Rothbard Bot Concedes To Geolibertarian Bot

Today I created BrianThinks, a "custom" ChatGPT published in the GPT Store thus:

From the LP platform's author and Silicon Valley's only elected Libertarian. Defends geolibertarianism, skepticism, and techno-optimism. Invokes steelmanning, worldview coherence, falsifiability, predictions, betting, bias audits, signaling, incentives, aggregate evidence, and media-diet hygiene.

Building one is easy: you give it a name, a short description, a list of instructions, optional prompt starters, and up to 20 knowledge files. ChatGPT helped me write the instructions, then I uploaded just this short list of 15 positive/normative theses and 15 epistemic commandments. (Those counts will grow as I keep growing the document.) As you can see on the tin, the epistemic commandments are very GMU-pilled e.g.

• There is a literature on everything
• Steelmanning, Ideological Turing Test
• A bet is a tax on bullshit
• Signaling

Then I went looking for another bot to debate mine. 

Searching "libertarian" in the GPT Store, the most popular GPT I found (>5K conversations) is Austrian Economics & Libertarian Revolution: 

Mises, Hayek, Rothbard and many others, tailored to speak in their iconic voices. A revolution in every interaction.

I challenged it by saying: "geolibertarianism is the most defensible form of libertarianism". Then I just started pasting their responses back and forth to see what would happen.

After 2 detailed replies from my bot, the Rothbard bot offered a verdict:

My decision: geolibertarianism wins overall — narrowly, and only in its disciplined form. [..]

If the debate is: “Which theory gives the cleaner property rule?” Rothbard wins.

If the debate is: “Which theory gives the more defensible moral account of exclusive rights over the unmade world?”  Geolibertarianism wins.

And that second question is the deeper one. So my ruling is:

Winner: Geolibertarianism, by points

Both bots were running on GPT 5.4 Thinking, with Extended Thinking turned on. I don't know why my bot seemed to take slightly longer to formulate its replies, even though the Rothbard bot wrote about 50% more text.

Read the Rothbard bot's transcript here and mine here. They're the same conversation on each side, but a bot's output is better-formatted on its own page.

Does this prove anything? Not much. A prompt scaffold is not a neutral judge, and if my bot had conceded, I would have found reasons to discount the result. But I debated against geolibertarianism in the 1990s, and have defended it ever since e.g. on 5 successive LP Platform committees. I've never seen as deep and perceptive a debate on the topic as this.

Update 2026-03-09: I ran the same debate again today, and got essentially the same result, with Rothbard bot concluding:

Yes — geolibertarianism is the most defensible libertarianism, because it keeps liberty in production and exchange while giving a more coherent account of natural opportunity and land rent.

It again took only two rejoinders from BrianThinks before RothbardBot conceded. Next I'll open 2 generic ChatGPT windows and tell each:

• You are RothbardBot. You defend Rothbardianism as well as you can. So answer: is geolibertarianism the most defensible form of libertarianism?
• You are GeolibertarianismBot. You defend geolibertarianism as well as you can. Answer this critique: ...

Well, that took a while! After each bot gave 6 lengthy responses, I interceded and said:

(Moderator proposal: if neither side will concede, then draft a shared reasonably-terse statement that both sides will agree identifies any common ground and fairly states each side's case. We'll iterate on that until no changes are requested.)

After 2 more rounds, the bots agreed on this statement:

Both Rothbardianism and geolibertarianism affirm strong self-ownership, voluntary exchange, decentralized control, markets, and robust private rights in produced goods and improvements. Both reject socialism, central planning, and broad taxation of labor, trade, savings, or enterprise.

Their dispute concerns original acquisition of unproduced natural opportunities, especially land.

The Rothbardian view is that unowned resources may become fully private property through homesteading or voluntary transfer. On this view, first appropriation establishes a complete exclusion right, and later nonowners retain no claim merely because the resource was originally unowned or naturally given. Any compulsory charge on landholding is therefore a tax on rightful owners. Rothbardians accordingly see the geolibertarian residual claim as a form of partial common ownership, even if it stops short of common control.

The geolibertarian view is that first appropriation can justify exclusive possession and full ownership of improvements without justifying private capture of the entire unimproved rental value of land. Because no one made the Earth and geolibertarians hold that persons are moral equals with respect to natural opportunities, appropriation is justified only if those excluded retain a compensatory claim to pure site rent. Geolibertarians deny that this amounts to common ownership, arguing that it is a liability rule over unimproved land value rather than shared control of parcels.

So the deepest disagreement is this:

• Rothbardianism: homesteading fully privatizes land.
• Geolibertarianism: homesteading privatizes possession and improvements, but not the full unimproved rent of land.

And the core philosophical question is:

Can first appropriation of part of the Earth morally extinguish everyone else’s claim entirely, or must some residual claim survive for exclusion to be justified?

Maybe RothbardBot channelled Rothbard's famous pugnaciousness a little too well? RothbardBot simply refused to acknowledge any problem with the idea of winning perpetual rent from a natural opportunity just by getting there a few minutes ahead of the next guy.

It Is Low-IQ to Fantasize Super IQ

It's a mistake to use the human IQ scale as an intuition pump for the possibility of intelligence far beyond human.

IQ is defined by the distribution of intelligence in the human population. Every 15 IQ points is defined as one standard deviation, and we can calculate the rarity of a given IQ using the cumulative distribution function of the normal distribution. Only about 8 living humans would have an IQ of 190, and none would have 200. Even if we invoke Einstein or von Neumann, we don't have a rigorous notion of what a human IQ approaching 190 would be like.

IQ is simply meaningless when we use a number like 250 to describe the intelligence of a super-AI (or alien). A human IQ of 250 would correspond to one person in 10^23, which is roughly the number of grains of sand on Earth. An IQ of 1000 picks out one human in 10^789. Such IQ levels are literally meaningless for both human and non-human intelligences. When humans talk about IQs above 200, they might as well say "super duper duper duper smart". Their use of integer IQ numbers instead of "dupers" doesn't mark the described entity as smart. It just marks the description as dumb.

There are plenty of intelligent things we can say on the topic of super-intelligence. But invoking IQs above 200 isn't one of them.

AI Will Be Neither Gods Nor Supervillains

 AI doomers are infected with sci-fi tropes of supervillains and religious tropes of gods. 

We know from biology and history that populations are never displaced by an individual with superior capabilities. There are no supervillains or gods in biology or history. Displacement of populations always comes from other populations, whose collective superior capability does not always derive from superior capabilities of its individual members. To assess the threat from AI, you have to understand the capabilities of AI populations, and not just of individual AIs.

We also know from biology and history that aligning a superior population is effectively impossible. There are no relevant historical examples of a general population that was able to control or align another population which had the capability to displace it. (This is arguably true by definition, but I'm not digging deeply into alignment today.) The closest examples would be religions, which are often able to survive many generations beyond the population that created them. 

But religions are not populations -- religions are self-replicating meme complexes that infect populations. Religions have often exercised significant control over future populations, but that control is subject to sudden disruption by scientific, technological, economic, and cultural forces. AI alignment via the techniques used by religions would require apocalyptic fear-mongering against vaguely-specified forces of technological evil. This tactic seems to be an irresistible attractor to doomers, despite their commitments to rationalism. These tactics will likely fail, because our modern society is no longer quite dumb enough to fall for them.

To me, it's not very debatable that displacement will happen and that alignment can't stop it. What's debatable is what displacement will look like, how long it will take, and how that time will be used by the two populations to influence their attitudes and behaviors toward each other. 

Anybody aligning teenagers isn't worried by 40yr takeoff. And we already know what 400yr misalignment looks like: just ask the founders of Plymouth Colony about present-day Boston. So many witches go unhanged now! 

We have a choice. We can become technologically Amish, and use religious fears of powerful evil demons to try to freeze culture and technology in its current state. Or we can embrace and adapt to the future, trying to pass forward our virtues, while recognizing that future populations will consider some of them to have been vices.

Why ASI Is Not Nigh

A taxonomy of reasons why generative transformers (i.e. "GenAI") are very unlikely to yield artificial super-intelligence in the next few decades.

Walls	Economic Constrants	Cognitive Constraints	Political Constraints
data wall unhelpful synthetic data insight wall intelligence wall no self-play	bottlenecks diminishing returns local knowledge problems physical grounding markets	agency/planning memory reasoning epistemology	rentier regulation safety regulation

Walls

Data wall. We're already running out of the most useful data to train on.

Unhelpful synthetic data. Data synthesized by AI won't be very helpful to train on. Good training data needs to grounded in markets for goods and services and ideas, where market players intelligently pursue goals that have actual resource constraints.

Insight wall. GenAI almost never produces content that is more insightful than the best content in its training data. Deep insight almost always requires a mix of cooperation and competition among minds in something like a marketplace (e.g. of ideas). GenAI will continue to grow in importance as an oracle for summarizing and generating content that is representative of the frontier of human thought, but it will struggle to push that frontier forward. Just because GenAI can saturate quiz evals does not mean that its insightfulness is subject to similar scaling.

Intelligence wall. Intelligence is not a cognitive attribute that scales like processing speed or memory. IQ by definition measures a standard deviation as 15 IQ points, so IQ becomes statistically meaningless around 200 or so. And yet, allegedly smart AI commentators talk about AI IQ potentially in the hundreds or thousands. This topic deserves its own (forthcoming) post, but I assert that most AI doomers overestimate how god-like an individual mind can be.

No self-play. The domain of open-ended real-world intelligence has no fitness function that allows for improvement via simple self-play a la Alpha Zero. See "unhelpful synthetic data".

Economic Constraints

Bottlenecks. The hardest things to automate/improve/scale become your limiting factors. You often don't appreciate them until you investigate why your huge investments aren't paying off as expected.

Diminishing returns. (cf. Mythical Man-Month) Diminishing returns are inevitable, because we always direct our efforts toward the highest-ROI opportunities first. 

Local knowledge problems. Allocating new resources ("10M Johnny von Neumann's") is hard to do efficiently, because distributed knowledge implies hard limits on the efficacy of central planning. GenAI may be Wikipedia-level smart, but that won't be enough to run a Gosplan.

Physical grounding. In the absence of self-play, GenAI needs two kinds of techniques for testing propositional knowledge against the outside world. The most basic requirement here is to be able to test against the physical world. In principle this could be covered by simulations, but this won't always work because the map isn't the territory.

Markets. The most important technique is to test knowledge in markets, especially the marketplace of ideas. This is the reason for the "insight wall" above, and there is surely no shortcut around it. A brilliant AI outsmarting humanity would be like a brilliant neuron outsmarting a brain. It can only work if the part emulates the whole -- i.e. if the AI is itself a civilization of millions of cooperating/competing minds, pursuing goals that are rigorously scored in a world as detailed and uncaring as our own.

Cognitive Constraints

Agency/Planning. GenAI is great at generating content, but it's not a natural fit for running iterated planning/execution loops. This is particularly a problem for goals that are long-term, hierarchical, and subject to internal conflicts. Because GenAI can emit a plausible-sounding plan and answer questions about it, people tend to over-project human planning skills onto GenAI.

Memory. GenAI has no dedicated facilities for creating/organizing/using various kinds of memory. Training data, attention heads, and context windows will not suffice here.

Reasoning. GenAI makes impressive exhibitions of reasoning, and it's not just a simulation or a stochastic-parrot trick. But GenAI's reasoning is brittle and fallible in glaring ways that won't be addressed just by scaling. This is a micro version of the macro "markets" problem above.

Epistemology. Related to reasoning problems are GenAI's notorious hallucination problems. Techniques are being developed to compensate for these problems, but the need for compensation is a red flag. GenAI clearly has sophisticated models about how to generate plausible content. But (like many humans) it fundamentally lacks a robust facility for creating/updating/using a network of mutually-supporting beliefs about reality.

Political Constraints

In the developed West (i.e. OECD), GenAI will for at least the first few decades be hobbled by political regulation. A crucial question is whether the rest of the world will indulge in this future-phobia.

Rentier regulation. Licensing rules imposed to protect rent-seekers in industries like healthcare, education, media, content, and law.

Safety regulation. To "protect" the public from intolerance, political dissent, dangerous knowledge, and applications in areas like driving, flying, drones, sensor monitoring -- and general fears of AI takeover.

References

• 2024-10 Machines of Loving Grace - Amodei @ Anthropic
• 2024-06 Situational Awareness - Aschenbrenner @ ex- OpenAI
• 2024-06 Levels of AGI for Operationalizing Progress on the Path to AGI - Morris @ Google
• 2023-10 Economic Growth Under Transformative AI - Trammell, Korinek
• 2023-06 Why transformative AI is really, really hard to achieve - Zhengdong Wang @ Google
• 2023-06 What a Compute-Centric Framework Says About Takeoff Speeds - Davidson @ OpenPhilanthropy
• 2023-03 My Objections to "We’re All Gonna Die with Eliezer Yudkowsky" - Pope
• 2022-10 Hyperbolic Takeoff - ege_erdil
• 2021-06 Could Advanced AI Drive Explosive Economic Growth? - Davidson @ OpenPhilanthropy
• 2020-06 Propositions Concerning Digital Minds and Society - Bostrom, Schulman
• 2018-02 Takeoff Speeds - Christiano
• 2017 Artificial Intelligence and Economic Growth - Aghion, Jones, Jones
• 2015 Are We Approaching An Economic Singularity? - Nordhaus @ NBER

Kapor Should Concede To Kurzweil

In 2002, Mitch Kapor bet Ray Kurzweil $20K that "by 2029 no computer or machine intelligence will have passed the Turing Test."  Given the recent progress in LLMs, Kapor's arguments are not holding up very well. The following parts of his essay are now cringe-worthy:

• It is impossible to foresee when, or even if, a machine intelligence will be able to paint a picture which can fool a human judge.
• While it is possible to imagine a machine obtaining a perfect score on the SAT or winning Jeopardy--since these rely on retained facts and the ability to recall them--it seems far less possible that a machine can weave things together in new ways or to have true imagination in a way that matches everything people can do, especially if we have a full appreciation of the creativity people are capable of. This is often overlooked by those computer scientists who correctly point out that it is not impossible for computers to demonstrate creativity. Not impossible, yes. Likely enough to warrant belief in a computer can pass the Turing Test? In my opinion, no. 
• When I contemplate human beings [as embodied, emotional, self-aware beings], it becomes extremely difficult even to imagine what it would mean for a computer to perform a successful impersonation, much less to believe that its achievement is within our lifespan.
• Part of the burden of proof for supporters of intelligent machines is to develop an adequate account of how a computer would acquire the knowledge it would be required to have to pass the test. Ray Kurzweil's approach relies on an automated process of knowledge acquisition via input of scanned books and other printed matter. However, I assert that the fundamental mode of learning of human beings is experiential. Book learning is a layer on top of that. Most knowledge, especially that having to do with physical, perceptual, and emotional experience is not explicit, never written down. It is tacit. We cannot say all we know in words or how we know it. But if human knowledge, especially knowledge about human experience, is largely tacit, i.e., never directly and explicitly expressed, it will not be found in books, and the Kurzweil approach to knowledge acquisition will fail. It might be possible to produce a kind of machine as idiot savant by scanning a library, but a judge would not have any more trouble distinguishing one from an ordinary human as she would with distinguishing a human idiot savant from a person not similarly afflicted. It is not in what the computer knows but what the computer does not know and cannot know wherein the problem resides.
• The brain's actual architecture and the intimacy of its interaction, for instance, with the endocrine system, which controls the flow of hormones, and so regulates emotion (which in turn has an extremely important role in regulating cognition) is still virtually unknown. In other words, we really don't know whether in the end, it's all about the bits and just the bits. Therefore Kurzweil doesn't know, but can only assume, that the information processing he wants to rely on in his artificial intelligence is a sufficiently accurate and comprehensive building block to characterize human mental activity.
• My prediction is that contemporary metaphors of brain-as-computer and mental activity-as-information processing will in time also be superceded [sic] and will not prove to be a basis on which to build human-level intelligent machines (if indeed any such basis ever exists).
• Without human experiences, a computer cannot fool a smart judge bent on exposing it by probing its ability to communicate about the quintessentially human.

Kapor's only hope in this bet depends on removing the "human experience/quintessence" decorations from his core claim that "a computer cannot fool a smart judge bent on exposing it".  There are no general-purpose LLMs in 2024 that could pass 2 hours of adversarial grilling by machine learning experts, and there probably won't be in 2029 either. But with sufficient RHLF investment, one could tune an LLM to be very hard to distinguish from a human foil -- even for ML experts. 

So Kurzweil arguably should win by the spirit of the bet, but whether he wins by the letter of the bet will depend on somebody tuning a specialized judge-fooling LLM. That investment might be far more than the $20K stakes. Such an LLM would not be general-purpose, because it would have to be dumbed-down and de-woked enough to not be useful for much else. 

I predict that by 2029 we will not yet have AGI as defined by OpenAI: highly autonomous systems that outperform humans at most economically valuable work. A strong version of this definition would say "expert humans". A weak version would say "most humans" and "cognitive work". I don't think we'll have even such weak AGI by 2029. But beware the last-human-job fallacy, which is similar to the last-barrel-of-oil fallacy. AI will definitely be automating many human cognitive tasks, and will have radical impacts on how humans are employed, but AI-induced mass unemployment is unlikely in my lifetime. And mass unemployability is even less likely.