By Javier Surasky
Original in
Spanish; translated with ChatGPT assistance
.png)
Knowledge and Power in Artificial Intelligence
The
Massachusetts Institute of Technology (MIT) occupies a position in the global
artificial intelligence ecosystem in which its influence stems not only from
the research it produces, but also from its capacity to train talent, mobilize
resources, connect academia and industry, and turn scientific knowledge into
innovation, companies, and frameworks for thinking about public policy.
Viewing MIT
through a power map raises a question: how does an academic institution
accumulate and transform knowledge into influence? This perspective is similar
to the one we already applied in our AI power map to cases such as Tsinghua University and the Stanford Institute for Human-Centered Artificial Intelligence (HAI).
As with the
Chinese university, AI cuts across different institutional and academic spaces
at MIT, but two institutions stand out.
The Computer
Science and Artificial Intelligence Laboratory (CSAIL) is the Institute’s
main computer science research laboratory and works across fields including
artificial intelligence, machine learning, natural language processing,
robotics, computer vision, cryptography, and systems, among others. According to
the laboratory’s own figures, it brings together more than 1,700 members,
around 60 research groups, and more than 900 active projects (MIT CSAIL
Alliances, n.d.).
The MIT
Stephen A. Schwarzman College of Computing, meanwhile, was institutionally
launched in 2019 as part of a large-scale transformation. It was conceived
through a $1 billion commitment, including a foundational $350 million gift
from Stephen A. Schwarzman. The project envisioned creating 50 new faculty
positions and a structure designed to integrate computing and AI with the
Institute’s other academic fields (MIT News Office, 2018).
Its
benefactor deserves attention. Schwarzman is the co-founder and CEO of
Blackstone and has had a significant presence in U.S. politics. He chaired the Strategic
and Policy Forum created during Donald Trump’s first administration and has
been a major Republican donor; in 2024, he once again publicly backed Trump’s
presidential candidacy (Reuters, 2024; Viebeck, 2016).
The College's
history also introduces a geopolitical dimension. Schwarzman explained that his
decision to support its creation was prompted, at least in part, by trips to
China, where he observed the intensification of investment in artificial
intelligence and became concerned about ensuring that the United States
remained at the forefront of the technology. In his words, he sought to
strengthen U.S. competitiveness (Dizikes, 2019).
This
reminds us that the production of frontier knowledge depends on human talent,
but also on large concentrations of private capital, and that philanthropy
makes it possible to transform part of that economic capital into scientific
infrastructure, institutional prestige, and the capacity to intervene in
strategic fields such as AI.
MIT’s power
becomes more visible when we examine its connections. CSAIL maintains research
relationships with both public and private actors. U.S. agencies funding its
research include DARPA, the Department of the Air Force, NASA, the National
Institutes of Health, and the National Science Foundation; its non-federal
sponsors include companies and organizations such as Boeing, Intel, Lockheed
Martin, SAP, Toyota, IBM, and Pfizer (MIT CSAIL, n.d.). The Schwarzman College,
meanwhile, maintains collaborative programs with Amazon, Google, the Hasso
Plattner Institute, and the Mohamed bin Zayed University of Artificial
Intelligence, among other organizations (MIT Schwarzman College of Computing,
n.d.).
In February
2025, MIT also introduced the Generative AI Impact Consortium, whose six
founding members were Analog Devices, Coca-Cola, OpenAI, Tata Group, SK
Telecom, and TWG Global. MIT itself describes the initiative as a bridge
between academia and industry: companies contribute challenges, data, and
specialized knowledge, while researchers work on those problems. The consortium
includes workshops and discussions designed to identify and prioritize research
challenges (McDonnell, 2025).
Going
further, in April 2026 MIT also established the MIT-IBM Computing Research
Lab, expanding a relationship spanning nearly a decade. Its predecessor,
the MIT-IBM Watson AI Lab, had funded more than 210 projects, involved
more than 150 MIT faculty members and researchers and more than 200 IBM
researchers, produced more than 1,500 peer-reviewed papers, and funded more
than 500 students and postdocs (MIT News, 2026).
These
relationships reveal MIT as an actor with sufficient power and convening
capacity to bring scientific research, technology companies, capital, public
agencies, and talent development together within the same institutional space.
This
proximity between academia and industry raises a sensitive question. CSAIL
Alliances explicitly serves as the channel through which companies and
organizations gain access to the laboratory’s researchers, students, emerging
technologies, and startups. The program itself states that some initiatives
allow corporate members to provide input that helps shape research projects.
CSAIL also argues that industry funding fills an important need in sustaining research
advancement (MIT CSAIL Alliances, n.d.).
None of
this demonstrates that companies control MIT’s scientific agenda. It does,
however, allow us to pose a question that matters for a power map: who has the
resources and access needed to turn their problems into research problems for
the institution? And, conversely, which issues in need of research may be
pushed to the margins?
From this
perspective, a problem of dependence, access, and unequal capacity to
participate in shaping technological priorities emerges. Not as an anomaly
specific to MIT, but as a structural tension in contemporary artificial
intelligence research.
The Power to Define Problems
Another
form of influence is at work here, less tangible than funding: the capacity to
produce the categories through which other actors interpret artificial
intelligence.
MIT has
developed initiatives specifically aimed at connecting technical knowledge and
public policy. Created in late 2020 and convened by the Schwarzman College, the
AI Policy Forum brings together scientists, technologists, policymakers,
and business leaders to move debates about AI principles toward practical
implementation and develop guidance for governments and companies (MIT AI
Policy Forum, 2020, 2022).
This
capacity takes on another dimension in MIT’s work on risk. The AI Risk
Repository was publicly introduced in August 2024 as a database designed to
systematize risks identified across existing frameworks and taxonomies. The MIT
AI Risk Initiative, based at MIT FutureTech, expanded this work through a
range of tools designed to identify, prioritize, and manage AI-related risks
(MIT FutureTech, 2024; MIT AI Risk Initiative, n.d.).
Among these
tools is the AI Governance Map, which currently classifies more than a
thousand cases contained in documents on AI governance and regulation from the
United States and other parts of the world, examining dimensions such as the
risks covered, regulated sectors, actors involved, and stages of the systems’
life cycle (MIT AI Risk Initiative, 2026a).
These
initiatives clearly perform a knowledge-systematization function. But viewed
through the lens of power, they reveal something else: to classify is also to
intervene.
Defining
what counts as a risk, how problems are grouped, which actors are identified as
responsible, or which instruments are considered relevant helps structure the
field within which governments, companies, and organizations subsequently
debate their decisions.
Here, one
finding is particularly revealing. An update published by the initiative itself
in April 2026 found that the more than one thousand documents analyzed devoted
considerably more attention to risks related to safety, privacy, and
transparency than to socioeconomic risks such as economic devaluation or the
centralization of power (MIT AI Risk Initiative, 2026b).
This
observation is particularly relevant to a power map: how AI is governed also
depends on what the institutions responsible for producing and organizing
knowledge can identify as a problem.
Academic Power Must Also Be Questioned
Major
universities often appear in debates about artificial intelligence as sources
of expert knowledge that are relatively distinct from governments and
companies. That distinction remains important, but it is insufficient to
understand how the technological ecosystem operates today.
MIT reveals
a far more interconnected structure: research funded by public agencies and
companies; programs designed to connect companies with researchers and
students; consortia that incorporate industry problems and knowledge; and
spaces aimed at translating academic knowledge into tools for governments and
businesses.
Precisely
for this reason, its position deserves critical scrutiny.
Interactions
with companies or governments can produce knowledge and socially valuable
applications, but they create a tension around preserving the university’s
intellectual autonomy and its critical capacity when the resources and
infrastructure needed to research increasingly complex technologies are
distributed in profoundly uneven ways.
It also
matters who remains outside these networks. If a significant part of the
process of defining technological problems circulates among elite universities,
large corporations, governments with substantial technological capacity, and
specialized organizations, we should ask what capacity other regions,
institutions, and social groups have to ensure that their problems are
recognized as legitimate research problems.
We are not
claiming that these voices are necessarily excluded from MIT. The issue is
structural: access to the spaces where knowledge, categories, and priorities
are produced is itself a form of power.
Legitimized Power
MIT does
not occupy the same position in artificial intelligence as a company that
controls a foundation model, a global platform, or large data centers. Its
power is different.
It is,
fundamentally, scientific, institutional, and epistemic power.
It consists
of producing knowledge, training those who will build future technologies,
connecting actors with different resources, and helping determine which
questions merit research and which categories we use to interpret their
consequences.
This is why
MIT is central to an artificial intelligence power map. Its case shows that
technological power does not reside solely in owning infrastructure, capital,
or models. It also lies in something prior: the capacity to produce legitimate
knowledge about technology and to influence the definition of the problems that
society will have to address.
And here a
question arises that extends beyond MIT itself: if a small number of
institutions simultaneously concentrate scientific prestige, economic
resources, talent, and access to decision-makers, who defines the agenda
through which we think about the future of artificial intelligence?
Key Facts
- The Massachusetts Institute of
Technology was founded in 1861 and opened its doors to students in 1865,
in a context shaped by U.S. industrialization and by a conception of
education oriented toward connecting scientific knowledge with practical
problems.
- In fiscal year 2025, MIT
reported an endowment of approximately $27.4 billion, with 80% of that
amount subject to restrictions on its use (MIT News, 2025).
- 106 people affiliated with MIT,
including alumni, faculty members, and other members of its community,
have received the Nobel Prize (MIT, n.d.-a).
- During fiscal year 2025, MIT
recorded 684 invention disclosures, filed 623 new U.S. patent
applications, was granted 282 U.S. patents, and supported 38 companies in
using MIT intellectual property (MIT, n.d.-b).
- MIT maintains institutional
research and education initiatives across different regions of the world
through programs such as MIT International Science and Technology
Initiatives (MISTI), which connects students and faculty with
universities, companies, and international organizations (MIT
International Science and Technology Initiatives, n.d.).
References
Dizikes, P.
(2019, March 1). For founders of new college of computing, the human element
is paramount. MIT News. https://news.mit.edu/2019/founders-new-college-computing-human-element-reif-schwarzman-0301
McDonnell,
L. (2025, February 3). Introducing the MIT Generative AI Impact Consortium.
MIT Computer Science and Artificial Intelligence Laboratory. https://www.csail.mit.edu/news/introducing-mit-generative-ai-impact-consortium
MIT.
(n.d.-a). Awards & honors. MIT Facts. https://facts.mit.edu/awards-honors/
MIT.
(n.d.-b). Technology licensing. MIT Facts. https://facts.mit.edu/
MIT AI
Policy Forum. (2020, October 19). A global collaboration to move artificial
intelligence principles to practice. MIT Schwarzman College of Computing. https://computing.mit.edu/news/a-global-collaboration-to-move-artificial-intelligence-principles-to-practice/
MIT AI
Policy Forum. (2022, September 23). Q&A: Global challenges surrounding
the deployment of AI. MIT Schwarzman College of Computing. https://computing.mit.edu/news/the-global-challenges-surrounding-the-deployment-of-ai/
MIT AI Risk
Initiative. (2026a). How is AI being governed? https://airisk.mit.edu/ai-governance
MIT AI Risk
Initiative. (2026b, April 9). Mapping the AI Governance Landscape: April
2026 Update. https://airisk.mit.edu/blog/mapping-the-ai-governance-landscape-april-2026-update
MIT CSAIL.
(n.d.). Strategic partners. Massachusetts Institute of Technology. https://www.csail.mit.edu/sponsors/strategic-partners
MIT CSAIL
Alliances. (n.d.). CSAIL Alliances. Massachusetts Institute of
Technology. https://www.csail.mit.edu/engage/csail-alliances
MIT CSAIL
Alliances. (n.d.). CSAIL
by the numbers. Massachusetts Institute of Technology. https://cap.csail.mit.edu/engage/why-csail-alliances/csail-numbers
MIT
FutureTech. (2024, August 14). Presenting the AI Risk Repository.
Massachusetts Institute of Technology. https://futuretech.mit.edu/news/presenting-the-ai-risk-repository
MIT AI Risk
Initiative. (n.d.). About us. Massachusetts Institute of Technology. https://airisk.mit.edu/about
MIT
International Science and Technology Initiatives. (n.d.). MISTI is the hub
for global experiences at MIT. About MISTI, Massachusetts Institute of
Technology. https://misti.mit.edu/about
MIT News.
(2025, October 10). MIT releases financials and endowment figures for 2025.
Massachusetts Institute of Technology. https://news.mit.edu/2025/mit-releases-financials-and-endowment-figures-1010
MIT News.
(2026, April 29). The MIT-IBM Computing Research Lab launches to shape the
future of AI and quantum computing. Massachusetts Institute of Technology. https://news.mit.edu/2026/mit-ibm-computing-research-lab-launches-0429
MIT News
Office. (2018, October 15). MIT reshapes itself to shape the future.
Massachusetts Institute of Technology. Massachusetts Institute of Technology. https://computing.mit.edu/news/mit-reshapes-itself-to-shape-the-future/
MIT
Schwarzman College of Computing. (n.d.). External collaborations.
Massachusetts Institute of Technology. https://computing.mit.edu/external-collaborations/
Reuters.
(2024, May 24). Blackstone CEO Schwarzman to back Trump, Axios reports. https://www.reuters.com/world/us/blackstone-ceo-schwarzman-back-trump-axios-reports-2024-05-24/