What if the most valuable jobs in an AI-dominated future aren’t the white-collar knowledge worker positions we’ve spent decades optimizing for, but the blue-collar trades and human connections we’ve undervalued?
My guest today is venture capitalist Josh Wolfe. We explore how AI is transforming society, from the key obstacles slowing its progress to the unexpected factors that will determine who wins. We discuss why access to quality data beats having better AI models, how genuine human relationships become more valuable as machines take over tasks, and what it means when AI can run experiments and make discoveries around the clock.
Wolfe argues that in a world of abundant AI, the persistent advantage isn’t intelligence but behavioral understanding and authentic human relationships.
Available now: YouTube | Apple Podcasts | Spotify | Transcript
Key takeaways
- AI is currently bottlenecked by the speed of the web and biological processes, not just compute power, because even with faster processing, you can’t move faster than the underlying systems allow.
- The real moat in AI isn’t the models but proprietary data repositories because models will trend toward commoditization while unique datasets remain defensible.
- Blue-collar workers, like plumbers and maintenance workers, are more protected from AI disruption than white-collar professionals because physical tasks requiring tacit knowledge remain difficult for robots.
- Memory players like SK Hynix may become more important than compute providers as 30-50% of AI inference shifts to on-device processing rather than cloud-based systems.
- Human connection and shared experiences become the scarcest resource in an AI-abundant world because machines can’t replicate the value of commiserating about shared memories or inside jokes.
- Meta’s open-source strategy with Llama is the smartest play because they’re betting on data advantage from WhatsApp, Instagram, and Facebook while letting others improve the models.
- Passive indexation has fundamentally changed markets, with algorithms now making most investment decisions, creating potential systemic risks during market shocks.
- The asymmetry of modern warfare means a $300,000 drone can potentially neutralize a trillion-dollar aircraft carrier, fundamentally changing military strategy.
- Immigration of talented individuals remains crucial for competitiveness: 38% of AI researchers in the US are from China, while China graduates 50% of global AI undergrads.
- Everything in culture is a remix of something that came before, which raises fundamental questions about copyright and ownership when AI creates content.
- Success isn’t measured by wealth or status but by whether your children are proud of you and feel you were present for them.
- Countries that restrict AI development while competitors don’t may find themselves at a significant disadvantage, similar to military rules of engagement.
- The ability to “DVR your life” through tools like screen recording and AI memory will be a generational change in how we process and recall information.
- Machines doing science 24/7 through automated labs and hypothesis testing represents one of the most transformative near-term applications of AI.
- Africa, particularly the Sahel and Maghreb regions, represents both enormous opportunity and risk as the next potential geopolitical flashpoint.
- DOGE’s spotlight on government inefficiency is virtuous regardless of motivations because competition and transparency make systems better.
Current Obsessions and Technology Speed Limits
Josh opens by discussing his current technological obsessions, particularly the physical bottlenecks constraining progress. He explains that nature has evolved enzymes and catalysts to speed up biological reactions, but you fundamentally can’t move faster than the speed of biology. Similarly, AI agents like ChatGPT Operator can only move at the speed of the web, creating natural physics-like limits even in digital domains. This leads him to explore what technologies could accelerate these constrained systems.
“If you’ve tried ChatGPT Operator, it can only move at the speed of the web.”
AI’s Impact on Web Design and APIs
The conversation turns to how AI is forcing a fundamental redesign of web interfaces. Wolfe predicts businesses will emerge to optimize websites not for human clicks but for AI agents conducting research and making decisions. This shift extends to the potential death of APIs, as AI agents might simply interact with front-end interfaces as if they were users, bypassing the complex negotiations and restrictions of traditional API integrations. He shares his moral dilemma of clicking “I’m not a bot” while using Operator to conduct research on his behalf.
“APIs, which have been the plumbing of everything in SaaS and software and negotiating behind the scenes, may start to lose influence and power to AIs.”
The Shift to On-Device Intelligence
Wolfe challenges conventional wisdom about AI scaling, arguing that while training still requires massive GPU clusters, inference is shifting to devices. He predicts 30-50% of inference will happen on smartphones, particularly Android devices, making memory players like SK Hynix potentially more important than compute providers. This shift could fundamentally change the economics of AI, with Samsung following Intel’s trajectory of decline while Korean companies exploit regulatory arbitrage.
Data as the Ultimate Moat
The discussion reveals Wolfe’s contrarian view that proprietary data, not models, will determine AI winners. Meta’s strategy of open-sourcing Llama while maintaining control of WhatsApp, Instagram, and Facebook data exemplifies this approach. Bloomberg’s financial data, hospital systems’ medical records, and pharmaceutical companies’ failed experiments all represent untapped gold mines. He argues that models will trend toward commodity status while unique datasets remain defensible.
“The real value is going to be in the repository of data, longitudinal data, deep data.”
The Paradox of White-Collar Vulnerability
In a surprising reversal of conventional wisdom, Wolfe argues that blue-collar workers are more protected from AI disruption than elite professionals. While AI can already perform differential diagnoses better than many doctors and analyze legal documents with precision, robots still can’t effectively fold laundry or serve meals. The tacit knowledge embedded in craft and maintenance work (knowing where to tap the pipe) remains difficult to replicate.
“A robot today can’t really fully serve a meal, and they cannot effectively, even though every humanoid robot tries, fold laundry.”
The Future of Coding and Engineering
Wolfe agrees with Zuckerberg’s prediction that AI will match mid-level engineers by mid-2025, but frames this through the Jevons Paradox: making something more efficient increases rather than decreases total demand. While companies might reduce their coding workforce by 10-30%, the remaining developers become 20-100X more productive, and millions more people gain the ability to create software through natural language.
Scientific Automation and Discovery
One of the most transformative near-term applications Wolfe envisions is machines doing science 24/7. Scientists could conjure hypotheses from a beach in the Bahamas, have AI search literature, then instruct cloud labs to run experiments robotically. The challenge is engineering controlled randomness into these systems, since many breakthroughs from penicillin to Viagra emerged from serendipitous errors.
“I know something that nobody else knows and they won’t know until I tell them.”
Human Connection as Ultimate Scarcity
Drawing from a dinner with Daniel Kahneman before his death, Wolfe explores how human connection becomes irreplaceable in an AI world. Barbara Tversky‘s insight that losing people means losing partners to amplify shared memories highlights what AI can never replicate. The value isn’t just connection but the ability to commiserate about experiences only you and another person shared.
“A person will get the inside joke. And so, to your question about the advantage, the advantage is that human connection, because we are still human and we want that.”
Tesla, Truth, and Entrepreneurial Ethics
Wolfe’s critique of Tesla centers not on the technology but on what he sees as Elon Musk’s questionable relationship with truth, particularly around accounting and stock sales. He contrasts this with Jeff Bezos’s approach at Amazon, praising SpaceX as an extraordinary American asset while remaining skeptical of Tesla’s financial engineering. His broader point: the best entrepreneurs prove critics wrong through performance, not by attacking short sellers.
The Transformation of Financial Markets
The rise of passive indexation has fundamentally altered market dynamics, with algorithms making most investment decisions through indiscriminate buying and selling based on flows. Wolfe worries about systemic risks when market-cap-weighted indices accelerate both booms and busts. He predicts a return to active management as rising capital costs wash out the “funny money” of the past decade.
Geopolitical Realignments and Defense Innovation
Wolfe identifies Africa’s Sahel and Maghreb regions as potential flashpoints, with Russian mercenaries, Chinese infrastructure, and violent extremists creating conditions for “the next Afghanistan.” He advocates for hemispheric hegemony in the Americas and increased investment in defense technology, noting that companies like Anduril are inspiring a new generation to see defense work as both patriotic and profitable.
“The best way to avoid conflict is to have deterrence. If Ukraine had nuclear weapons, Putin wouldn’t have invaded.”
The Immigration Imperative
With China graduating 50% of global AI undergrads and 38% of US AI researchers being Chinese nationals, Wolfe argues for aggressive talent acquisition. He proposes automatic visas for anyone educated in America who commits to working for US companies for five years. The historical lesson is clear: attracting brilliant minds won World War II and will determine future competitiveness.
Information Processing and Tools
Wolfe describes his intensive information diet: reading 40 international newspapers daily through PressReader, using Rewind to DVR his digital life, and running parallel queries across multiple AI models. He primes AIs with specific expertise, asks for non-obvious insights versus clichés, and uses the divergence between models to surface unique perspectives. His goal is finding meta-information—what editors bury on page C22 that might actually matter.
“You are the world’s greatest expert in neuroscience. You have read every paper that has been published. You have a skeptical eye to new claims, but you are also open-minded to interesting correlations.”
The Maintenance Economy
As capital costs rise, Wolfe predicts a shift from growth to maintenance. Companies will focus on extending the life of existing assets ( think satellites, infrastructure, HVAC systems) rather than constant replacement. He’s amazed that complex systems don’t constantly break, seeing opportunity in technologies that help maintain aging infrastructure longer.
Breaking Free from Algorithmic Constraints
Wolfe explores a philosophical challenge: as AI creates increasingly accurate models of individuals based on their digital footprints, how do people break free from these algorithmic constraints? Just as college provides four years to reinvent yourself, he wonders if we need mechanisms to escape the “information mosaic” that perfectly predicts our preferences but limits our evolution.
Copyright, Creativity, and Ownership
Everything is a remix, Wolfe argues, from Beatles riffs borrowed from 1940s blues to genetic recombination creating children. If an art student can study at the Louvre, why can’t AI train on internet data? He believes prompts should confer ownership just as brush strokes do, though he worries more about breaking free from stylistic constraints than copyright protection.
The Architecture of National Competitiveness
For countries seeking a competitive advantage, Wolfe prescribes: fund basic research without directing it (preserving serendipity), implement flat taxes for simplicity, maintain strong military deterrence, create predictable regulations for capital markets, and aggressively recruit global talent. Singapore provides the model: rules-based systems where capital and people are welcome and well-treated.
Defining Success Through Family
Despite his intellectual competitiveness and drive to discover “legal secrets,” Wolfe’s ultimate measure of success is his children’s pride. Having grown up with an absent father, he prioritizes being present for every concert, game, and science fair. He shares insider information about portfolio companies with his kids, not to educate them but because he wants them to think he’s cool—a refreshingly honest admission about the selfish roots of seemingly selfless behavior.
“My dad did that. My dad made that. My dad was present for me.”
Resources
- Noise (Book)
- Westworld (Film)
- The Boys (Film)
- Everything Is a Remix (Film)
- Caught (Performance by David Parsons)
- Planet of the Apes (Film)
- Deadwood (Film)
- Game of Thrones (Film)
- Daniel Kahneman
- Barbara Tversky
