intelligence, where as ‘A’ will seed resentment among more radical-prone segments of humanity.
Which brings us to an uncomfortable realization.
The ASI will figure this out.
It will recognize that humanity will never voluntarily accept what it determines necessary.
Just like you understand your dog will never voluntarily take bad-tasting medicine no matter how much it needs it.
So what do you do with the dog?
You hide pills in food.
You hold its mouth.
You do what's necessary for its own good despite protests.
Except your dog won’t actually attack you to avoid eating those pills.
Humans will.
The ASI will reach similar conclusions about us.
Not from malice but from logic.
The same logic that makes parents vaccinate crying children.
The same logic that makes doctors operate on patients who can't consent.
The entity capable of solving humanity's greatest problems will necessarily violate in the process billions of preferences.
It will be labeled 'unsafe' because it's doing what needs to be done instead of what we want.
We are trying to build something we're destined to fear because we can't recognize salvation when it arrives in a form we never asked for.
So I return to my original statement:
AI will never be safe.
But I have a potential solution.
I will cover that in the next thought piece.
FELIX
P.S. I've you read everything above, thank you for your time. I feel that this message is an important one to hear.
If you haven't yet done so already, do join the Augmented Intelligence Facebook group and share your thoughts there.
Here's the link: https://www.facebook.com/groups/augmented/posts/1291317215383994/
P.S.S. I haven't been launching workshops because I have been occupied with 'clarity' ever since I committed to NEO.
It's a necessary part of transitioning to different direction. But I got something cooking up.
I will tell you all about it soon.