I’m writing this sitting in the SeaTac airport lounge waiting for a flight, looking at Bernie Sanders’ latest op-ed about AI, where he proposes that the solution to America’s AI woes is for the government to take a 50% ownership stake in the big tech companies — half the shares, half the board seats, packaged as a “sovereign wealth fund” funded by a one-time tax paid in stock.
OMG … Noooooo !!!
One of the many ironies that’s hit me lately, as AGI moves toward realization and the world at large starts to grasp that AGI really is coming soon, is this: the left wing of US politics — which in many ways I’m deeply sympathetic with, and agree with on many important matters — is more or less diametrically opposed to what I think is right on the two technologies most central to my own work, AI and blockchain.
On social policy, I lean strongly toward the government helping out the poor and the disadvantaged. My mother ran social welfare programs for most of her career, and I came away from watching her triumphs and struggles with a solid understanding that it’s usually not someone’s fault that they’re in a bad economic spot — and that we do a lot of good by helping both adults and kids improve their circumstances with education , with counseling and coaching, and also with practical material assistance. I’m also pacifist by default, if not by dogma; most of America’s military adventures, I think, have made things worse rather than better.
But on blockchain, it seems right now most of the left wants to ban it outright or regulate it into dysfunction, and I think that would be very badly wrong, both morally and pragmatically. Yes, crypto can be abused. It also has amazing potential for good — for coordinating people and AI agents to do cool and useful things in genuinely novel ways.
And on AI, it also seems most of the left wants to regulate it into submission. The implicit logic seems to be: if advanced AI is going to take people’s jobs, then kill the AI, because people need jobs. That’s a terrifyingly short-sighted way of looking at things.
Over-regulating AI is going to be far worse than letting it self-organize and evolve. There could, in principle, be helpful AI regulations — rules that actually guide AI development in a positive direction — but essentially none of the regulations being proposed are of that nature. Almost all of them would make things worse. We’re far better off with minimal AI regulation than with foolish, paranoid or dogmatic AI regulation that holds back progress, or that simply hands ownership to the big companies best at capturing the regulators.
Sanders’ proposal, I think, sits squarely in the large family of well-meaning but wrong-headed ideas about AI and the government’s relationship to it — ideas from left-leaning thinkers whose heart is in the right place but whose mind is not.
White all this seems extremely obvious to me, I can see it isn’t obvious to everyone because I am reading quite a few online comments in support of Sanders’ unintentionally absurdist proposal.
So let’s roll through a few of the details…
What Sanders gets right
Sanders’ core suggestion is to give agency, control, and a share of the AI wealth to the general public, on the grounds that AI models are built on everybody’s data. And that is a noble thing to want, for sure. Modern AI has been built on the accumulated knowledge, creativity, labor, culture, language, science, code, art, and conversation of our entire species — the hate speech and the trolling very much included. So if AI becomes a central productive force on the planet, which indeed is in the midst of happening, its benefits should not accrue only to a handful of billionaires, VCs, and big companies.
The moral intuition here is fairly sound. Sanders is, at last, seeing AI as a civilization-shaping force. AI will change everything.
Where Sanders goes insanely wrong
But the remedy Bernie proposes — the government owning half the shares and appointing half the boards — goes in entirely the wrong direction.
One thing I find interesting is the assumption baked into his proposal: that big companies are going to dominate the future of AI. That doesn’t have to be the case.
AGI is powerful, fast-moving, and not something we fully understand yet. It’s a fragile assumption that AGI will be dominated by a few big firms.
Sanders has understood that AI is going to be big; he hasn’t understood that the moat around the big companies is neither that wide nor that deep. We have open-source AGI projects. We have decentralized AGI projects. There’s an enormous and growing amount of AGI innovation happening all over the place — including my own efforts with SingularityNET, the ASI Alliance, OmegaClaw, and so forth. There is also plenty of serious AGI work outside the US entirely.
Trying to nationalize today’s AI leaders, among its many other problems, builds a long-term policy around a short-term reality — around the assumption that those leaders won’t simply get innovator’s-dilemma’d out of the lead they currently hold.
But let’s grant Bernies “Big Tech Uber Alles” assumption for the sake of argument. Suppose today’s big tech companies do remain utterly central to pushing AGI forward. What are the risks of handing the government a 50% stake and board-level control?
The obvious one: it would slow these companies down and make them dysfunctional and incompetent at the very things they do best.
It’s not that I think government is always bad at fostering innovation — far from it. Public funding built the internet, GPS, modern computing, and a lot of space exploration (much of it military funding, which is another story). US military funding literally invented AI. The NIH has driven amazing medical progress — not perfect, too conservative in many ways, but epically valuable nonetheless. Government is genuinely good at certain kinds of powerful and critical invention.
However, AGI, at this stage, is not likely to fall into this category. AGI research right now is exploratory, uncertain, weird, and extremely fast-moving. It depends on rapid iteration, fierce technical disagreement, speculative architectures, and open-ended experimentation. A company half-governed by political appointees is going to optimize for optics, compliance, and political acceptability. It’s going to move very, very slowly. It’s going to turn into a utility, like the power company.
And then what happens? The center of AI progress moves elsewhere — to open-source networks, smaller companies, foreign companies. At which point the government faces a choice: accept that innovation has moved beyond its control, or try to extend the same control regime to the new innovators — nationalizing every startup the moment it shows promise, or simply outlawing them. Either way, it would be a huge mess which would slow down innovation and lead to all manner of complex new practical dilemmas.
The importance of asking suitably deep questions…
What Sanders’ proposal overlooks is that ownership is not really the crux of the matter. Who holds the stock, who has the board votes — those matter, but they’re surface. The deeper relevant question is: what species of governance and guidance system can actually do what AGI needs? What kind of system can make wise, adaptive, transparent, participatory decisions at the speed and complexity AGI demands?
Our current political system is not the one, that’s for sure. Legislatures move slowly. Regulatory agencies are routinely captured by the industries they’re meant to regulate. Public debate on the big social networks is fragmented, emotionally charged, and easily manipulated. (And nationalizing the social networks wouldn’t solve any of that; it would just create something even more dysfunctional. You have to get far more creative.)
I’ve been thinking about this sort of issue a fair bit lately. My father, Ted Goertzel, is a political sociologist, and he and I have been working on a book called AGI Nation, about what public policy in the US and other modern democracies should look like as AGI unfolds. One of the central notions we’re exploring is what we call creative decentralized governance.
We don’t want to replace human democracy with AI, and nor do we want to let corporations rule the future. What we want is for AI systems to help people enact democracy more intelligently. The question isn’t which federal appointees get to sit on OpenAI’s board. The questions are things like:
How can citizens meaningfully participate in AI governance?
How can expert knowledge be integrated without sliding into technocracy?
How can local communities shape AI deployment in their own lives?
How can AI systems make their reasoning transparent enough to be challenged?
How can competing AI systems audit each other?
How do we avoid both corporate monopoly and government monopoly?
How can AGI help humanity deliberate better, rather than just shifting power from one institution to another?
These are the sorts of issues my dad and I are trying to address. They don’t seem to be things Bernie Sanders — wonderful, good-hearted human being though he is — is considering.
So what should we actually do?
It’s a long story (which is why we’re writing a book about it), but here are a few clues:
Support decentralized AI infrastructure. Don’t try to ban it. Credit where it’s due: the Trump administration’s openness to crypto is, at least, not standing in the way here….
If the government wants to invest, don’t just buy shares of big tech companies. Invest in open-source and open-science AI. Fund decentralized AI networks and open AGI architectures. Create public AI infrastructure that is genuinely open and decentralized — not monolithically owned and controlled.
Use the tools the government already has well. Antitrust law and interoperability policy are reasonably powerful instruments for preventing lock-in. The point isn’t for the government to own the monopolists; the point is to stop monopolies and oligopolies from forming in the first plac, and bust them up as needed. Foster decentralized tooling and open science, and use antitrust to keep the monopolies from dominating.
Build mechanisms for broad economic sharing that don’t freeze innovation. Maybe we do want a sovereign wealth fund that helps provide a basic income. But that can be done without the government owning shares of a few big companies — it could be done, say, through dividend flows from decentralized AI networks the government helped fund. Public stakes in publicly funded decentralized ventures. We don’t need to replace a corporate monopoly with a government-plus-corporate monopoly that’s just as oligopolistic and even less efficient.
Think globally. Focusing on the US makes sense for those of us here, but AI governance is global — it has to be species-wide. If we reform AI in a way that’s confined to the US, then whether we intend to or not, we’re feeding the AI-arms-race narrative, which is destructive and harmful at best and existentially catastrophic at worst. The right long-term frame isn’t US government ownership of US AI companies. It’s: how do we foster the emergence of open, cooperative, international, decentralized institutions capable of guiding AI for the benefit of all sentient beings?
Not rigidly-constrained nationalized quasi-AGI bureacracy — self-organizing global decentralized AGI for broad human benefit
Summing up…
Bernie is right that it’s wrong for a handful of billionaires running a handful of tech companies to decide the future of AI — and thus, potentially, the future of our species and its successors. He’s right that the public deserves a share in the wealth generated from humanity’s collective knowledge. He’s right that AI governance must be democratized.
But a mechanism as blunt as giving the government 50% control of the big tech companies would almost certainly cause more harm than good. Fortunately, it has essentially no chance of actually happening.
To say it differently:
We don’t want the government to nationalize big AI.
We want to make sure AGI never becomes “big AI” in the first place.
A fortunate fact is, we more or less do know how to achieve this. The AGI research community (including my team in the SingularityNET ecosystem) does actually know better ways of building AGI than what big tech is doing — including ways suited to running on open, decentralized networks — and we’re getting more and more people interested in building such things. We could use help, including from the government. But reifying today’s big-AI oligarchy into an oligarchic government bureaucracy is not actually going to be helpful.
It’s disappointing, as someone who’s been sympathetic to many aspects of the political left for a long time, to watch the left try to impose an archaic way of thinking onto something fundamentally new. The left often has its heart in the right place — but you have to have your mind in the right place too. My hope is that the book my dad and I are writing can add a little more depth of thinking across the whole political spectrum on these critical issues.
In the end, if we’re going to have a positive outcome, the best route is for humanity to build AGI together. This has to go beyond left versus right, and it has to go beyond US versus China versus Russia. It has to transcend ideological, national, and global divisions, and return to the source of what brings us together in the first place: our common human bodies, human minds, human species — our compassion, our ambition, our fears. We have to go back to that commonality, and from there work together to think up genuinely innovative, creative, powerful solutions for guiding AGI’s growth… solutions that match (and can adapt with) the power of the astonishing technologies we’re bringing into this world. We need an innovative energetic decentralized community to help turn rough proto-solutions like the ones I’ve sketched above — and my dad and I are fleshing out in more detail in our book — into impactful realities…
Sanders is often associated with a demonstrative political arrogance, grounded in a limited understanding of economic realities and an idealized view of socialism, in my view. I do not mean to ridicule any political view but this is too serious of an issue to not be critical of Sanders even if he seems to have a heart.
“return to the source of what brings us together in the first place: our common human bodies, human minds, human species — our compassion, our ambition, our fears. We have to go back to that commonality”
This is exactly right. You are building a strong foundation but the key ingredient in a powerful alternative is the constitution of this commonality: who is represented; who has agency; what are the terms of participation; and how do they exit if the structure does not serve them? Blockchain fits here as the key democratic tool.
You call out your mother’s social work and the Ubuntu tradition: the constitution of the helping relationship is prior to and more important than its content. Who is represented in the room. Who has agency over the terms. What happens when the relationship fails the person it was supposed to serve. Whether they can exit without losing everything they contributed to it.
Your mother didn’t just deliver services. She built and maintained covenants — relational structures in which the dignity of the person being helped was architecturally prior to the efficiency of the system helping them. When those structures failed — when the agency became extractive, when the caseworker’s incentives diverged from the client’s needs, when the person had no exit except deeper dependency — she saw what that cost. Not just in outcomes but in the specific moral injury of a system that consumed the suffering of the dispossessed without return.
That’s the missing ingredient in your decentralization vision. You’ve correctly identified that Bernie’s nationalization proposal answers the wrong question — ownership is surface, governance is structure. You’ve correctly identified that open-source and decentralized AGI networks are more promising than government-plus-corporate oligopoly. But decentralization without covenant is just fragmentation at smaller scale. A thousand small extractive models are not better than one large extractive model. What makes the difference is not the size of the node but the constitution of its terms: who contributed to the corpus, under what governance rights, with what enforced exit if those terms are violated.
This is where your blockchain substrate becomes architecturally essential rather than ideologically preferred. The covenant needs teeth. Specifically:
Corpus provenance on-chain: every contribution to a local training corpus — a clinical note, a crisis transcript, a care worker’s documented dignity refusal — carries its consent terms as an immutable hash. The model trained on that corpus carries a verifiable provenance chain that determines what uses are permitted under whose governance. Not as a legal agreement but as an architectural constraint.
Cryptographic exit as the covenant’s enforcement mechanism: if the governance structure violates the terms under which a community contributed its signal, they invoke exit. Their contributions are flagged in the provenance chain. The model’s outputs derived from those contributions are marked revoked. Compensation flows trigger automatically. You don’t sue — you exit, and the exit costs the violator their most valuable signal. This is what makes democratic accountability real rather than procedural.
Federation governance weighted by acuity and marginalization rather than token holdings: when local models federate into shared models, the governance voting structure weights contribution by the structural depth of what the corpus captures — not by data volume, which privileges the already-advantaged, but by diagnostic specificity, which privileges the populations whose experience reveals what the system most needs to know about itself. The lessons of the dispossessed are most epistemically dear to the general covenant.
I am a psychiatrist working on building this around Harborview Medical Center for the benefit of the dispossessed of King County. This is your front yard (beyond the moat) and a natural partnership to demonstrate the power of locally constituted training corpora and governance. We should talk.