NextFin News - Stanford professor, entrepreneur and longtime AI researcher Dr. Fei‑Fei Li sat down with Mishal Husain for Bloomberg Weekend in a conversation recorded at Bloomberg Audio Studios and published on November 21, 2025. Over roughly 44 minutes she set out how she sees the present moment in artificial intelligence, the responsibilities of those who build it, and the research she now calls the next frontier.
The discussion ranged from Li's personal origin story as an immigrant and student of physics to the work that made modern AI possible, and on to practical concerns about jobs, energy and the governance of powerful systems. Throughout the interview she returned to two central themes: that AI is a technology that touches civilization and that humans must retain agency over its development and use.
AI as a civilizational technology
Li framed the current era as historically consequential: This is a civilizational technology
. She said the phrase captures not just technical power but "how many people it impacts," adding that "everyone's life, work, wellbeing, future will somehow be touched by or impacted by AI." At the same time she cautioned that technology is "a double‑edged sword" and that both intentional misuse and unintended consequences must be considered.
Power, democratization and agency
Asked about concentration of capability in a few large firms, Li acknowledged the imbalance: "The major tech companies hold so much of the tech itself." She expressed a desire to see AI "much more democratized" and said individuals in this era should feel they "have the agency to impact this technology." She stressed that responsible stewardship is required "no matter who builds or holds the profound impact of this technology."
From ImageNet to spatial intelligence
Reflecting on her early research, Li described the leap that ImageNet represented: moving from tiny image datasets of a few classes to something far larger and data‑driven. She explained why image recognition was her "first North Star," and how linguistic taxonomies such as WordNet suggested ways to organize visual concepts. Looking ahead, she described World Labs' focus on "spatial intelligence" — the capacity for AI to perceive, reason about and interact in three‑dimensional spaces — and why that matters as a complement to language models.
Marble and frontier world models
Li introduced Marble as a "frontier world model" that can generate coherent 3‑D environments from simple prompts or photos. She outlined multiple use cases: designers and architects using generated worlds to ideate; game developers and VFX artists lowering production costs; robotic simulation and evaluation; and immersive education where, for example, a student could "walk into" a cell to learn biology. On purpose she emphasised that creating 3‑D worlds is a foundational capability for both human and machine intelligence.
Personal origins, resilience and curiosity
Li recounted coming to the United States at 15 with little English and helping run her parents' dry‑cleaner for seven years, joking that she was the "CEO" of that business. She linked those experiences to the resilience necessary for scientific work: "As a scientist, you have to be resilient." She credited a high school teacher, Mr. Bob Sabella, with creating opportunities that nurtured her curiosity and mathematical interests.
Global competition and collaboration
On the international landscape Li said China and the United States are both "powerhouses in AI" and described the field as global. She observed the ambition and energy across regions and countries that are investing to have roles in AI, and said the discipline's horizontal reach makes it naturally international.
Jobs, labour and the need for nuance
Addressing whether AI will destroy large numbers of jobs, Li acknowledged that "a technology as impactful and profound as AI will have a profound impact in jobs" and pointed to roles already being affected, from contact centres to software engineering tasks. But she urged a nuanced view: historical technological shifts re‑landscape labour and responses are needed at individual, company and societal levels — including upskilling and institutional responsibility.
Existential risk, governance and human responsibility
Confronted with concerns that superintelligent AI could replace humanity, Li said she "respectfully disagree[s]" with alarmist conclusions about inevitable human extinction. Her argument focused on agency and governance: even if superintelligent systems were conceivable, the real question is "why would humanity as a whole allow this to happen?" She emphasised the need for international agreements, responsible development, and collective systems of governance rather than treating the risk as purely technical inevitability.
Energy, data centres and climate considerations
On the environmental cost of large models Li acknowledged increasing energy needs for training but noted that data centres do not have to rely on fossil fuels and that energy innovation and renewable build‑out (including in regions such as the Middle East) are part of the solution. She called on countries planning large data centres to examine their energy policies as part of responsible deployment.
Teachers, communication and social impact
Li identified teachers as among her primary concerns: "My one worry is our teachers," she said, arguing they are "the backbone of our society." She asked whether the public discourse has brought teachers along, whether educators are being "superpowered" by AI tools, and whether society is communicating effectively about AI's opportunities and risks.
Role, values and guidance for young people
Li said she is conscious of the responsibility that comes with being both a researcher and a startup CEO: "Everything I do has a consequence. And that's a responsibility I shoulder." Addressing parents, she urged cultivating timeless human values and agency: teach children to be curious, honest, hardworking and critically thinking. Her simple, practical advice about devices and AI tools was: "Don't do stupid things with your tools" and "don't be lazy" — use AI to prompt better questions and deeper learning, not as a shortcut to avoid learning.
Personal reading and family life
On a lighter note, Li said she still reads technical papers but that she also enjoys reading to her children and that her favourite book for bedtime reading is Harry Potter
. She mentioned speaking Chinese with her children while their father speaks to them in Italian — a small echo of the global life she now leads.
References and further reading:
- Fei‑Fei Li Helped Create AI, Now She Feels the Responsibility — The Mishal Husain Show (iHeart), published 21 November 2025
- The Mishal Husain Show — Apple Podcasts (episode published 21 November 2025)
- ListenNotes entry for the episode (transcript available)
- Background: Reuters — Fei‑Fei Li launches World Labs (Sept 13, 2024)
Explore more exclusive insights at nextfin.ai.
