Podcast Summary
Podcast: Duologue with Leslie Heaney
Episode: Thinking With Machines: AI, Human Judgment, and the Future of Intelligence with Vasant Dhar
Date: February 25, 2026
Host: Leslie Heaney
Guest: Vasant Dhar (Professor at NYU Stern School of Business, Founder of SCT Capital, AI pioneer, author of Thinking with the Brave New World of AI)
Episode Overview
This episode explores the evolution of Artificial Intelligence (AI), its current capabilities, and its far-reaching implications for society, work, medicine, and human agency. Leslie and Vasant delve into the fundamental shifts in AI, the dichotomy between expertise and common sense, the promise and limitations of AI in various domains, and the urgent need for society to reckon with questions of power, control, regulation, and responsibility as AI becomes ever more embedded in daily life.
Key Discussion Points & Insights
1. The Origins and Evolution of AI
- AI’s Beginnings:
- AI formally began in 1956, with first key applications in areas like internal medicine (e.g., Internist system at Stanford).
- Early AI relied heavily on encoding human expertise and rules.
“My first experience was...witnessing this interaction unfold between Jack Myers...and Internist the system. There was a dialogue going on...and at some point, the computer said, ‘because this question will help me discriminate between my top two hypotheses...’ and I was like, ‘holy smokes, how is a computer doing this?’” (Vasant, 07:13)
- From Expert Systems to Data-Driven AI:
- Shift from hand-coded rules to learning from data in the late 80s/90s.
- Explosion in capabilities with deep learning, allowing computers to “see” and process images as well as text.
2. The Breakthrough of Large Language Models (LLMs) and ‘General Intelligence’
- LLMs as a Paradigm Shift:
- Systems like ChatGPT solved “sentence completion” so effectively they unlocked something deeper—an ability to mimic common sense alongside expertise.
“That’s what I mean by serendipity...to do sentence completion well, [the system] had to solve a much larger problem, which is to understand the world in general.” (Vasant, 14:02)
- Systems like ChatGPT solved “sentence completion” so effectively they unlocked something deeper—an ability to mimic common sense alongside expertise.
- Common Sense vs. Expertise Dissolved:
- LLMs don’t distinguish between small talk and complex expertise; they handle both, revolutionizing interfaces and access.
“Now everyone can identify with it and use it in their own way. And that’s a huge deal…” (Vasant, 09:59)
- LLMs don’t distinguish between small talk and complex expertise; they handle both, revolutionizing interfaces and access.
3. The Bifurcation of Society: Who Thrives in an AI World?
- AI as an Amplifier of Ability:
- Those with a strong existing base of knowledge can use AI to become even more capable. Those relying solely on AI risk disempowerment.
“It’s almost like the rich get richer phenomenon. The more I know, the more I can use the intelligence to learn even more…” (Vasant, 16:32)
- Those with a strong existing base of knowledge can use AI to become even more capable. Those relying solely on AI risk disempowerment.
- Warning Against Over-Reliance:
- Using AI as a crutch, without personal understanding, leads to vulnerability in the workforce.
“If you can’t hold your own in a professional conversation without ChatGPT, you’re not adding any value…” (Vasant, 20:19)
- Using AI as a crutch, without personal understanding, leads to vulnerability in the workforce.
- Advice for Students and Learners:
- Build core competency first; then let AI amplify your skills, instead of replacing your critical thinking.
4. AI and the Future of Work
- Job Displacement and Creation:
- Routine knowledge-based jobs (not just manual labor) are threatened, akin to automation’s impact on physical labor in the past.
“This time around, it’s the brain...the machine is coming for...anyone who does routine kind of work.” (Vasant, 25:07)
- Routine knowledge-based jobs (not just manual labor) are threatened, akin to automation’s impact on physical labor in the past.
- AI as a Creative and Analytical Tool:
- Example: AI-generated financial reports allow analysts to cover 500 companies, amplifying their capabilities (21:30).
- Human oversight and interpretation are crucial even when using AI-generated outputs.
5. AI in Medicine: Promise and Limitations
- Opportunities and Cautions:
- AI can automate low-stakes medical tasks, but should not be fully trusted with high-stakes clinical judgments—AI still makes mistakes.
“I’m not comfortable at the moment, let’s say, trusting my health completely to the machine.” (Vasant, 29:41)
- AI can automate low-stakes medical tasks, but should not be fully trusted with high-stakes clinical judgments—AI still makes mistakes.
- Collaboration, Not Replacement:
- Physicians and experts now operate alongside AI—both must improve their skills accordingly.
- Data sharing and infrastructure (e.g. coordinated medical records) are lagging but crucial for future AI capabilities.
6. Regulation, Power, and Responsibility
- Company vs. Country Power:
- With few companies/operators controlling vast resources and platforms, traditional notions of government regulation may not suffice.
“...decisions about AI should not be left to a few people in Silicon Valley...” (Vasant, 44:08)
- With few companies/operators controlling vast resources and platforms, traditional notions of government regulation may not suffice.
- Urgency of Ethical and Regulatory Conversations:
- The pace of AI development calls for urgent societal debate: Who should control AI? What rights and responsibilities should agents have?
- Legal and Fiduciary Accountability:
- The need for AI systems to adopt standards similar to fiduciary duties in finance; lawsuits likely to drive (not just legislation) early regulation.
7. Existential Risks and the Question of Control
- Can We Govern More Intelligent Systems?
- As AI progresses, can a less intelligent species (humans) control a more intelligent one (AI)?
“Can a less intelligent species control a more intelligent one? ...And ChatGPT said...not really.” (Vasant, 51:00)
- As AI progresses, can a less intelligent species (humans) control a more intelligent one (AI)?
- Examples and Scenarios:
- AI agents acting autonomously, value drift in institutions (e.g., foundations run by ‘bots’), and concentration of wealth and agency.
8. Broader Societal Implications
- Environmental Impact:
- Growing AI capacity requires enormous energy and infrastructure (data centers, energy, water), which will become a global competition.
- Education & Family-Level Engagement:
- The most important place to address AI’s influence is the family, with parents engaged in how their children use and learn with AI.
“This question needs to be asked at the family level...can you tell whether [your kids] are getting smarter or dumber with AI?” (Vasant, 54:42)
- Proficiency in AI usage is easy; wisdom and judgment in using it is the new differentiator.
- The most important place to address AI’s influence is the family, with parents engaged in how their children use and learn with AI.
Notable Quotes & Memorable Moments
- On the Inescapability of AI:
“Our future of humanity, whether we like it or not, is about thinking with machines. There’s no opting out.” (Vasant, 03:41)
- On AI’s Surprising Emergent Abilities:
“...to do sentence completion well, it had to solve a much larger problem...to understand the world in general…” (Vasant, 14:02)
- On Social Bifurcation and Empowerment:
“This bifurcation that I’m talking about is something we really should be mindful of...how do we use it to get even better?” (Vasant, 17:06)
- On Regulation and Societal Choice:
“It’s time for people to really understand this technology and get engaged and get involved in shaping its future...decisions about AI should not be left to a few people in Silicon Valley...” (Vasant, 44:08)
- On The Limits of AI in Empathetic Roles:
“When feelings are involved, we should be careful with computers...mental health is very subtle, very nuanced...simulation isn’t as good as the real thing.” (Vasant, 35:49, 36:47)
- On Parental and Educational Roles:
“Can you tell whether [your kids] are getting smarter or dumber with AI? It’s an essential question...” (Vasant, 54:42)
- On Agency and Value Drift with AI:
“Would Musk say, you know what, I don’t want like a Musk foundation run by humans because I don’t trust them. I trust my Bot more than I would trust these people...” (Vasant, 41:56)
- On the Urgency of Societal Response:
“We need to all ask ourselves that question and have that conversation.” (Leslie, 58:55)
Key Timestamps
- Origins of AI, Story of Internist (05:19–08:11)
- Shift to Data-Driven AI, Emergence of Deep Learning (08:59–10:56)
- Large Language Models, Sentence Completion Breakthrough (10:56–15:29)
- Advice for Students, Dangers of Over-Reliance (18:43–21:00)
- Bot-Generated Financial Analysis Example (21:30–24:10)
- Jobs, Displacement, White Collar Automation (25:01–27:11)
- AI in Medicine, Limitations & High/Low Stakes (28:01–33:01)
- AI Agents, Autonomy, and Societal Power (37:12–42:49)
- Regulation, Duty of Care, Lawsuits (43:57–49:03)
- Environmental/Energy Impact (52:10–53:42)
- Education, Parenting, and Future Skills (54:39–57:13)
- Existential Questions and Final Takeaways (57:46–58:56)
Takeaways
- We are at a pivotal crossroads with AI. Its promise and risk are intertwined.
- The primary societal challenge is not just technological but human: ensuring we amplify rather than abdicate our own judgment, agency, and values.
- Close, informed, and ongoing societal, family, and policy engagement is essential to keep AI aligned with human interests.
For a deeper dive, read Vasant Dhar’s book “Thinking with the Brave New World of AI” which blends professional memoir and accessible AI manual, and continues the conversation opened in this must-listen episode.