Artificial Intelligence and Geopolitics: A Reading of the Fifth RD Debate Meeting
The fifth RD Debate meeting brought together six panelists who agreed only on the diagnosis of education and split over the outlook for artificial intelligence, the role of the State, and the country's place in the value chain. From the Tabuga Think Tank, a reading based on The Inputs Trap.

The fifth episode of RD Debate, produced by CODES together with El Nuevo Diario and broadcast live from Radio Televisión Dominicana on September 20, 2026, brought together six voices to discuss the Dominican Republic's place in the global race for artificial intelligence. Víctor Gómez Casanova, Iluminada Muñoz and Mite Nishio moderated; the debaters were José Laluz, Rafael Paz, Hidekel Morrison, Jimmy Rosario as rector of ITLA, Darian Vargas and Fernando Abreu, the latter as a presidential hopeful for Única Vía. The program covered five areas: the existential-risk warning attributed to researcher Jacob Coxon, the promises in health and science, the economic impact on the country, education, and the geopolitical reordering that the technology is producing.
The most revealing finding came from the contrast among those present. The six panelists converged on a single diagnosis, education, when Darian Vargas pointed out that 92 out of every 100 Dominican students fall below the basic level in mathematics according to PISA, and Rafael Paz added that the figure rises to 98 among the poorest. On everything else, the technology's time horizon, the role of the State and the country's position in the value chain, the panel split into positions that are hard to reconcile.
Six views on the same phenomenon
José Laluz took the most accelerated end of the spectrum. He argued that artificial general intelligence will arrive before 2030, that any attempt at alignment will fail against a superintelligence, and that recursive self-improvement already writes most of the code in the labs. He also claimed that OpenAI crossed the AGI threshold on September 3 with a model he called Astra, and that before 2030 blind people will regain their sight. From that premise he derived a radical agenda: every economic sector will enter a crisis within ten years, capitalism will dissolve in the face of abundance, elite universities will disappear, and the Government should operate a digital twin of every public institution. He closed by calling for an artificial intelligence cabinet, as a matter of urgency.
Hidekel Morrison took the opposite bank with an argument about method. He called Coxon's warning alarmism and set against that individual opinion the consensus of the International AI Safety Report 2026, which rules out that current capabilities threaten the species or escape human control. He introduced a concept that ran through the rest of the debate, regulatory capture: executives at large companies exaggerate remote risks so they can write the rules themselves and close the door on emerging competitors. His most substantive contribution came on geopolitics, when he described the layered value chain, rare-earth refining in China, microprocessor design in the United States, lithography in the Netherlands, manufacturing in Taiwan, and proposed that the country identify its own niche in data centers and environmental management of its mineral resources. His final conclusion pointed to data governance as the priority missing from the whole discussion.
Rafael Paz built an intermediate position with a historical anchor. He presented artificial intelligence as a platform of power capable of rebalancing the global order, the way Colossus altered the course of the Second World War, and placed the real leap in the 2040s, when a general intelligence connects with quantum hardware. Leaning on Turing's theoretical limits, he bounded the catastrophic scenario. On the economy he supplied the most uncomfortable figures of the debate: 15 percent of graduates in STEM fields against Costa Rica's 38 percent, and 700 professionals per million inhabitants against more than 2 500 in the United States and China. He defended strategic alignment with the United States, described the Western model as public-private collaboration embodied in Stargate and the Chinese model as centralized state planning in the Hefei laboratory, and firmly upheld examinations as the basis of a meritocratic society.
Jimmy Rosario spoke from public administration. He acknowledged that the country has no chance of competing in compute when 87 of the 133 centers worldwide with capacity to process artificial intelligence belong to the United States, and placed the Dominican response in talent training: 52 000 scholarships awarded in seven months, more than 25 000 graduates and an employability rate above 90 percent before graduation. He cited the arrival of Google, the Center of Excellence in Artificial Intelligence with NVIDIA and the National Artificial Intelligence Strategy as evidence of a favorable moment, and called for an international oversight body, recalling Putin's 2017 remark about whoever controls the technology.
Darian Vargas contributed a persistently empirical point of view. He dismissed the extinction scenario this decade with a comparative argument, fentanyl and drones will kill more people than any model, and shifted attention to present harms: the war in Ukraine as a real-time artificial intelligence laboratory and social media algorithms as amplifiers of harassment. Against Laluz's forecast he brandished the World Economic Forum report on the future of work, with 170 million jobs created, 92 million eliminated and a net positive balance of 78 million by 2030, and he described the physical chain of semiconductors, NVIDIA, ASML and TSMC, to remind the audience that no intelligence manufactures its own chips. His most concrete proposal was a free-trade zone for artificial intelligence and cybersecurity, conditional on an education system able to respond to the 21st century. He recognized public policies such as RD Inteligente and the AWS and Indotel scholarships, and demanded that they be measured by job placement and contribution to GDP.
Fernando Abreu defined himself as a techno-realist and argued that the outcome is epistemologically indeterminable, which requires preparing simultaneously for the best and the worst scenario. He admitted, against his own liberal conviction, that a basic income might prove necessary during a transition period. His economic diagnosis was the harshest on infrastructure: a country with blackouts and electricity distribution in the hands of the State lacks the conditions to attract chip manufacturing. He proposed privatizing the distribution companies, adopting the American free-market model, replicating the artificial intelligence tutors of the Salvadoran system and rewarding school performance with devices. He rejected the National Artificial Intelligence Strategy as statist and refuted Mazzucato's entrepreneurial-state thesis by asserting that the private sector preceded the State in every relevant invention.
The contrasts that defined the night
Five frictions structured the debate. Laluz and Morrison clashed over the distinction between current capabilities and future projections: Morrison insisted that an agent that trains itself does so on human instruction and within defined limits, while Laluz bet on self-improvement with no ceiling. Laluz and Vargas set faith against hardware, when Vargas closed with a memorable image, “if an artificial intelligence designs, prints and manufactures its own chips before 2030, hope to be alive to give it a hug”. Abreu and Paz disputed the historical origin of innovation, with Paz citing North American R&D funds and Stargate, and Abreu attributing every advance to private capital. Abreu and Laluz debated ends and means: for Abreu, abundance is an achievement of capitalism; for Laluz, abundance dissolves scarcity and with it capitalism itself, to which he gave ten years to live. Paz and Rosario disagreed over ITLA, democratized according to the rector and distorted according to the excommunicator, who recalled that the first Dominican university chip came out of PUCMM and the most advanced programs are at INTEC.
Response from the Tabuga Think Tank to the moderators' questions
Our reading starts from the thesis we have called The Inputs Trap, published in our first annual report in 2026. The Dominican Republic leads the region in connectivity and technology investment and yet is losing ground on innovation results: from 90 to 97 place in the Global Innovation Index and from sixth to ninth in the Latin American Artificial Intelligence Index, the same position Mite Nishio cited when opening the geopolitical segment. The debate confirmed the thesis by omission. Five of six panelists spoke about adopting, consuming and training; almost none spoke about producing, measuring and governing.
On Coxon's warning and the total-replacement scenario. The relevant question for a country like ours differs from the one that was asked. The existential risk, if it exists, is decided in four laboratories and two governments, and the Dominican Republic has a concrete avenue to influence that conversation: the United Nations Global Dialogue on Artificial Intelligence Governance, where Tabuga presented a formal contribution in Geneva in July 2026. There the country should join, with its own evidence, the bloc that demands transparency on the safety evaluations of frontier models. On the substance, Morrison set a limit by countering one individual's opinion with a documented scientific consensus. It is worrying that several high-impact claims, the Astra model, AGI being crossed on a precise date, a universal cure for blindness within four years, circulated unverified on public television. A technology debate format deserves real-time source verification, and that absence was the structural weakness of the participants.
On advances in health and science. AlphaFold and the interpretation of diagnostic images are real, measurable achievements. The acceleration of clinical trials will remain limited by biology and by regulation, as Vargas noted. For the Dominican Republic, the opportunity is concentrated on three fronts: interoperability of the Servicio Nacional de Salud's data, because no diagnostic model works without structured data; imaging-based screening pilots in primary care; and participation in regional multicenter trials that currently exclude us. The prerequisite, health data governance, has yet to be built.
On economic impact and the gap between companies. The panel debated training and scholarships. Tabuga measures a different variable. About 99 percent of a market estimated at 300 million dollars in artificial intelligence solutions is purchased outside the country. A million scholarships produce competent users and, on their own, zero Dominican companies able to capture that spending. Our analysis of 417 software tenders published by the DGCP in 2025 shows more than 80 institutions buying technology independently and without interoperability standards. Each isolated purchase adds an input and subtracts an output. The gap between companies with and without artificial intelligence opens first within the State, before it does in the private market. We propose two instruments. First, public procurement of artificial intelligence with a local capacity-transfer clause and open data by default. Second, measuring scholarships through three public indicators: employment placement at twelve months, relative salary and companies created. But the instrument that answers it is missing. On semiconductor nearshoring, Abreu identified the physical constraint, energy, and Paz identified the human constraint, STEM talent. Both are measurable and neither appears with quantified targets in the Estrategia Nacional de Inteligencia Artificial.
On education and artificial intelligence tutors. PISA offers the answer: the gap predates the technology, as Morrison and Vargas agreed. Regarding tutors, the think tank's position is clear: a pilot yes, a blind replication of the Salvadoran model no. The conditions are a design with a control group, data on minor students hosted under Dominican jurisdiction, and teachers with prior training in a supervisory role. El Salvador's program still lacks a published independent evaluation. The Dominican Republic can design the region's first pilot with open evaluation, and that design would be, in itself, an exportable output. Paz's point about exams deserves support: to evaluate is to measure, and measuring is the condition of any serious public policy.
On geopolitics, digital sovereignty and regulation. Three answers. On the value chain, Morrison described the layers precisely and the conclusion is unavoidable: the country has no realistic path toward chip design or frontier models. It does have three verifiable niches. Data centers, where NAP Caribe already operates and renewable energy is the bottleneck. Applied artificial intelligence services for the Spanish-speaking market, a cognitive nearshoring that complements the manufacturing kind. And high-value sectoral data, tourism, remittances, health, which no foreign model holds and which constitute our differentiating asset. On sovereignty, the commercial alignment with the United States that Paz puts forward describes reality; sovereignty, however, is defined by a different question: where is Dominican data, under what law and with what auditing capacity. Today the State signs agreements with hyperscalers without a data inventory or localization clauses. Dominican participation in LatamGPT, the regional initiative led by CENIA Chile, is the only model-production asset where the country has a seat –via Tabuga, and it should be defended with its own budget. On regulation, Abreu and Morrison agree, from opposite shores, on the risk of regulatory capture. Tabuga adds an observation: the Estrategia Nacional, as currently formulated, creates structure before metrics. Regulating the consumption of a technology that others produce yields little; measuring inputs and outputs yields a lot. The minimum proposal is a national artificial intelligence observatory with three annual public series: investment, adoption by sector, and local companies and patents. With that series, ninth place among nineteen countries stops being a ranking figure and becomes a policy target.
Food for thought
The panel produced the most honest education diagnosis to have aired on Dominican free-to-air television this year. At the same time, it produced zero measurable proposals on local production of artificial intelligence. That asymmetry, plenty of knowledge about what the country consumes and silence about what it could create, is The Inputs Trap that the Tabuga Think Tank has been pointing to. The national conversation will move forward when the next edition asks, with the same intensity, which Dominican company, which Dominican data and which Dominican model will come out of all this effort.