Return-Path: Delivered-To: alvarochf@traceseis.com Received: from cs510.bluehost.com by cs510.bluehost.com with LMTP id OCPBFf0rtGlbjgUAWiBQxw (envelope-from ) for ; Fri, 13 Mar 2026 09:23:41 -0600 Return-path: Envelope-to: alvarochf@traceseis.com Delivery-date: Fri, 13 Mar 2026 09:23:41 -0600 Received: from v540.v5375b7fa.use4.send.mailgun.net ([159.112.244.40]:40710) by cs510.bluehost.com with esmtps (TLS1.2) tls TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256 (Exim 4.98.2) (envelope-from ) id 1w14MX-00000001lZv-0Oev for alvarochf@traceseis.com; Fri, 13 Mar 2026 09:23:41 -0600 DKIM-Signature: a=rsa-sha256; v=1; c=relaxed/relaxed; d=mg-d1.substack.com; q=dns/txt; s=k1; t=1773415404; x=1773422604; h=List-Unsubscribe-Post: List-Unsubscribe: List-Post: List-Id: List-Archive: List-Owner: Reply-To: In-Reply-To: References: Sender: Sender: Date: Message-Id: To: To: From: From: Subject: Subject: Content-Type: Mime-Version; bh=qFt7a3wSB3AfH7sAOVAAnKxTQj60ui9vVmWvC1qdSkM=; b=J9rhmmJ+3UFKj8p99SZ/ak8Ue67A0D52TeWH28PnobQnlM2gTXJKTt9FGpky5x1iBZBg0yd8pbkYRYx+MHDE7DHBleNIeFbruGukeT2PE2E4klzcGNFepccbFItqGua7D1Axsd2auDk9RmI7umsc5KEQZ38ITzGQ+GCuK97rdEE= X-Mailgun-Sid: WyJiYTg1MiIsImFsdmFyb2NoZkB0cmFjZXNlaXMuY29tIiwiMzliMDhjNiJd Received: by da4ca03ce82b737b214fe7c354262fe87e6fd3c4aceb4246efe1752d0d6d0e8c with HTTP id 69b42bec8cb7f18f58aad26e; Fri, 13 Mar 2026 15:23:23 GMT X-Mailgun-Sending-Ip: 159.112.244.40 X-Mailgun-Batch-Id: 69b42becf7f3a8e2a2fa81ba Mime-Version: 1.0 Content-Type: multipart/alternative; boundary="b54437abcfea1cb634abd247e21b8de8de9493d14f9c4c4a14e0056f2ef0" Subject: =?UTF-8?q?AI_won=E2=80=99t_fix_central_planning?= From: Cosmos Institute To: alvarochf@traceseis.com X-Mailgun-Tag: post X-Mailgun-Track-Clicks: false Message-Id: <20260313152308.3.45e984aecf7d72e9@mg-d1.substack.com> Date: Fri, 13 Mar 2026 15:23:08 +0000 Feedback-Id: post-190832174:cat-post:pub-2225794:substack Sender: Cosmos Institute References: In-Reply-To: Reply-To: Cosmos Institute List-Owner: List-Url: List-Archive: List-Id: List-Post: List-Unsubscribe: List-Unsubscribe-Post: List-Unsubscribe=One-Click X-Mailgun-Variables: {"category": "post", "email_generated_at": "1773415403694", "is_freemail": "true", "post_audience": "everyone", "post_id": "190832174", "post_type": "newsletter", "pub_community_enabled": "true", "publication_id": "2225794", "subdomain": "cosmosinstitute", "user_id": "347258985"} --b54437abcfea1cb634abd247e21b8de8de9493d14f9c4c4a14e0056f2ef0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable View this post on the web at https://blog.cosmos-institute.org/p/ai-wont-fi= x-central-planning In 1962, Victor Glushkov pitched the Soviet authorities on a nationwide cyb= ernetics network to solve the oldest problem in socialist economics: how to= allocate resources without private property and market prices. Washington = was alarmed enough for the CIA to create a special taskforce. Just as Soviet industrialization was giving way to stagnation, the first se= rious computers gave the planned economy a shot in the arm. The central pla= nners believed that they=E2=80=99d struggled to process information fast en= ough, so increased computing power promised a solution. In 1969, the Polish economist Oskar Lange described the market as =E2=80=9C= a computing device of the pre-electronic age,=E2=80=9D suggesting we could = put =E2=80=9Csimultaneous equations on an electronic computer and obtain th= e solution in less than half a second.=E2=80=9D In 1971, Chile=E2=80=99s so= cialist government attempted a version of this managed via telex machine, o= nly for it to be cut short by the coup of 1973. Thirty years after the fall of the Soviet Union, enthusiasm for these ideas= has returned. In 2016, Jack Ma predicted that [ https://substack.com/redir= ect/90aa387c-1022-4c6b-8476-b739add61f48?j=3DeyJ1IjoiNXFxeXF4In0.h9dEDbGj64= cTNudluPwbMYWklSu8UKGoFv6eWS-HyoA ] =E2=80=9Cthe planned economy will becom= e increasingly big =E2=80=A6 because with access to all kinds of data, we m= ay be able to find the invisible hand of the market.=E2=80=9D Marxist econo= mists have written with surprising enthusiasm [ https://substack.com/redire= ct/08e0cd86-3844-4558-b3fe-87651909a8ac?j=3DeyJ1IjoiNXFxeXF4In0.h9dEDbGj64c= TNudluPwbMYWklSu8UKGoFv6eWS-HyoA ] about Walmart and Amazon, viewing them a= s technologically-enabled planned economies. The greatest excitement has been reserved for advanced AI. Zvi Mowshowitz h= as argued [ https://substack.com/redirect/c3c830cb-7780-4398-a6fe-3d8df7e66= a1a?j=3DeyJ1IjoiNXFxeXF4In0.h9dEDbGj64cTNudluPwbMYWklSu8UKGoFv6eWS-HyoA ] t= hat AI =E2=80=9Ccan embody the preferences and knowledge of many or even al= l humans, in a way an individual human or group of humans never could.=E2= =80=9D Meanwhile, Erik Brynjolfsson and Zo=C3=AB Hitzig have made the case = that [ https://substack.com/redirect/1e21853a-6060-42fa-99a3-ac0e75512809?j= =3DeyJ1IjoiNXFxeXF4In0.h9dEDbGj64cTNudluPwbMYWklSu8UKGoFv6eWS-HyoA ], by co= mbining immense processing capacity with the ability to codify tacit knowle= dge through computer vision, language, and sensor data, AI could erode the = traditional advantages of decentralization. The optimists attack the case for traditional markets and decentralization = from multiple directions: AI can match or exceed the information-processing= advantages of markets, capture knowledge embedded in human judgment, simul= ate competition without running it, assess outcomes markets model badly thr= ough proxies, or simply replace the human participants whose limitations cr= eated the problem in the first place. Despite their diversity, many of these arguments fall into the same traps. = They routinely misstate the case for decentralization and flatten the disti= nction between different kinds of knowledge, while treating any unsolved pr= oblems as an engineering detail. The pursuit of knowledge=20 The most influential case against central planning was made on epistemologi= cal grounds by Friedrich Hayek. Oskar Lange and many of his successors read= him as making a simple point about transmission: the useful knowledge in a= ny economy is spread across millions of minds, so no central authority can = collect it fast enough to act on it. This is dangerously wrong. In =E2=80=9CThe Use of Knowledge in Society,=E2= =80=9D [ https://substack.com/redirect/ab427b6e-2838-421b-909c-84b9177893c1= ?j=3DeyJ1IjoiNXFxeXF4In0.h9dEDbGj64cTNudluPwbMYWklSu8UKGoFv6eWS-HyoA ] Haye= k distinguishes between two different types of knowledge. The first is scie= ntific or theoretical knowledge, which can be stated in general rules or pr= inciples. In theory, this kind of knowledge could be effectively concentrat= ed in a single mind or system =E2=80=93 arguably, LLMs already do this very= well. The second type is what he calls =E2=80=9Cknowledge of the particular circu= mstances of time and place.=E2=80=9D This is knowledge that is embedded in = practice, judgment, and context. For example, a farmer may know that a spec= ific field drains poorly in its southeast corner or a sales rep may notice = a change in body language with a long-standing client. This kind of knowled= ge is derived from the experience of a specific context, rather than from t= heoretical training or by performing a regression analysis. The person who = has this knowledge will frequently struggle to explain how they acquired it= =2E This concept of tacit knowledge was expressed in more detail by Michael Pol= anyi, a chemist turned philosopher of science. Polanyi famously formulated = tacit knowledge as the idea that =E2=80=9Cwe can know more than we can tell= =2E=E2=80=9D There are lots of things we c= an do that we would struggle to art= iculate. The theoretical account of riding a bike =E2=80=93 adjusting angul= ar momentum through micro-corrections in steering =E2=80=93 bears little re= lationship to the knowledge that you possess and exercise when doing it. Polanyi=E2=80=99s view is that tacit knowledge is not just knowledge that h= appens to be unstated, but instead has a distinctive architecture. Imagine a bank manager in a meeting with a local business owner, asking for= an extension to their loan. In the course of this interaction, she senses = that something isn=E2=80=99t right. The bank manager is picking up on a series of cues, such as the business ow= ner=E2=80=99s posture, the rhythm of his speech, or the differences from th= eir past interactions. The bank manager experiences all of this as a single= act of perception, rather than a series of data points. If we tried to unp= ick this knowledge and asked the bank manager to list out all the data poin= ts she picked up on, it would be akin to asking a pianist to state the prec= ise angle of each finger while she=E2=80=99s playing. Of course, the planner at this point could argue that even if the bank mana= ger can=E2=80=99t articulate these cues, we could train a model across vide= o, audio, and biometrics to detect the same patterns she=E2=80=99s detectin= g. The model doesn=E2=80=99t need to have the same experience as her, it ju= st needs to produce the same outputs. For example, driving was long held up= as an example of inalienable tacit knowledge; Polanyi himself argued that = =E2=80=9Cthe skill of a driver cannot be replaced by a thorough schooling i= n the theory of the motorcar.=E2=80=9D Despite this, we now have highly per= formant self-driving cars. This only tells us so much. Driving is mostly about seeing things and movin= g your body in response. The environment is physically constrained and the = action space is narrow. While complicated, it is markedly less ambiguous th= an navigating a dense web of human intentions and social meanings. AI excel= s at chess, but falters in complicated social reasoning games. The market i= s far more like the latter. The distinction here goes beyond difficulty. A car navigates spatial relati= onships and physical dynamics, whereas what the bank manager does is catego= rically different: she is interpreting, drawing on a framework of meaning b= uilt up through years of situated experience that organizes her perception = before any calculation begins. That framework is the structure through whic= h the interaction becomes intelligible to her at all. More processing power= has no bearing on a gap like this. By being situated in both the conversation and the wider social context, ou= r bank manager also has a few other advantages versus an impersonal system.= For a start, she has skin in the game. If she gets this call wrong, her re= putation and business could suffer; in extreme cases, her physical safety c= ould be at risk. Secondly, unlike a system observing a bunch of cues, she i= s an active participant in the interaction. Her tone and method of formulat= ing questions are all part of the interaction. Finally, she=E2=80=99s situa= ted in the community. She knows both the social norms and the realities of = running a business in that area. The snake devours its tail Eventually, we get stuck in a loop. Even if we could train an AI on these i= nteractions, what would the training data consist of? Someone has to decide= to record certain things and not others. For example, we may include the t= ranscript of the conversation, some financial metrics, and the outcome of t= he loan, but not the handshake or some of the pauses between words. The dat= a is already a selective compression of the interaction, shaped by prior hu= man decisions about what matters. The tacit knowledge that made those framing decisions is invisible to the s= ystem trained on their outputs. No dataset encounters raw reality. It arriv= es pre-shaped by decisions about what to measure and what to discard. When = you tell the system to look for indicators of trustworthiness, you=E2=80=99= ve already decided what the relevant features are, which is precisely the j= udgment you were hoping the system would replicate. We make direct perceptual contact with the world in a way that AI can=E2=80= =99t. We determine the very concepts needed to carve up the world intelligi= bly, invent new ones constantly, and make normative and aesthetic judgments= all throughout. If the data is always post-conceptual, then every training= pipeline inherits the tacit knowledge of whoever decided what to measure, = which means the system can never fully escape human judgment, even in princ= iple. The optimist could object here. Perhaps a reinforcement learning function c= ould create something functionally equivalent to skin in the game. A dynami= c AI system interacting with clients would also develop its own tone and me= thod of probing. Big enough systems don=E2=80=99t simply compute over pre-s= pecified data =E2=80=93 they learn from experience and may develop somethin= g akin to internal world models. They may absorb tacit knowledge wholesale = without anyone specifying what to look for. Even granting all of that, the question is whether it can replicate the mod= el of engagement that produces our bank manager=E2=80=99s particular sensit= ivity. She must sit inside the tension between reputational risk, relations= hip preservation, institutional obligation, and commercial judgment without= the ability to collapse them into a single metric. This is what makes her = attention so acute =E2=80=93 she can=E2=80=99t keep everyone happy. This points to something general. A system that lacks our direct perceptual= access to the world can detect statistical regularities within whatever fr= amework it=E2=80=99s been given. It won=E2=80=99t recognize the moment when= existing categories fail to capture what=E2=80=99s really happening, becau= se its measure of =E2=80=9Cwhat matters=E2=80=9D is determined by the frame= work. A reward function can=E2=80=99t evaluate its own weights. Only someon= e embedded in the situation =E2=80=93 who personally feels the weight of co= mpeting, immeasurable human risks, obligations, and norms =E2=80=93 is capa= ble of making that judgment. Information versus action The economist Israel Kirzner tells the story [ https://substack.com/redirec= t/3bde78cf-f212-43ba-ba07-b3a9d5c74f78?j=3DeyJ1IjoiNXFxeXF4In0.h9dEDbGj64cT= NudluPwbMYWklSu8UKGoFv6eWS-HyoA ] of a mother struggling with a teething ch= ild. The mother has tried everything she can to soothe or pacify the child,= but to no avail. A travelling salesman knocks on her door and offers her a= colorful toy at a price of five dollars. Her child is delighted and calms = down. On closer inspection, she is dismayed to realize that the toy is noth= ing more than a collection of marbles in a clear plastic container =E2=80= =93 something that she could have assembled in her kitchen for less than a = dollar. She =E2=80=9Ccould kick herself for not having done so.=E2=80=9D On one level, this is understandable. The mother knew she had marbles in he= r kitchen and had the wherewithal to put them in a container, but this know= ledge =E2=80=9Cdid not inspire her to action.=E2=80=9D In Kirzner=E2=80=99s= framing, she had information-knowledge (all the facts), but not action-kno= wledge (the alertness to act on the information). Even if you build a system that can collect every production function and e= very consumer preference in the economy, all you have done is assemble a co= mprehensive stock of information-knowledge. Alertness is hard to program, b= ecause it isn=E2=80=99t a process of inference from known data. A successfu= l entrepreneur acts speculatively in response to a suspected opportunity, w= hich by definition, hasn=E2=80=99t already been recognized. Hayek went further. He observed that [ https://substack.com/redirect/bb5ae5= ad-9a5f-4185-9f0e-f6e6d94b8e4c?j=3DeyJ1IjoiNXFxeXF4In0.h9dEDbGj64cTNudluPwb= MYWklSu8UKGoFv6eWS-HyoA ] much of the knowledge that matters for economic c= oordination doesn=E2=80=99t exist at all until the competitive process gene= rates it. If you are an entrepreneur who tries a new approach, whether you succeed or= fail, you have created new knowledge. As a result of the risk you took, pe= ople know whether a specific combination of resources, aimed at a set of cu= stomers, at a particular price point is viable. This knowledge didn=E2=80= =99t exist somewhere in the ether waiting to be discovered by a more powerf= ul algorithm. Instead, it was brought into existence by speculation in cond= itions of genuine uncertainty. The firm that succeeds reveals the value of = its approach retroactively. The price signals that result are not the transmissions of pre-existing dat= a, but the outputs of a distinct process. The indeterminacy runs deeper sti= ll: economic agents are not carrying fixed utility functions. Participation= in exchange changes the participants =E2=80=93 human interactions themselv= es generate new kinds of choices. The system a planner would need to model = doesn=E2=80=99t hold still, because the process of market coordination is p= artly constitutive of the preferences and knowledge it produces. Akio Morita, the co-founder of Sony, launched the Walkman in 1979 against t= he objections of the rest of the company. Morita had observed that people t= ook large stereos to the beach and listened to music in their cars, and sen= sed a market opportunity. Sony=E2=80=99s own market research and consumer s= urveys consistently suggested that there was no consumer demand for a tape = player that couldn=E2=80=99t record, no matter how portable it was. What Morita did was not pattern recognition on a richer dataset. He changed= the conceptual space, reconceiving what a music device could be and who it= could be for. No amount of data about what consumers said they wanted woul= d have produced the Walkman, because the preference for it was partly a con= sequence of the product=E2=80=99s existence. The market revealed his conjec= ture to have value, and it now retrospectively seems obvious. But obviousne= ss after the fact is the signature of knowledge that could only have been c= reated through action. The launch of the Walkman then had a series of downstream consequences =E2= =80=93 for competitors, component manufacturers, and music sales alike. The= relationship between music and everyday life changed, with enduring social= and economic consequences. If it ain=E2=80=99t broke Even if many of the foundational technological challenges could be solved b= y a future system, we would also need to believe that it would be worth the= risk. An entrepreneur who bets wrong loses his own capital. A society that= dismantles its price system has made an irreversible collective wager. We = would need to have confidence that this future model would outperform marke= ts =E2=80=93 an order that, given basic institutions, no one has to design. For the traditional Marxist, the case is straightforward. If technology can= finally solve the calculation problem, it vindicates the claim that capita= lism contains the seeds of its own succession. But as we saw earlier, the A= I alternative to traditional markets increasingly appeals to those who don= =E2=80=99t hope for the final triumph of the proletariat. Oliver Klingefjord and Joe Edelman from the Meaning Alignment Institute arg= ue that [ https://substack.com/redirect/ee441b0c-b0ca-4574-bc5f-4237eb07d00= 7?j=3DeyJ1IjoiNXFxeXF4In0.h9dEDbGj64cTNudluPwbMYWklSu8UKGoFv6eWS-HyoA ] adv= anced AI systems could correct a number of shortcomings in current markets.= They believe that markets systematically contract on proxy metrics like ho= urs, subscriptions, engagement, rather than outcomes actually delivered. Th= is is partly because doing so would be prohibitively expensive across milli= ons of consumers, but also because there is an asymmetry of power between s= uppliers and consumers. Big suppliers can write =E2=80=9Ctake it or leave i= t=E2=80=9D contracts and have huge information advantages versus those that= they are contracting with. Klingefjord and Edelman argue that replacing markets with AI could collapse= these measurement and bargaining costs, making it feasible to pay supplier= s for delivered benefit via competitive, voluntary AI intermediaries [ http= s://substack.com/redirect/9a1f902a-8961-4c81-9c26-39d8be062b0a?j=3DeyJ1Ijoi= NXFxeXF4In0.h9dEDbGj64cTNudluPwbMYWklSu8UKGoFv6eWS-HyoA ] that pool consume= rs, assess outcomes qualitatively, and negotiate enterprise-level deals. This is much more sophisticated than technosocialism, but runs into some of= the same problems. This approach tries to maintain a price system, but changes what the prices= track. But this entire system relies on an intermediary being able to asse= ss whether a good outcome was delivered. Unlike market prices, which emerge= from entrepreneurial bids under genuine uncertainty and personal risks, th= ese assessed prices would reflect a system=E2=80=99s operationalized defini= tion of human benefit. While the user might set the guardrails, the system = has to turn these into assessable criteria. In essence, it=E2=80=99s a syst= em=E2=80=99s approximation of a person=E2=80=99s approximation of what cons= titutes human flourishing.=20 This would also result in an enormous amount of discretionary judgment embe= dded in an infrastructure layer that most people would never inspect. While= you could try to mitigate this by having a world of competing AI intermedi= aries, it=E2=80=99s hard to see users choosing between rival theories of th= eir own good =E2=80=93 as operationalized by AI systems that they can=E2=80= =99t design =E2=80=93 as an obvious improvement on choosing between rival p= roducts in a traditional market. Magical thinking It is, of course, possible to argue that these objections could all be over= come. Maybe we will build systems that can collect all dispersed and local = knowledge, model genuine alertness, simulate exchange, and anticipate the o= utputs of the price discovery process without running it. It may then lead = to more efficient resource allocation. But we haven=E2=80=99t built these s= ystems, and nothing in the current trajectory suggests we are close. The bu= rden of proof lies with those who believe otherwise. The case for planning, by necessity, assumes away all real-world constraint= s while simultaneously reversing the burden of proof. In response to argume= nts about the importance of markets, it hypothesizes a system that by stipu= lation overcomes any individual objection and then challenges opponents to = prove that it=E2=80=99s impossible. Thought experiments that grant one or two premises can be genuinely useful = when thinking about advanced technology, but with each additional hypotheti= cal, the value diminishes. Any theory of how the world could work becomes p= lausible once you assume the existence of an omniscient machine god. In the end, the CIA didn=E2=80=99t need to worry about Soviet cybernetics. = Glushkov=E2=80=99s proposed cybernetics system, budgeted to cost the equiva= lent of over one trillion dollars in today=E2=80=99s money, never saw the l= ight of day. The Soviet authorities, unconvinced by his argument that the s= ystem would pay for itself several times over, balked at the price tag. Cyb= ernetics would not get to play its role in the inevitable triumph of the pr= oletariat. It seemed that even the Soviets had lost faith in the planned economy. On t= his, if little else, they were right, even if it was not for reasons that t= hey fully understood. The fundamental obstacle was never processing power o= r data collection. It was that the economy a planner would need to model is= constitutively shaped by the expectations and interpretive frameworks of t= he people who participate in it. Those frameworks shift in response to the = very act of observation and intervention. There is no fixed economy waiting= to be measured.=20 The system that a planner would need to model is the same system that the p= lan would destroy. Unsubscribe https://substack.com/redirect/2/eyJlIjoiaHR0cHM6Ly9ibG9nLmNvc21= vcy1pbnN0aXR1dGUub3JnL2FjdGlvbi9kaXNhYmxlX2VtYWlsP3Rva2VuPWV5SjFjMlZ5WDJsa0= lqb3pORGN5TlRnNU9EVXNJbkJ2YzNSZmFXUWlPakU1TURnek1qRTNOQ3dpYVdGMElqb3hOemN6T= kRFMU5EQXpMQ0psZUhBaU9qRTRNRFE1TlRFME1ETXNJbWx6Y3lJNkluQjFZaTB5TWpJMU56azBJ= aXdpYzNWaUlqb2laR2x6WVdKc1pWOWxiV0ZwYkNKOS5zVDRmNk9CbHRZejdEOFhWNi0tRDFEQ1V= sRGNoNzBTZU1NRWcyWmZPbm9VIiwicCI6MTkwODMyMTc0LCJzIjoyMjI1Nzk0LCJmIjp0cnVlLC= J1IjozNDcyNTg5ODUsImlhdCI6MTc3MzQxNTQwMywiZXhwIjoyMDg4OTkxNDAzLCJpc3MiOiJwd= WItMCIsInN1YiI6ImxpbmstcmVkaXJlY3QifQ.rr7pTQCBQPgJ1v_gxlbyAkkVFaZWwFOl5ASZZ= 1faJ3w? --b54437abcfea1cb634abd247e21b8de8de9493d14f9c4c4a14e0056f2ef0 Content-Type: text/html; charset="utf-8" Content-Transfer-Encoding: quoted-printable AI won’t fix centr= al planning 3D""
Even superintelligence needs a price
͏   ̳= 9; ­͏     ­͏     ­͏ &nb= sp;   ­͏     ­͏     ­&= #847;     ­͏     ­͏    = ; ­͏     ­͏     ­͏ &nbs= p;   ­͏     ­͏     ­&#= 847;     ­͏     ­͏    = ­͏     ­͏     ­͏  = ;   ­͏     ­͏     ­= 47;     ­͏     ­͏     = ­͏     ­͏     ­͏  =   ­͏     ­͏     ­T= 7;     ­͏     ­͏     &= #173;͏     ­͏     ­͏   =   ­͏     ­͏     ­͏= ;     ­͏     ­͏     &#= 173;͏     ­͏     ­͏   &= #8199; ­͏     ­͏     ­͏=     ­͏     ­͏     = 73;͏     ­͏     ­͏   &#= 8199; ­͏     ­͏     ­͏ =     ­͏     ­͏     = 3;͏     ­͏     ­͏   = 199; ­͏     ­͏     ­͏ &= nbsp;   ­͏     ­͏     ­= ;͏     ­͏     ­͏   Q= 99; ­͏     ­͏     ­͏ &n= bsp;   ­͏     ­͏     ­= ͏     ­͏     ­͏   ̳= 9; ­͏     ­͏     ­͏ &nb= sp;   ­͏     ­͏     ­&= #847;     ­͏     ­͏    = ; ­͏     ­͏     ­͏ &nbs= p;   ­͏     ­͏     ­&#= 847;     ­͏     ­͏    = ­͏     ­͏     ­͏  = ;   ­͏     ­͏     ­= 47;     ­͏     ­͏     = ­͏     ­͏     ­͏  =   ­͏     ­͏     ­T= 7;     ­͏     ­͏     &= #173;͏     ­͏     ­͏   =   ­͏     ­͏     ­͏= ;     ­͏     ­͏     &#= 173;͏     ­͏     ­͏   &= #8199; ­͏     ­͏     ­͏=     ­͏     ­͏     = 73;͏     ­͏     ­͏   &#= 8199; ­͏     ­͏     ­͏ =     ­͏     ­͏     = 3;͏     ­͏     ­͏   = 199; ­͏     ­͏     ­͏ &= nbsp;   ­͏     ­͏     ­= ;͏     ­͏     ­͏   Q= 99; ­͏     ­͏     ­͏ &n= bsp;   ­͏     ­͏     ­= ͏     ­͏     ­͏   ̳= 9; ­͏     ­͏     ­͏ &nb= sp;   ­͏     ­͏     ­&= #847;     ­͏     ­͏    = ; ­͏     ­͏     ­͏ &nbs= p;   ­͏     ­͏     ­&#= 847;     ­͏     ­͏    = ­͏     ­͏     ­͏  = ;   ­͏     ­͏     ­= 47;     ­͏     ­͏     = ­͏     ­͏     ­͏  =   ­͏     ­͏     ­T= 7;     ­͏     ­͏     &= #173;͏     ­͏     ­͏   =   ­͏     ­͏     ­͏= ;     ­͏     ­͏     &#= 173;͏     ­͏     ­͏   &= #8199; ­͏     ­͏     ­͏=     ­͏     ­͏     = 73;͏     ­

<= a href=3D"https://substack.com/app-link/post?publication_id=3D2225794&post_= id=3D190832174&utm_source=3Dpost-email-title&utm_campaign=3Demail-post-titl= e&isFreemail=3Dtrue&r=3D5qqyqx&token=3DeyJ1c2VyX2lkIjozNDcyNTg5ODUsInBvc3Rf= aWQiOjE5MDgzMjE3NCwiaWF0IjoxNzczNDE1NDAzLCJleHAiOjE3NzYwMDc0MDMsImlzcyI6InB= 1Yi0yMjI1Nzk0Iiwic3ViIjoicG9zdC1yZWFjdGlvbiJ9.vKX7QTQNGQ9BbZ-i3m1ZkI_JSc6tv= 4lGUTYj-VbM6W0" style=3D"color: rgb(54,55,55);text-decoration: none;">AI wo= n’t fix central planning

Even superintelligence needs a price

=
<= td align=3D"right" style=3D"height:20px;">
Forwar= ded this email? Subscribe here for= more
<= table role=3D"presentation" width=3D"auto" border=3D"0" cellspacing=3D"0" c= ellpadding=3D"0">
Alex Chalmers
<= td align=3D"right">
 
<= tr>
3D""
3D""
<= tr>
 
3D""=
New Planet by Konstantin Yuon (1921)

In 1962, Victor Glushkov pitched the Soviet authorit= ies on a nationwide cybernetics network to solve the oldest problem in soci= alist economics: how to allocate resources without private property and mar= ket prices. Washington was alarmed enough for the CIA to create a special t= askforce.

Just as Soviet industrialization was giving way t= o stagnation, the first serious computers gave the planned economy a shot i= n the arm. The central planners believed that they’d struggled to pro= cess information fast enough, so increased computing power promised a solut= ion.

In 1969, the Polish economist Oskar Lange described th= e market as “a computing device of the pre-electronic age,” sug= gesting we could put “simultaneous equations on an electronic compute= r and obtain the solution in less than half a second.” In 1971, Chile= ’s socialist government attempted a version of this managed via telex= machine, only for it to be cut short by the coup of 1973.

= Thirty years after the fall of the Soviet Union, enthusiasm for these= ideas has returned. In 2016, Jack Ma predicted that “the plan= ned economy will become increasingly big … because with access to all= kinds of data, we may be able to find the invisible hand of the market.= 221; Marxist economists have written with surprising enthusiasm abou= t Walmart and Amazon, viewing them as technologically-enabled planned econo= mies.

The greatest excitement has been reserve= d for advanced AI. Zvi Mowshowitz has argued that AI “can embo= dy the preferences and knowledge of many or even all humans, in a way an in= dividual human or group of humans never could.” Meanwhile, Erik Brynj= olfsson and Zoë Hitzig have made the case that, by combining im= mense processing capacity with the ability to codify tacit knowledge throug= h computer vision, language, and sensor data, AI could erode the traditiona= l advantages of decentralization.

The optimists atta= ck the case for traditional markets and decentralization from multiple dire= ctions: AI can match or exceed the information-processing advantages of mar= kets, capture knowledge embedded in human judgment, simulate competition wi= thout running it, assess outcomes markets model badly through proxies, or s= imply replace the human participants whose limitations created the problem = in the first place.

Despite their diversity, many of these = arguments fall into the same traps. They routinely misstate the case for de= centralization and flatten the distinction between different kinds of knowl= edge, while treating any unsolved problems as an engineering detail.

The pursuit of knowledge

The most influential= case against central planning was made on epistemological grounds by Fried= rich Hayek. Oskar Lange and many of his successors read him as making a sim= ple point about transmission: the useful knowledge in any economy is spread= across millions of minds, so no central authority can collect it fast enou= gh to act on it.

This is dangerously wrong. In “Th= e Use of Knowledge in Society,” Hayek distinguishes between= two different types of knowledge. The first is scientific or theoretical k= nowledge, which can be stated in general rules or principles. In theory, th= is kind of knowledge could be effectively concentrated in a single mind or = system – arguably, LLMs already do this very well.

The second type is what he calls “knowledge of the particular ci= rcumstances of time and place.” This is knowledge that is embedded in= practice, judgment, and context. For example, a farmer may know that a spe= cific field drains poorly in its southeast corner or a sales rep may notice= a change in body language with a long-standing client. This kind of knowle= dge is derived from the experience of a specific context, rather than from = theoretical training or by performing a regression analysis. The person who= has this knowledge will frequently struggle to explain how they acquired i= t.

This concept of tacit knowledge was expressed in more de= tail by Michael Polanyi, a chemist turned philosopher of science. Polanyi f= amously formulated tacit knowledge as the idea that “we can know more= than we can tell.” There are lots of things we can do that we would = struggle to articulate. The theoretical account of riding a bike – ad= justing angular momentum through micro-corrections in steering – bear= s little relationship to the knowledge that you possess and exercise when d= oing it.

Polanyi’s view is that tacit knowledge is no= t just knowledge that happens to be unstated, but instead has a distinctive= architecture.

Imagine a bank manager in a meeting with a l= ocal business owner, asking for an extension to their loan. In the course o= f this interaction, she senses that something isn’t right.

The bank manager is picking up on a series of cues, such as the busin= ess owner’s posture, the rhythm of his speech, or the differences fro= m their past interactions. The bank manager experiences all of this as a si= ngle act of perception, rather than a series of data points. If we tried to= unpick this knowledge and asked the bank manager to list out all the data = points she picked up on, it would be akin to asking a pianist to state the = precise angle of each finger while she’s playing.

Of = course, the planner at this point could argue that even if the bank manager= can’t articulate these cues, we could train a model across video, au= dio, and biometrics to detect the same patterns she’s detecting. The = model doesn’t need to have the same experience as her, it just needs = to produce the same outputs. For example, driving was long held up as an ex= ample of inalienable tacit knowledge; Polanyi himself argued that “th= e skill of a driver cannot be replaced by a thorough schooling in the theor= y of the motorcar.” Despite this, we now have highly performant self-= driving cars.

This only tells us so much. Driving is mostly= about seeing things and moving your body in response. The environment is p= hysically constrained and the action space is narrow. While complicated, it= is markedly less ambiguous than navigating a dense web of human intentions= and social meanings. AI excels at chess, but falters in complicated social= reasoning games. The market is far more like the latter.

T= he distinction here goes beyond difficulty. A car navigates spatial relatio= nships and physical dynamics, whereas what the bank manager does is categor= ically different: she is interpreting, drawing on a framework of meaning bu= ilt up through years of situated experience that organizes her perception b= efore any calculation begins. That framework is the structure through which= the interaction becomes intelligible to her at all. More processing power = has no bearing on a gap like this.

By being situated in bot= h the conversation and the wider social context, our bank manager also has = a few other advantages versus an impersonal system. For a start, she has sk= in in the game. If she gets this call wrong, her reputation and business co= uld suffer; in extreme cases, her physical safety could be at risk. Secondl= y, unlike a system observing a bunch of cues, she is an active participant = in the interaction. Her tone and method of formulating questions are all pa= rt of the interaction. Finally, she’s situated in the community. She = knows both the social norms and the realities of running a business in that= area.

The snake devours its tail

Eventu= ally, we get stuck in a loop. Even if we could train an AI on these interac= tions, what would the training data consist of? Someone has to decide to re= cord certain things and not others. For example, we may include the transcr= ipt of the conversation, some financial metrics, and the outcome of the loa= n, but not the handshake or some of the pauses between words. The data is a= lready a selective compression of the interaction, shaped by prior human de= cisions about what matters.

The tacit knowledge that made t= hose framing decisions is invisible to the system trained on their outputs.= No dataset encounters raw reality. It arrives pre-shaped by decisions abou= t what to measure and what to discard. When you tell the system to look for= indicators of trustworthiness, you’ve already decided what the relev= ant features are, which is precisely the judgment you were hoping the syste= m would replicate.

We make direct perceptual contact with t= he world in a way that AI can’t. We determine the very concepts neede= d to carve up the world intelligibly, invent new ones constantly, and make = normative and aesthetic judgments all throughout. If the data is always pos= t-conceptual, then every training pipeline inherits the tacit knowledge of = whoever decided what to measure, which means the system can never fully esc= ape human judgment, even in principle.

The optimist could o= bject here. Perhaps a reinforcement learning function could create somethin= g functionally equivalent to skin in the game. A dynamic AI system interact= ing with clients would also develop its own tone and method of probing. Big= enough systems don’t simply compute over pre-specified data – = they learn from experience and may develop something akin to internal world= models. They may absorb tacit knowledge wholesale without anyone specifyin= g what to look for.

Even granting all of that, the question= is whether it can replicate the model of engagement that produces our bank= manager’s particular sensitivity. She must sit inside the tension be= tween reputational risk, relationship preservation, institutional obligatio= n, and commercial judgment without the ability to collapse them into a sing= le metric. This is what makes her attention so acute – she can’= t keep everyone happy.

This points to something general. A = system that lacks our direct perceptual access to the world can detect stat= istical regularities within whatever framework it’s been given. It wo= n’t recognize the moment when existing categories fail to capture wha= t’s really happening, because its measure of “what matters̶= 1; is determined by the framework. A reward function can’t evaluate i= ts own weights. Only someone embedded in the situation – who personal= ly feels the weight of competing, immeasurable human risks, obligations, an= d norms – is capable of making that judgment.

Infor= mation versus action

The economist Israel Kirzner tells= the story of a mother struggling with a teething child. The moth= er has tried everything she can to soothe or pacify the child, but to no av= ail. A travelling salesman knocks on her door and offers her a colorful toy= at a price of five dollars. Her child is delighted and calms down. On clos= er inspection, she is dismayed to realize that the toy is nothing more than= a collection of marbles in a clear plastic container – something tha= t she could have assembled in her kitchen for less than a dollar. She ̶= 0;could kick herself for not having done so.”

= On one level, this is understandable. The mother knew she had marbles in he= r kitchen and had the wherewithal to put them in a container, but this know= ledge “did not inspire her to action.” In Kirzner’s frami= ng, she had information-knowledge (all the facts), but not action-knowledge= (the alertness to act on the information).

Even if you bui= ld a system that can collect every production function and every consumer p= reference in the economy, all you have done is assemble a comprehensive sto= ck of information-knowledge. Alertness is hard to program, because it isn&#= 8217;t a process of inference from known data. A successful entrepreneur ac= ts speculatively in response to a suspected opportunity, which by definitio= n, hasn’t already been recognized.

Hayek went f= urther. He observed that much of the knowledge that matters for econ= omic coordination doesn’t exist at all until the competitive process = generates it.

If you are an entrepreneur who tries a= new approach, whether you succeed or fail, you have created new knowledge.= As a result of the risk you took, people know whether a specific combinati= on of resources, aimed at a set of customers, at a particular price point i= s viable. This knowledge didn’t exist somewhere in the ether waiting = to be discovered by a more powerful algorithm. Instead, it was brought into= existence by speculation in conditions of genuine uncertainty. The firm th= at succeeds reveals the value of its approach retroactively.

Akio Morita, the co-founder of Sony, launched the Walkman in 1979 agai= nst the objections of the rest of the company. Morita had observed that peo= ple took large stereos to the beach and listened to music in their cars, an= d sensed a market opportunity. Sony’s own market research and consume= r surveys consistently suggested that there was no consumer demand for a ta= pe player that couldn’t record, no matter how portable it was.

What Morita did was not pattern recognition on a richer dataset. = He changed the conceptual space, reconceiving what a music device could be = and who it could be for. No amount of data about what consumers said they w= anted would have produced the Walkman, because the preference for it was pa= rtly a consequence of the product’s existence. The market revealed hi= s conjecture to have value, and it now retrospectively seems obvious. But o= bviousness after the fact is the signature of knowledge that could only hav= e been created through action.

The launch of the Walkman th= en had a series of downstream consequences – for competitors, compone= nt manufacturers, and music sales alike. The relationship between music and= everyday life changed, with enduring social and economic consequences.

=

If it ain’t broke

Even if many of the= foundational technological challenges could be solved by a future system, = we would also need to believe that it would be worth the risk. An entrepren= eur who bets wrong loses his own capital. A society that dismantles its pri= ce system has made an irreversible collective wager. We would need to have = confidence that this future model would outperform markets – an order= that, given basic institutions, no one has to design.

For = the traditional Marxist, the case is straightforward. If technology can fin= ally solve the calculation problem, it vindicates the claim that capitalism= contains the seeds of its own succession. But as we saw earlier, the AI al= ternative to traditional markets increasingly appeals to those who don̵= 7;t hope for the final triumph of the proletariat.

Ol= iver Klingefjord and Joe Edelman from the Meaning Alignment Institute argue th= at advanced AI systems could correct a number of shortcomings in = current markets. They believe that markets systematically contract on proxy= metrics like hours, subscriptions, engagement, rather than outcomes actual= ly delivered. This is partly because doing so would be prohibitively expens= ive across millions of consumers, but also because there is an asymmetry of= power between suppliers and consumers. Big suppliers can write “take= it or leave it” contracts and have huge information advantages versu= s those that they are contracting with.

Klinge= fjord and Edelman argue that replacing markets with AI could collapse these= measurement and bargaining costs, making it feasible to pay suppliers for = delivered benefit via competitive, voluntary AI intermediaries that = pool consumers, assess outcomes qualitatively, and negotiate enterprise-lev= el deals.

This is much more sophisticated than techn= osocialism, but runs into some of the same problems.

This a= pproach tries to maintain a price system, but changes what the prices track= =2E But this entire system relies on an= intermediary being able to assess whe= ther a good outcome was delivered. Unlike market prices, which emerge from = entrepreneurial bids under genuine uncertainty and personal risks, these as= sessed prices would reflect a system’s operationalized definition of = human benefit. While the user might set the guardrails, the system has to t= urn these into assessable criteria. In essence, it’s a system’s= approximation of a person’s approximation of what constitutes human = flourishing.

This would also result in an enormous amount = of discretionary judgment embedded in an infrastructure layer that most peo= ple would never inspect. While you could try to mitigate this by having a w= orld of competing AI intermediaries, it’s hard to see users choosing = between rival theories of their own good – as operationalized by AI s= ystems that they can’t design – as an obvious improvement on ch= oosing between rival products in a traditional market.

Ma= gical thinking

It is, of course, possible to argue that th= ese objections could all be overcome. Maybe we will build systems that can = collect all dispersed and local knowledge, model genuine alertness, simulat= e exchange, and anticipate the outputs of the price discovery process witho= ut running it. It may then lead to more efficient resource allocation. But = we haven’t built these systems, and nothing in the current trajectory= suggests we are close. The burden of proof lies with those who believe oth= erwise.

The case for planning, by necessity, assumes away a= ll real-world constraints while simultaneously reversing the burden of proo= f. In response to arguments about the importance of markets, it hypothesize= s a system that by stipulation overcomes any individual objection and then = challenges opponents to prove that it’s impossible.

T= hought experiments that grant one or two premises can be genuinely useful w= hen thinking about advanced technology, but with each additional hypothetic= al, the value diminishes. Any theory of how the world could work becomes pl= ausible once you assume the existence of an omniscient machine god.

In the end, the CIA didn’t need to worry about Soviet cybern= etics. Glushkov’s proposed cybernetics system, budgeted to cost the e= quivalent of over one trillion dollars in today’s money, never saw th= e light of day. The Soviet authorities, unconvinced by his argument that th= e system would pay for itself several times over, balked at the price tag. = Cybernetics would not get to play its role in the inevitable triumph of the= proletariat.

It seemed that even the Soviets had lost fait= h in the planned economy. On this, if little else, they were right, even if= it was not for reasons that they fully understood. The fundamental obstacl= e was never processing power or data collection. It was that the economy a = planner would need to model is constitutively shaped by the expectations an= d interpretive frameworks of the people who participate in it. Those framew= orks shift in response to the very act of observation and intervention. The= re is no fixed economy waiting to be measured.

The system that a planner would need to model is the same system that= the plan would destroy.

 
<= td style=3D"vertical-align:middle;">
3D""Like
3D""Comment
<= /tr>
3D""Restack
 

© 2026 Cosmos Institute=
500 W 2nd Street, Suite 1900, Austin, TX 78701
Unsubscribe

3D"Get3D"Start

<= /td>3D""3D"" --b54437abcfea1cb634abd247e21b8de8de9493d14f9c4c4a14e0056f2ef0--