EO StudioAI Will Create New Wealth, But Not Where You Think | Carnegie Mellon University, Po-Shen Loh
CHAPTERS
- 0:00 – 2:09
AI can already replace skilled coaching—so where does human value go?
Po-Shen Loh opens by noting that today’s AI can solve advanced problems instantly, even encroaching on roles like sophisticated math coaching. He frames the central tension: if AI can do the “skill,” the future hinges on what uniquely motivates and defines humans.
- •AI can now solve complex academic problems rapidly
- •Even expert-like coaching can be automated
- •Raises the question: what remains distinctly valuable about humans?
- •Sets up the talk’s focus on opportunity and society after AI
- 2:09 – 3:09
A surprising discovery in high-poverty rural classrooms
Loh recounts visiting an impoverished rural fourth-grade classroom and being shocked by the students’ speed, curiosity, and respect for one another. The experience challenges assumptions about where talent and future economic value will come from.
- •Rural, high-poverty students answer and collaborate impressively
- •Classroom culture: respectful listening and idea-sharing
- •Contrasts observed capability with outside perceptions of the community
- •Suggests major overlooked talent pools exist
- 3:09 – 4:10
No phones, no internet—yet more creativity and authentic problem-solving
He learns the students often lack phones and reliable internet, so they create their own games and challenges. Loh argues that standard curricula under-train this kind of non-standard thinking, even though it’s exactly what the AI era demands.
- •Limited tech access can push kids toward inventiveness
- •Students make their own games instead of consuming apps
- •Standard curriculum optimizes for “standard problems”
- •Post-AI world rewards non-standard problem solving
- 4:10 – 4:40
From rural America to Africa: talent exists, but recognition networks don’t
On a trip to Africa, Loh observes many capable people and asks why economic development lags despite obvious human potential. He concludes the key missing ingredient is not talent but trusted visibility—others don’t know who to invest in or hire.
- •Africa’s growing population increases global importance
- •He observes strong capability on the ground
- •Development gap linked to lack of external awareness/visibility
- •Resource flows often go to intermediaries, not proven individuals
- 4:40 – 5:41
Building high-trust links via peer teaching: a practical network model
Loh describes a system where high schoolers coach middle schoolers in pairs, often across countries, selected for both care and real-time problem-solving ability. Co-teaching creates authentic relationships that later become channels for opportunity.
- •High schoolers are selected for empathy + on-the-spot thinking
- •Teaching in pairs builds mutual knowledge and trust
- •Cross-country pairing reveals competence on both sides
- •Relationships formed now can convert into future collaboration
- 5:41 – 7:11
Remote work + economic arbitrage: a new global wealth pipeline
He predicts that in 5–10 years, these trust-based relationships can directly shape hiring and entrepreneurship through remote work. Wage and cost-of-living differences create “split-the-difference” wins, enabling new wealth creation outside traditional hubs.
- •Remote work works; the bottleneck is trust and discovery
- •People hire those they already trust and know are capable
- •Cross-border wage differences can benefit both parties
- •Income stability can enable entrepreneurship in developing regions
- 7:11 – 7:42
Networks as the post-AI economic engine
Loh generalizes the idea: large, high-trust networks of thoughtful, service-minded people can become the foundation for opportunity in the 21st century. He positions this as a societal upgrade beyond test-ranking and credential gatekeeping.
- •“High-trust networks” can reroute resources efficiently
- •Networks reduce friction in collaboration and investment
- •Alternative to centralized selection via exams and rankings
- •Proposes a new economic flow system built on connection
- 7:42 – 8:42
The paradox of AI: as skills get automated, even ‘safe jobs’ erode
He argues AI will outperform humans across many skills and that even blue-collar work may be disrupted as humanoid robots scale. The implication: “having a skill” is less protective than having character, adaptability, and trustworthiness.
- •AI keeps improving at whatever humans do well
- •Humanoid robots threaten blue-collar “safe job” assumptions
- •Corporate incentives accelerate automation (e.g., manufacturing)
- •Shifts focus from task training to flexible human capability
- 8:42 – 9:13
What’s special about humans: visible care, trust, and moral intent
Loh claims a key human advantage is discernible intention—people can often tell, face-to-face, who genuinely cares about the bigger picture. In a risky, automated world, roles that require trust and ethical reliability become more valuable.
- •Humans can perceive genuine care through interaction
- •Robots/AI don’t provide the same trust signals
- •High-trust roles grow in importance as systems automate
- •Employability shifts toward intent + learning capacity
- 9:13 – 11:16
Interconnected tech creates new failure modes—so ‘safety people’ become essential
Using EV software as an example, he warns that increasingly networked systems are vulnerable to hacks, bugs, and AI-generated code problems. This expands demand for trustworthy people to secure, audit, and safeguard critical infrastructure.
- •Software-updated devices create systemic risk surfaces
- •Hacking or subtle code changes can cause real-world harm
- •AI-written code can produce unexpected behavior
- •Creates jobs for trusted security/safety stewards
- 11:16 – 12:16
AI democratizes learning—but motivation and critical thinking are the new bottlenecks
Loh celebrates that anyone can now learn almost anything with tools like ChatGPT or Claude, and he uses his own workflow as proof. But he cautions that AI can sound authoritative while being wrong, so curiosity and judgment matter more than ever.
- •AI provides broad, low-friction access to knowledge
- •He uses advanced models to solve problems and generate hints
- •Risk: authoritative-sounding misinformation and overreliance
- •Core question shifts from ‘can you learn?’ to ‘why/for what?’
- 12:16 – 14:16
Against ‘AI for test prep’: don’t turn students into human robots
He critiques AI-driven education that optimizes for exam performance, arguing it produces people who mimic machines rather than develop authentic thinking. He warns that the old path—rank, elite university, stable job—is breaking, fueling a mental health crisis.
- •China example: AI apps built to boost standardized exam scores
- •This trains compliance/performance rather than original thought
- •Old 20-year “study hard for a job” contract is collapsing
- •Job scarcity after credentialing can trigger major distress
- 14:16 – 17:31
A new societal operating system: thoughtful communities, entrepreneurship, and language access
Loh proposes a healthier model: connect thoughtful, caring people so they naturally create value and solve real pain points—entrepreneurship as applied empathy. He offers concrete advice (e.g., master English) and closes with a blunt forecast: stability is fading, so adaptation is required.
- •Thoughtful people cluster—networks amplify opportunity
- •Entrepreneurship = finding others’ pain points and solving them
- •Practical lever: learn English to access broader systems
- •Break out of rigid exam systems; stability will be rarer under AI
- 17:31 – 18:43
Preview: the rise of the AI-native engineer and multi-agent development
In the episode teaser, Mihail describes a Stanford course centered on AI across the software development lifecycle. He highlights the emerging “AI-native engineer” who manages fleets of agents—positioned as a new elite competency for junior developers entering the workforce.
- •AI is reshaping every stage of the SDLC
- •Stanford course demand signals rapid industry shift
- •AI-native engineers treat AI as a new operating language
- •Top performance comes from managing multiple agents effectively
