StarZero All shows Try the podcast clipper →

Aug 21, 2017 · 30m · 20vc

20VC: YC's Daniel Gross on How YC Can Democratise AI & Reduce Incumbency Advantages, Why ML Enabled Software Will Eat The Software That Ate The World & Whether AI Will Produce Independent Companies or Be Technology within Incumbents

Daniel Gross · 22m spoken Harry Stebbings · 6m spoken
0:00 / 0:00

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

In this episode of The 20 Minute VC, Y Combinator partner Daniel Gross discusses the practical realities of artificial intelligence, strategies for AI startups to compete with tech incumbents, and his personal journey from entrepreneur to venture investor. Gross introduces the YCAI initiative while sharing actionable insights on startup data moats, machine learning business models, and productivity practices.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 23.4% of the talking time here. How this is scored →

Harry as informed peer 3.4 Guest teaching 4.2 Guest disagreement 2.0 Harry pushing back 2.2
05100:0010:0020:0030:002:15–8:01 · Harry as informed peer 1/10 Daniel Gross's Startup Journey and Transition to YC Daniel shares his background coming to Silicon Valley from Israel, pivoting his startup Greplin 48 hours before YC Demo Day, selling to Apple, and returning as a YC partner. Harry remains almost entirely in listening mode, allowing Daniel to tell his story uninterrupted.8:01–12:41 · Harry as informed peer 3/10 Defining AI Realities vs. Hype Harry introduces a provocative quote from Aaron van der Vende claiming AI is mostly a scam. Daniel reframes the claim, educating the host on how to categorize AI technology into commoditized models, speculative research, and newly practical applications like computer vision and speech recognition.12:41–16:46 · Harry as informed peer 4/10 AI Startup Viability and the Launch of YCAI Harry asks a structured question about whether AI will produce standalone companies or remain a sustaining innovation for incumbents. Daniel breaks down the three structural moats large incumbents hold (talent, compute capital, and data) and explains how YCAI is designed to democratize those resources for startups.16:46–21:11 · Harry as informed peer 5/10 Overcoming Data Moats, Transfer Learning, and Ambiguity Harry challenges Daniel's stance on data moats by citing Dennis Mortensen's strategy of deep domain specialization, and later asks about Aaron van der Vende's quote on ambiguity. Daniel politely dissents on specialization being a total cure due to activation energy barriers, and educates on transfer learning.21:11–29:00 · Harry as informed peer 4/10 Business Models in Machine Learning Harry asks if Daniel is concerned that ML lacks a foundational business model, leading Daniel to directly disagree and reframe ML as software eating software. The segment transitions into a quickfire round covering personal productivity, meditation, rejection frameworks, and decaf coffee.2:15–8:01 · Guest teaching 1/10 Daniel Gross's Startup Journey and Transition to YC Daniel shares his background coming to Silicon Valley from Israel, pivoting his startup Greplin 48 hours before YC Demo Day, selling to Apple, and returning as a YC partner. Harry remains almost entirely in listening mode, allowing Daniel to tell his story uninterrupted.8:01–12:41 · Guest teaching 6/10 Defining AI Realities vs. Hype Harry introduces a provocative quote from Aaron van der Vende claiming AI is mostly a scam. Daniel reframes the claim, educating the host on how to categorize AI technology into commoditized models, speculative research, and newly practical applications like computer vision and speech recognition.12:41–16:46 · Guest teaching 5/10 AI Startup Viability and the Launch of YCAI Harry asks a structured question about whether AI will produce standalone companies or remain a sustaining innovation for incumbents. Daniel breaks down the three structural moats large incumbents hold (talent, compute capital, and data) and explains how YCAI is designed to democratize those resources for startups.16:46–21:11 · Guest teaching 5/10 Overcoming Data Moats, Transfer Learning, and Ambiguity Harry challenges Daniel's stance on data moats by citing Dennis Mortensen's strategy of deep domain specialization, and later asks about Aaron van der Vende's quote on ambiguity. Daniel politely dissents on specialization being a total cure due to activation energy barriers, and educates on transfer learning.21:11–29:00 · Guest teaching 4/10 Business Models in Machine Learning Harry asks if Daniel is concerned that ML lacks a foundational business model, leading Daniel to directly disagree and reframe ML as software eating software. The segment transitions into a quickfire round covering personal productivity, meditation, rejection frameworks, and decaf coffee.2:15–8:01 · Guest disagreement 0/10 Daniel Gross's Startup Journey and Transition to YC Daniel shares his background coming to Silicon Valley from Israel, pivoting his startup Greplin 48 hours before YC Demo Day, selling to Apple, and returning as a YC partner. Harry remains almost entirely in listening mode, allowing Daniel to tell his story uninterrupted.8:01–12:41 · Guest disagreement 2/10 Defining AI Realities vs. Hype Harry introduces a provocative quote from Aaron van der Vende claiming AI is mostly a scam. Daniel reframes the claim, educating the host on how to categorize AI technology into commoditized models, speculative research, and newly practical applications like computer vision and speech recognition.12:41–16:46 · Guest disagreement 1/10 AI Startup Viability and the Launch of YCAI Harry asks a structured question about whether AI will produce standalone companies or remain a sustaining innovation for incumbents. Daniel breaks down the three structural moats large incumbents hold (talent, compute capital, and data) and explains how YCAI is designed to democratize those resources for startups.16:46–21:11 · Guest disagreement 3/10 Overcoming Data Moats, Transfer Learning, and Ambiguity Harry challenges Daniel's stance on data moats by citing Dennis Mortensen's strategy of deep domain specialization, and later asks about Aaron van der Vende's quote on ambiguity. Daniel politely dissents on specialization being a total cure due to activation energy barriers, and educates on transfer learning.21:11–29:00 · Guest disagreement 4/10 Business Models in Machine Learning Harry asks if Daniel is concerned that ML lacks a foundational business model, leading Daniel to directly disagree and reframe ML as software eating software. The segment transitions into a quickfire round covering personal productivity, meditation, rejection frameworks, and decaf coffee.2:15–8:01 · Harry pushing back 0/10 Daniel Gross's Startup Journey and Transition to YC Daniel shares his background coming to Silicon Valley from Israel, pivoting his startup Greplin 48 hours before YC Demo Day, selling to Apple, and returning as a YC partner. Harry remains almost entirely in listening mode, allowing Daniel to tell his story uninterrupted.8:01–12:41 · Harry pushing back 2/10 Defining AI Realities vs. Hype Harry introduces a provocative quote from Aaron van der Vende claiming AI is mostly a scam. Daniel reframes the claim, educating the host on how to categorize AI technology into commoditized models, speculative research, and newly practical applications like computer vision and speech recognition.12:41–16:46 · Harry pushing back 2/10 AI Startup Viability and the Launch of YCAI Harry asks a structured question about whether AI will produce standalone companies or remain a sustaining innovation for incumbents. Daniel breaks down the three structural moats large incumbents hold (talent, compute capital, and data) and explains how YCAI is designed to democratize those resources for startups.16:46–21:11 · Harry pushing back 4/10 Overcoming Data Moats, Transfer Learning, and Ambiguity Harry challenges Daniel's stance on data moats by citing Dennis Mortensen's strategy of deep domain specialization, and later asks about Aaron van der Vende's quote on ambiguity. Daniel politely dissents on specialization being a total cure due to activation energy barriers, and educates on transfer learning.21:11–29:00 · Harry pushing back 3/10 Business Models in Machine Learning Harry asks if Daniel is concerned that ML lacks a foundational business model, leading Daniel to directly disagree and reframe ML as software eating software. The segment transitions into a quickfire round covering personal productivity, meditation, rejection frameworks, and decaf coffee.

speaking balance: gold is Harry, purple is the guest (3 minute bins)

0:00 · Harry 87.5% · guest 12.5%0:00 · Harry 87.5% · guest 12.5%3:00 · Harry 0% · guest 100%3:00 · Harry 0% · guest 100%6:00 · Harry 14.9% · guest 85.1%6:00 · Harry 14.9% · guest 85.1%9:00 · Harry 0% · guest 100%9:00 · Harry 0% · guest 100%12:00 · Harry 11.8% · guest 88.2%12:00 · Harry 11.8% · guest 88.2%15:00 · Harry 18.2% · guest 81.8%15:00 · Harry 18.2% · guest 81.8%18:00 · Harry 10.9% · guest 89.1%18:00 · Harry 10.9% · guest 89.1%21:00 · Harry 17.8% · guest 82.2%21:00 · Harry 17.8% · guest 82.2%24:00 · Harry 10.8% · guest 89.2%24:00 · Harry 10.8% · guest 89.2%27:00 · Harry 46% · guest 54%27:00 · Harry 46% · guest 54%30:00 · Harry 100% · guest 0%30:00 · Harry 100% · guest 0%
Sharpest disagreement ▶ 21:22 Disagreeing on ML Business Models

Daniel directly rejects Harry's question premise about a lack of ML business models, stating 'Well, I would disagree' and arguing ML-enabled software will eat traditional software.

 Hardest push from Harry ▶ 16:38 Challenging Data Moat Assumptions

Harry explicitly pushes back on Daniel's perspective regarding data moats by citing x.ai founder Dennis Mortensen's counterargument on deep domain specialization.

 Biggest teaching moment ▶ 8:23 Categorizing AI Tech Realities

Daniel dismantles the provocative 'AI is a scam' premise by methodically categorizing machine learning into legacy commodities, far-off research, and actionable middle-tier breakthroughs.

 Harry holds his own ▶ 16:38 Citing Specialized Counterarguments

Harry demonstrates strong preparation and industry knowledge by pitting Dennis Mortensen's tactical insights against Daniel's broad assertions about data advantages.
the scores for every segment, with the reasoning behind each
ChapterTopicHarry as informed peerGuest teachingGuest disagreementHarry pushing backWhy
Daniel Gross's Startup Journey and Transition to YC 1100 Daniel shares his background coming to Silicon Valley from Israel, pivoting his startup Greplin 48 hours before YC Demo Day, selling to Apple, and returning as a YC partner. Harry remains almost entirely in listening mode, allowing Daniel to tell his story uninterrupted.
Defining AI Realities vs. Hype 3622 Harry introduces a provocative quote from Aaron van der Vende claiming AI is mostly a scam. Daniel reframes the claim, educating the host on how to categorize AI technology into commoditized models, speculative research, and newly practical applications like computer vision and speech recognition.
AI Startup Viability and the Launch of YCAI 4512 Harry asks a structured question about whether AI will produce standalone companies or remain a sustaining innovation for incumbents. Daniel breaks down the three structural moats large incumbents hold (talent, compute capital, and data) and explains how YCAI is designed to democratize those resources for startups.
Overcoming Data Moats, Transfer Learning, and Ambiguity 5534 Harry challenges Daniel's stance on data moats by citing Dennis Mortensen's strategy of deep domain specialization, and later asks about Aaron van der Vende's quote on ambiguity. Daniel politely dissents on specialization being a total cure due to activation energy barriers, and educates on transfer learning.
Business Models in Machine Learning 4443 Harry asks if Daniel is concerned that ML lacks a foundational business model, leading Daniel to directly disagree and reframe ML as software eating software. The segment transitions into a quickfire round covering personal productivity, meditation, rejection frameworks, and decaf coffee.

Statements from this episode (11)

Opinion
Gross: AI in 2017 mirrors the 2005 big data hype cycle
“AI today is a little bit like that kind of big data trend everyone talked about in 2005 and six, where it's changing the world in a bunch of different ways that I'll detail in a moment. But it's not really living up to, I think, all the hype, and that's probab…”
Daniel Gross Aug 21, 2017 ▶ 8:50
Insight
Gross: Humans consistently fail to predict when research becomes a commodity
“I also think that we are very, very bad as a species at predicting when things that are the far end of research are about to become a commodity.”
Daniel Gross Aug 21, 2017 ▶ 11:56
Prediction Not checkable as stated
Gross: Breakthroughs in reinforcement learning will transition to market quickly
“Because when something really big happens with, say, reinforcement learning, which is an area mostly attributed to research today, it's going to happen really quickly.”
Daniel Gross Aug 21, 2017 ▶ 12:25
Assertion Not checkable as stated
Gross: No independent company exists solely because of modern ML
“Currently, the way you kind of look at the world today, there is yet to be an independent entity that you could really say is an AI company, or I think a better framework for this only exists because of the kind of latest and greatest machine learning technolo…”
Daniel Gross Aug 21, 2017 ▶ 12:57
Assertion Supported
Gross: Autonomous driving startups are operating without LIDAR
“And there certainly are startups that are working on autonomy, literally with cameras and deep learning alone without the help of LIDAR.”
Daniel Gross Aug 21, 2017 ▶ 13:38
Disclosure
Gross: Y Combinator launches YCAI as its first AI vertical
“We launched this thing called YCAI, which is an experiment, and it's a really, it's our first vertical of YC that's dedicated to AI”
Daniel Gross Aug 21, 2017 ▶ 15:11
What-if
Gross: Airbnb likely would not exist without Michael Seibel and YC
“It is less likely that Airbnb would have been created had Michael Seibel not recruited the founders to Y Combinator, introduced them to Paul Graham, the program had been set up in place, the investment been available, demo day being available”
Daniel Gross Aug 21, 2017 ▶ 17:32
Insight
Gross: Bad engineers over-apply AI to problems lacking ambiguity
“Really great machine learning engineers have the same characteristic that really great all engineers have, which is they're kind of lazy, and they want to find the kind of Occam's razor, the dumbest way to fix the problem, and the bad ones want to kind of over…”
Daniel Gross Aug 21, 2017 ▶ 20:41
Prediction Not checkable as stated
Gross: Machine learning-enabled software will eat traditional software
“I think that there are going to be business models of taking AI and applying it to a vertical in the same way that you could literally take the advent of the database and apply it to a vertical. You know, software eats the world. Is a quote I think originally …”
Daniel Gross Aug 21, 2017 ▶ 21:24
Prediction Not checkable as stated
Gross: AI platform startups face tough competition from TensorFlow and incumbents
“Second is for the platform companies, I do think there's a less obvious but likely business model as well. Less obvious because unlike the database analogy I provided, it's not immediately clear that there'll be An Oracle or a MySQL, that is to say, a company …”
Daniel Gross Aug 21, 2017 ▶ 21:50
Disclosure
Daniel Gross reveals angel investment in Parker Conrad's Rippling
“My most recently announced investment is a company called Rippling, which was started by the CEO of Zenefits, Parker Conrad”
Daniel Gross Aug 21, 2017 ▶ 27:46
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 1,200 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more