Why I strongly disagree with your arguments, Daniel

[acq23]

On the video Why We Need Some AI Model Controls.

It is also a bit scary to see how many people agree with you.

The problem is real. So is it for every dual-use tool.

I am not saying the risk isn't there. It is there in the same way it is there for any tool with two uses: knives, rifles, guns, social media.

Social media are weapons too. They can be aimed at one person, and children have died from that. They can be aimed at a population: Facebook ran an undisclosed behavioral experiment on roughly 700,000 users, which became public in 2014 through publication, not through a leak.

So: harm is demonstrable, and none of it required a global authority to identify. It required researchers publishing and national institutions acting.

"For your protection" is the oldest formula there is.

The narrative that a measure exists to protect people from something has been used to erode democracy and freedom throughout history, and recent events bear that out.
The reasons to reject this run from the practical to the strategic to the philosophical, and I will not list them all here. Pick whichever resonates with you.

The specific version you are proposing, that a knowledgeable group should hold controls for the good of everyone else, is paternalistic, and paternalism has a documented record. In 1927 the United States Supreme Court upheld compulsory sterilization in Buck v. Bell, on public-health reasoning, by eight votes to one. Tens of thousands of Americans were sterilized under state programs that followed. The decision has never been formally overturned. It was not the work of villains, but of the most credentialed and well-intentioned people of their era, acting on the best expert consensus available to them, for the good of a population that had no say.

That is the failure mode of expert control: confident consensus applied to people who cannot refuse it.

Ask where this reasoning has already arrived.

Things that have happened:

• Communications scanning. The EU permission for platforms to scan private messages lapsed on 4 April 2026, and in July 2026 Parliament extended it to 3 April 2028. More MEPs voted against that extension than for it: the motion to reject drew 314 votes against 276 in favour, with 17 abstentions, and failed only because blocking required an absolute majority of 361. The permanent mandatory-scanning regulation is not law yet and failed its June 2026 negotiation round. A measure most voting members opposed survived on a procedural threshold.
• Identity verification. Introduced for minors on social platforms, which is where internet identity verification for everyone begins.
• Behavioral scoring and control apps. Already deployed, already normalized during the last emergency.
• Medical mandate. Coordinated treatment policy in a pandemic, set above the individual's right to decide about their own health.
• Carbon taxes and credits. Allocation of a permitted quantity of activity per person.

Each one arrived with a protective justification, and each one was easier to introduce because the one before it had already been accepted.

Who controls the controllers?

The governance structure you are proposing has a track record, and it is poor.

The UN-inspired model does not answer to the people it governs, and that is not a theoretical complaint. UN peacekeepers introduced cholera to Haiti in 2010, in an epidemic that killed thousands. The organization denied responsibility for six years, and when victims sued, it asserted absolute immunity under its 1946 convention and the claim was dismissed. Nobody was accountable to anyone. In 1995, Srebrenica was overrun while under UN protection as a designated safe area.

Now judge them on the goals they set for themselves.

The accountability problem is one half. The other half is performance on the problems this model claimed, and the numbers come from the institution itself.

In June 2026 the UN reported that 655 million people still live without electricity, that on current policies 640 million will still lack it in 2030, and that roughly 85 percent of those people are in sub-Saharan Africa. Its own assessment is that the pace of electrification has to triple to reach a target it set for itself. Water, sanitation and schooling read the same way, decade after decade, in report after report published by the same offices.

These are self-assigned goals, with self-chosen deadlines, backed by decades of funding and the united agreement of the world's governments. On its own scorecard, the united-leaders model has not delivered basic electricity, water or schooling to the people it speaks for.

So when the same structure is proposed as the safeguard for AI, the question is not whether the intentions are good. It is why anyone expects a different result from the same design.

Look at what EU institutions do to EU citizens, by unelected commissioners, and ask the same question at global scale. Who controls these people? At present, by treaty, the answer is no one.

You cannot put an unaccountable body in charge of a technology this consequential and call that a safeguard. An arrangement that cannot be held to account is not governance. It is only the promise of governance.

What actually moves this

This class of problem is solved by the growth of consciousness in enough people. Every attempt to force it is, at best, instrumental to that growth in the long run, and capable of severe damage in the short and medium term.

Technology has evolved, for reasons I do not claim to know, far faster than humanity, or faster than whoever is steering humanity. That gap is the real problem, and no committee closes it.

Which leaves the part nobody can do on your behalf:

• Take responsibility for your own life instead of delegating it to someone or something. No top-down solution will work.
• Educate your children.
• Keep educating and improving yourself.
• Don't assume voting for X or Y solves your problems. Act in your day-to-day life.
• Don't be consumed by what you cannot affect.
• Help yourself before helping others.

Sincerely,

[chrisglick]

I read what your AI wrote based on some initial prompt which I would have preferred to read.

I agree that "who watches the watchman" is a big concern, one which we haven't and probably can't solve within current paradigms.

I want to reply with something, but I didn't actually see what you are proposing? It reads like a disagreement, but without actually stating what you stand for? Free open models for everyone?

[acq23]

Thank you for the effort.
I gave it the points it developed with examples: so you are reading my sauce 70% with AI historical examples and rearrangement.
So seems you, and someone else, do not understood nor the file rouge nor the conclusions probably because biased toward "AI-slop" mindset:

This class of problem is solved by the growth of consciousness in enough people. Every attempt to force it is, at best, instrumental to that growth in the long run, and capable of severe damage in the short and medium term.

Technology has evolved, for reasons I do not claim to know, far faster than humanity, or faster than whoever is steering humanity. That gap is the real problem, and no committee closes it.

Which leaves the part nobody can do on your behalf:

Take responsibility for your own life instead of delegating it to someone or something. No top-down solution will work.
Educate your children.
Keep educating and improving yourself.
Don't assume voting for X or Y solves your problems. Act in your day-to-day life.
Don't be consumed by what you cannot affect.
Help yourself before helping others.

Or plainly: you can't control that in any way. People must carry out control mastering what they are able to master.

Kind regards

[chrisglick]

I read like 100 pages of AI slop in my work and personal projects every day, so I have a pretty strong filter against it.

I've read your 6 points, but how do those apply to limiting access to AI models? This is the "technology outrunning wisdom" problem which I've also had Claude bring up in response to some of my questions.

I feel like you're mixing together a personal accountability goal with a shared accountability goal. Restricting AI models won't change YOUR ability to do any of the 6 things you listed? And having open models for everything won't really further the 6 goals you listed?

So: are you sharing this as a public service? Are you saying that our personal control is more important than gov't, business, family controls? OR that, for us to help ourselves and help others, that we cannot control models?

It's interesting you talk about growth of consciousness, but it sounds like the word "consciousness" isn't in the spiritual sense, but is in the known failings of gov'ts, regulators and private industry to regulate themselves?

I'm more responding out of curiosity than trying to disagree with you at this point.

IT seems like your focus is this list:
Take responsibility for your own life instead of delegating it to someone or something. No top-down solution will work.
Educate your children.
Keep educating and improving yourself.
Don't assume voting for X or Y solves your problems. Act in your day-to-day life.
Don't be consumed by what you cannot affect.
Help yourself before helping others.

Which I think is generally good advice!

[acq23]

They don't limit access to AI.

The growth of individual consciousness is the basis of a future better society: no matter the field.
Delegating our responsibilities as individuals to gvt and external entities that are restricting AI access using authoritarian methods like the ones described by Daniel or those already in place globally will not produce a better society.

A solution may be the manufacturers will come to an agreement to test each other's models before public release and let governments block the release worldwide if needed.
But this requires an agreement among manufacturers and USA - CHINA - RUSSIA, that, as of today, is more utopia than a practical solution.

Maybe all of this happens to force people to evolve and understand that they are in charge of their own lives. They have the duty and the power to control gvt and the entities that have direct impact on them: not the other way around.

[woutervanderwal-cloud]

You might find this initiative quite interesting: https://pauseai.info/ they have some really good reasoning on their website of what their goal is and how to achieve this.

[acq23]

Thank you for the link