Context for responding to this post. Also watched Alvin Graylin: The Realities of the U.S. China AI Arms Race, Stanford Digital Economy Lab Video.
Graylin refers to the Kuznets Curve, speculating that with an increase in intelligence, AI may become better at distinguishing between right and wrong. Later, he mentions one of several proposed risk mitigations: separating civilian and national security AI systems.
1. The Bill of Rights Parallel?
In my view, the closest opportunity we have had to understand how to construct an AI Bill of Rights would have been to actually practice the original U.S. Bill of Rights in daily life. That would mean a system where the people truly ran the show, with a functioning, respected feedback loop. Unfortunately, fewer citizens care about their rights today, and even fewer can define them until those rights are already lost. For those paying attention, we can see the degradation of the Bill of Rights through its corporatization into “terms of service.”
I agree with David that implementation is difficult. I would add that, under the principles of realpolitik and “he who has the gold,” it is not going to happen under the existing control structures and structures we neither see nor influence. Imho- with most systems, there will be two versions of AI governance: theirs and ours, or perhaps one you never “voted for.”
2. Why a Separation in AI Governance?
Graylin’s presentation included a slide entitled Actions for AI Risk Mitigation which suggested separating civil and national security AI. But why should such a separation be necessary at all?
If the ultimate goal of AI is to create a system that is broadly or tries to approach benevolence in its allocation of resources, then most, if not all, human needs would be met. The struggle for resources would fall to near zero, removing the primary driver of conflict, including wars fought over scarcity.
If advanced AI (ASI) were designed to optimize for the benefit of all sentient beings, then an “AI Bill of Rights” might not even be needed. By the logic of a Kuznets Curve, outcomes should naturally converge toward greater equity in resource allocation and reduce the influence of politicians or a technocratic class. In theory, such a system could self-direct toward peace and prosperity without the need for a separate governance layer.
But AI is programmed by humans and more importantly, by those who fund it. History suggests the outcome will be less about universal benevolence and more about preserving existing power structures or defeating competing ones. In other words, the status quo and no Ai Bill of Rights.
I'm reminded of the old "radical transparency" idea of mine. Basically, in order to prevent the formation of a surveillance state by ruling powers that keep their privacy while everyone else has none, privacy itself, for individuals and groups, would need to be abolished, that information is freely acquired by everyone. I'm also tepid at the idea since it feels very extreme to me. Though it's also a bottom-up approach to a sousveillance society if I recall the word right. Everyone has phones with cameras and recorders, so people use those more often and share it more often, especially at government and corporations.
It's still something that spooks me so I haven't really develop the idea that much.
I also asked Deepseek by translating my question into Chinese (to get it to give me a more Chinese PoV answer), and then translated it back into English.
Deepseek continues to have a turd of a UI that's not aimed at adoption... So I just pastebin'd the translated response for sharing:
My main point here would be that whatever one thinks of their moral character, the Chinese government is "a competent government".
And "being competent at governance", itself has pre-reqs.
Like, top-down commands without information from people who understand the issues well because they live with them, will typically just lead to policy failures.
You also need lots of clarity to enable coordination across diverse groups.
So, not being transparent is just unprofessional, and a form of self-harm.
—
Plus, the makeup of the Chinese government is much more STEM / engineering heavy, so their basic attitude towards governance is different.
Like them introducing smart traffic lights, and other tools to enable better performance by self driving cars and buses... Thus there being lots of actual self driving buses in China.
Ultimately, the Chinese are a non-western, "great civilization"... Their philosophies also enable advanced human organization — and if they have flaws, there is no human group without flaws.
Couldn't we just strengthen the existing US Constitution and federal statutory landscape to achieve the same thing? For example, the 1st Amendment should be revised or extended to protect digital speech and expression from censorship, thus protecting cognitive sovereignty. The 4th Amendment should extend over digital data and biometrics as "effects," prohibiting warrantless searches and data seizure ("digital and biological sovereignty"). The 5th and 14th Amendments' due process clauses protect against unreviewable AI-driven seizure of assets by allowing private causes of action against algorithmic discrimination and associated government tech platforms ("right to universal access"). Keeping rights grounded in the Constitution ensures adjudication of these rights in American courts, not whatever unnamed UN global council Graylin mentioned in his article.
@David, Thanks for taking interest in the draft AI Bill of Rights I posted. Given the limited length, many of the risks and abuses that this bill is intended to protected us from are not expanded upon. Happy to do a call with you sometime so you can better understand the underlying reasoning behind each bill.
If you have time, you may enjoy reading my book OurNextReality.com which gives more details on the benefits and risks of AI and their impact on society.
There’s also an earlier post on the Abundanist Substack on “Abundanism” which you may find interesting.
Thanks, Alvin. My critique isn’t about length; it’s about architecture and viability. The social risks you’re pointing to are already widely recognized in the AI safety and governance communities; the work is to translate them into legally justiciable and politically workable obligations. If you want to advance this, publish an addendum that defines terms without relying on ‘AGI,’ names duty-bearers and enforcement for each article, and outlines a credible adoption path and disclosures.
To be candid, in its current form I do not see a viable path to adoption for you Bill of Rights. I’ve offered my feedback; if you choose to address it in a public addendum, I’m open to reading it. Otherwise, I’ll let my analysis stand and wish you the best.
I’m afraid the AI safety community is overly focused on super AI take over and not enough on human misuse/abuse. If you’d like to chat more happy to discuss. Or we can discuss it on your show. I’ve been researching/delivering AI/XR and cybersecurity related projects for over thirty years, and now teaching and doing research at MIT and Stanford.
Thanks, Alvin. I appreciate the invitation and the effort behind your draft. I receive a steady stream of “frameworks” and have to triage carefully; I engage when the author shows receptivity to critique and when the text translates big risks into enforceable obligations that are legally justiciable and politically workable.
Your reply hasn’t engaged those specifics, and the document remains an aspirational manifesto rather than a policy-grade proposal. It is better than many of the submissions that come across my desk, but it still wouldn’t meet the editorial bar I use for serious governance work. Given that, a call or a spot on my show isn’t a good use of time.
If you choose to publish a public addendum that defines terms without relying on “AGI,” names duty-bearers and enforcement for each article, and lays out a credible adoption path and disclosures, I’m open to reading it. Otherwise, I’ll let my analysis stand and wish you the best with your work.
Taking cues from Larry Harvey... (Okay, Burning Man is for a festival/community but hey....not a bad way to structure your behavior either). https://burningman.org/about/10-principles/
Context for responding to this post. Also watched Alvin Graylin: The Realities of the U.S. China AI Arms Race, Stanford Digital Economy Lab Video.
Graylin refers to the Kuznets Curve, speculating that with an increase in intelligence, AI may become better at distinguishing between right and wrong. Later, he mentions one of several proposed risk mitigations: separating civilian and national security AI systems.
1. The Bill of Rights Parallel?
In my view, the closest opportunity we have had to understand how to construct an AI Bill of Rights would have been to actually practice the original U.S. Bill of Rights in daily life. That would mean a system where the people truly ran the show, with a functioning, respected feedback loop. Unfortunately, fewer citizens care about their rights today, and even fewer can define them until those rights are already lost. For those paying attention, we can see the degradation of the Bill of Rights through its corporatization into “terms of service.”
I agree with David that implementation is difficult. I would add that, under the principles of realpolitik and “he who has the gold,” it is not going to happen under the existing control structures and structures we neither see nor influence. Imho- with most systems, there will be two versions of AI governance: theirs and ours, or perhaps one you never “voted for.”
2. Why a Separation in AI Governance?
Graylin’s presentation included a slide entitled Actions for AI Risk Mitigation which suggested separating civil and national security AI. But why should such a separation be necessary at all?
If the ultimate goal of AI is to create a system that is broadly or tries to approach benevolence in its allocation of resources, then most, if not all, human needs would be met. The struggle for resources would fall to near zero, removing the primary driver of conflict, including wars fought over scarcity.
If advanced AI (ASI) were designed to optimize for the benefit of all sentient beings, then an “AI Bill of Rights” might not even be needed. By the logic of a Kuznets Curve, outcomes should naturally converge toward greater equity in resource allocation and reduce the influence of politicians or a technocratic class. In theory, such a system could self-direct toward peace and prosperity without the need for a separate governance layer.
But AI is programmed by humans and more importantly, by those who fund it. History suggests the outcome will be less about universal benevolence and more about preserving existing power structures or defeating competing ones. In other words, the status quo and no Ai Bill of Rights.
I'm reminded of the old "radical transparency" idea of mine. Basically, in order to prevent the formation of a surveillance state by ruling powers that keep their privacy while everyone else has none, privacy itself, for individuals and groups, would need to be abolished, that information is freely acquired by everyone. I'm also tepid at the idea since it feels very extreme to me. Though it's also a bottom-up approach to a sousveillance society if I recall the word right. Everyone has phones with cameras and recorders, so people use those more often and share it more often, especially at government and corporations.
It's still something that spooks me so I haven't really develop the idea that much.
It should be noted that the west's presentation of China isn't how the Chinese themselves view the government and their lived experience in China.
Here's a ChatGPT link someone shared for the Chinese idea of "Whole Process Democracy":
https://chatgpt.com/share/68ae0514-eb24-8012-841c-634146eaa878
I also asked Deepseek by translating my question into Chinese (to get it to give me a more Chinese PoV answer), and then translated it back into English.
Deepseek continues to have a turd of a UI that's not aimed at adoption... So I just pastebin'd the translated response for sharing:
https://pastebin.com/TmZ9jRcS
—
My main point here would be that whatever one thinks of their moral character, the Chinese government is "a competent government".
And "being competent at governance", itself has pre-reqs.
Like, top-down commands without information from people who understand the issues well because they live with them, will typically just lead to policy failures.
You also need lots of clarity to enable coordination across diverse groups.
So, not being transparent is just unprofessional, and a form of self-harm.
—
Plus, the makeup of the Chinese government is much more STEM / engineering heavy, so their basic attitude towards governance is different.
Like them introducing smart traffic lights, and other tools to enable better performance by self driving cars and buses... Thus there being lots of actual self driving buses in China.
Ultimately, the Chinese are a non-western, "great civilization"... Their philosophies also enable advanced human organization — and if they have flaws, there is no human group without flaws.
Couldn't we just strengthen the existing US Constitution and federal statutory landscape to achieve the same thing? For example, the 1st Amendment should be revised or extended to protect digital speech and expression from censorship, thus protecting cognitive sovereignty. The 4th Amendment should extend over digital data and biometrics as "effects," prohibiting warrantless searches and data seizure ("digital and biological sovereignty"). The 5th and 14th Amendments' due process clauses protect against unreviewable AI-driven seizure of assets by allowing private causes of action against algorithmic discrimination and associated government tech platforms ("right to universal access"). Keeping rights grounded in the Constitution ensures adjudication of these rights in American courts, not whatever unnamed UN global council Graylin mentioned in his article.
@David, Thanks for taking interest in the draft AI Bill of Rights I posted. Given the limited length, many of the risks and abuses that this bill is intended to protected us from are not expanded upon. Happy to do a call with you sometime so you can better understand the underlying reasoning behind each bill.
If you have time, you may enjoy reading my book OurNextReality.com which gives more details on the benefits and risks of AI and their impact on society.
There’s also an earlier post on the Abundanist Substack on “Abundanism” which you may find interesting.
Thanks, Alvin. My critique isn’t about length; it’s about architecture and viability. The social risks you’re pointing to are already widely recognized in the AI safety and governance communities; the work is to translate them into legally justiciable and politically workable obligations. If you want to advance this, publish an addendum that defines terms without relying on ‘AGI,’ names duty-bearers and enforcement for each article, and outlines a credible adoption path and disclosures.
To be candid, in its current form I do not see a viable path to adoption for you Bill of Rights. I’ve offered my feedback; if you choose to address it in a public addendum, I’m open to reading it. Otherwise, I’ll let my analysis stand and wish you the best.
I’m afraid the AI safety community is overly focused on super AI take over and not enough on human misuse/abuse. If you’d like to chat more happy to discuss. Or we can discuss it on your show. I’ve been researching/delivering AI/XR and cybersecurity related projects for over thirty years, and now teaching and doing research at MIT and Stanford.
Thanks, Alvin. I appreciate the invitation and the effort behind your draft. I receive a steady stream of “frameworks” and have to triage carefully; I engage when the author shows receptivity to critique and when the text translates big risks into enforceable obligations that are legally justiciable and politically workable.
Your reply hasn’t engaged those specifics, and the document remains an aspirational manifesto rather than a policy-grade proposal. It is better than many of the submissions that come across my desk, but it still wouldn’t meet the editorial bar I use for serious governance work. Given that, a call or a spot on my show isn’t a good use of time.
If you choose to publish a public addendum that defines terms without relying on “AGI,” names duty-bearers and enforcement for each article, and lays out a credible adoption path and disclosures, I’m open to reading it. Otherwise, I’ll let my analysis stand and wish you the best with your work.
I do, however, offer paid consultations if you want some serious feedback. https://linktr.ee/daveshap
Taking cues from Larry Harvey... (Okay, Burning Man is for a festival/community but hey....not a bad way to structure your behavior either). https://burningman.org/about/10-principles/