Jan 2, 2019 · 40m · a16z
a16z Podcast | Talking Humans and Machines with NYT’s John Markoff
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of the a16z podcast, veteran New York Times technology journalist John Markoff joins hosts Sonal Chokshi and Michael Copeland to examine the historical evolution, philosophical divide, and socioeconomic consequences of artificial intelligence and physical automation. Markoff offers historical perspective on how tech industry shifts, demographic realities, and system design choices shape the future relationship between humans and autonomous systems.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The host holds 15.7% of the talking time here. How this is scored →
The host as informed peer 4.0 Guest teaching 4.8 Guest disagreement 1.4 The host pushing back 2.0
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 8:00 Rejection of host's false dichotomy claim
Markoff directly rejects Sonal's premise that the AI vs IA dichotomy is false, insisting that AI developers hold distinct values focused on replacing human labor for economic efficiency.
Hardest push from the host ▶ 7:26 Sonal challenges AI vs IA dichotomy thesisSonal uses her background in human-computer interaction to directly challenge Markoff's core premise, arguing that AI and IA are not opposing sides but interdependent design realities.
Biggest teaching moment ▶ 13:32 Dismantling the Kodak employment mythMarkoff corrects popular economic doom narratives by showing Kodak failed due to self-inflicted strategic mistakes rather than digital disruption alone, pointing out Fuji's survival and job creation from the internet.
The host holds their own ▶ 29:12 Reframing automation around subsurface softwareSonal demonstrates strong sector expertise by redirecting the discussion from physical humanoid robots to W. Brian Arthur's 'second economy' of invisible backend software automation.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| Silicon Valley's Generational Evolution and Shift | 3 | 5 | 1 | 1 | Markoff provides a historical overview of Silicon Valley's shift from manufacturing to design and outlines the foundational divide between McCarthy's AI and Engelbart's IA. The hosts ask open-ended questions and show basic familiarity with Engelbart's concept. | |
| Science Fiction Inspiration and Robotic Collaboration | 6 | 6 | 4 | 6 | Sonal draws on her HCI background to challenge Markoff's core premise, arguing that the AI vs IA dichotomy is largely false. Markoff politely but firmly disagrees, maintaining that the AI community operates with a distinct capitalist mindset aimed at eliminating labor. | |
| Demographic Realities and the Demand for Automation | 4 | 5 | 1 | 1 | Markoff explains his shift from an automation alarmist to viewing robotics as necessary for aging global demographics, citing Kahneman and Autor. The conversation is collaborative, with hosts citing demographic statistics and economic theories in agreement. | |
| Technology Disruption Myths and Net Income Distribution | 3 | 6 | 2 | 1 | Markoff dismantles the popular narrative that Instagram destroyed Kodak's workforce, explaining Kodak's strategic missteps and Fuji's survival while discussing broader productivity paradoxes. Hosts listen and prompt further discussion on acceleration. | |
| Physical Constraints and the Slowdown of Moore's Law | 5 | 6 | 2 | 4 | Markoff details technical roadblocks ending Moore's law like Dennard scaling and dark silicon. Sonal pushes back gently by pointing out that non-clock performance gains are still being achieved, which Markoff acknowledges before providing further chip physics details. | |
| Future Hardware Platforms Outside Silicon Valley | 3 | 5 | 1 | 1 | The hosts inquire about non-Valley platform shifts and ubiquitous computing. Markoff recounts historical context around Xerox PARC, Mark Weiser, and Steve Jobs's early implementation of embedded microprocessors. | |
| Human Attachment to Conversational AI | 5 | 4 | 1 | 2 | Sonal offers an insightful contrast between Weiser's invisible computing vision and humanity's actual emotional attachment to iPhones. Markoff agrees and illustrates the point with user metrics from Microsoft's Xiaoice chatbot in China. | |
| Corporate AI Strategy and Physical Robotics Reality | 2 | 5 | 1 | 1 | Michael asks about corporate winners, and Markoff shares insider reporting on Google's early robotics acquisitions under Andy Rubin while noting the sobering physical limits shown in the DARPA challenge. | |
| Subsurface Automation and In-Sourcing Tasks | 6 | 4 | 1 | 3 | Sonal reframes the automation debate away from physical robots toward invisible background software and W. Brian Arthur's second economy. Markoff supports her observation by describing how call center jobs are being insourced back to US data centers as software. | |
| Balancing Economic Efficiency with Human Centric Design | 3 | 5 | 1 | 1 | Markoff discusses the tension between capitalist efficiency and human values, contrasting Stanford's intense startup culture with CMU's academic focus and advocating for a GI Bill-style retraining model. | |
| The Roots of Silicon Valley Entrepreneurial Culture | 6 | 4 | 1 | 2 | Sonal sharply distinguishes modern startup culture from visionaries like Jobs and Gates who dropped out because they had a product vision rather than wanting to be entrepreneurs first. Markoff echoes her insight with historical anecdotes from 1981. | |
| Navigating Automation, Data Privacy, and Conclusion | 2 | 3 | 1 | 1 | Markoff offers closing advice for citizens regarding data privacy and the Faustian bargain with tech platforms, wrapping up a friendly and reflective interview. |
Statements from this episode (29)
Markoff: Silicon Valley investment center shifted from Santa Clara to San Francisco
“Richard Florida, about a year ago, did this wonderful bit of research where he basically found the center of Silicon Valley by geo-coordinating current investments, and, you know, once upon a time, the center of Silicon Valley was in Santa Clara, and now it's …”
Markoff: Silicon Valley shifted from manufacturing to marketing and design
“It used to be a manufacturing center, and now it's a marketing design center”
John McCarthy believed in 1962 that working AI was a decade away
“At that point in 62, 63, McCarthy thought it would take him a decade to build a working AI.”
Markoff: AI historically overpromises and underperforms repeatedly
“AI as a field has failed many times. Really quite remarkable how it over promises and underperforms over a long time”
Markoff: Symantec was originally founded as an AI company
“One was Tech Knowledge and the other was Symantec. You know, most people don't know that it started as an AI company, but it did.”
Kaplan and Brooks pursued AI after watching 2001: A Space Odyssey
“They both decided to go into AI because of Space Odyssey, 2001. They both saw Hal and said, I want to build that.”
Markoff: The AI community prioritizes replicating human capabilities to eliminate labor
“I think that the AI community has a set of values that, that are about building systems that basically replicate human capabilities. That's their value set. And oftentimes, you know, in a, in an economic system where you can take labor out of the equation, tha…”
Copeland: Past 40 years saw more job creation than any previous period
“The evidence is there's been more jobs created and more job creation in the last, you know, 40 years than, you know, ever, right?”
Markoff backed away from his alarmist stance on AI job loss
“I was an alarmist. And gradually, over the last four or five years, I've come way over.”
Markoff: Europe is spending $1 billion to develop elder care robotics
“Europe. They're spending a billion dollars to try to develop elder robots.”
Markoff: Enterprise software and the internet hollowed out white-collar corporate jobs
“The arrival of the internet coupled with companies like SAP and PeopleSoft and Oracle and IBM who are re-engineering the white-collar clerk class inside, ah, corporations has caused this, you know, the evaporation of, the hollowing out of modern corporations.”
Markoff: US productivity growth has been zero or negative since 2008
“Since 2008, you know, where you, The modern theory is capital's replacing labor. You should see that kind of productivity growth. We've seen no productivity growth. We've even seen negative productivity.”
Markoff: Kodak failed due to strategic errors, not Instagram's disruption
“13 programmers at Instagram take out a 140,000 workers at Kodak. You know, great story, except it's not true. I mean, first, you know, the guys at Kodak made a series of strategic mistakes, and they shot themselves in the head repeatedly until they were dead. …”
Markoff: The mature internet created at least 2.5 million jobs
“Instagram, as a company, couldn't exist until there was a mature internet. How many jobs did the mature internet create? Probably two and a half million, at least, good jobs.”
Markoff: Doug Engelbart identified semiconductor scaling five years before Moore's Law
“I'm the one who stumbled across the paper in the Stanford archives that Doug Engelbart wrote in 1960, where he looked at this brand new technology of semiconductors, and he pointed out that it scales.”
Markoff: Dennard scaling stopped dead in 2006, halting clock speed growth
“Denard scaling, which was really the most important part of Silicon Valley, stopped in 2006. Just stopped dead. Clock speed stopped advancing, and so you went to parallelism and all these other kinds of technique.”
Markoff: Transistor costs have stopped falling for non-Intel semiconductor companies
“For most of the industry the cost of transistors has stopped falling. And so as you scale, this is everybody but Intel in the industry says this. You, if you talk to Broadcom, if you talk to Qualcomm, all of the customers of the foundries say, you know, going …”
Markoff predicts semiconductor scaling will permanently stop at 5 nanometers
“I believe, you know, based on talking to all the wise men I've been able to talk to in Silicon Valley, that there are three more turns of the crank. We go from 14 to 10 to seven to five. The industry can see its way to five.”
Markoff: Power savings at 7nm and 5nm will create new industries
“And there is one more big deal. If you get to five or if you get to seven it's a significant savings in power. And it's, so it's not a performance, not its cost, but that turn of the crank on power, I think that could create new platforms that will create new …”
Markoff: Steve Jobs was first to commercialize ubiquitous ambient computing
“Steve Jobs was the first guy to get that. You take a microprocessor, and you combine it with a music player, and you've got an iPod. You take a microprocessor, you combine it with a telephone, you've got a smartphone. He was the first guy to really take advant…”
Markoff: Voice conversation will become the primary interface for everyday smart devices
“And I believe that we increasingly will talk to them because that's the most obvious interface.”
Markoff: Microsoft's Chinese AI chatbot Xiaoice reached 10 million intense daily users
“They've gotten twenty million users. Ten million of them are intense users who have multiple conversations within a day”
Markoff: 25% of Microsoft's Xiaoice chatbot users told it "I love you"
“25% of the Zhao Ice users have said, I love you, to the program.”
Markoff: Google acquired robotics startups to build an automated delivery network
“When Andy Rubin was going around the valley and buying and around the world and buying up, you know, the world's best roboticist, he was sketching out a vision which I would describe as a Google competing with Amazon. And you know, and if you can kind of see t…”
Markoff: Offshore call center jobs are returning to US data centers as software
“Because of falling cost of telecommunication, many of those jobs were outsourced to the Philippines and India. And now you have this Freakonomics reverse thing happening where those jobs are coming back to run in U.S. Data centers as software, not as U.S. Work…”
Markoff: Consumers will eventually prefer talking to machines over humans for efficiency
“First people will hate talking to machines, and then just like ATMs and bank tellers, they'll realize that it's a more efficient way to get the information you want. And we will spend much of our time talking to machines.”
Markoff: Stanford has replaced its academic ivory tower with startup entrepreneurship
“Stanford, you know, is become, it's become an entrepreneurial hotbed on a scale that you can't imagine. You know, there used to be this notion of an ivory tower, and that's so gone. People pass through on their way to their startup”