StarZero All shows Try the podcast clipper →

Sep 28, 2017 · 34m · mad

AI Startup Predictions // Bradford Cross - A fireside chat with Matt Turck (FirstMark's Data Driven)

Bradford Cross · 28m spoken Matt Turck · 2m spoken
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

At a Data Driven NYC fireside chat hosted by Matt Turck, investor and Merlon Intelligence CEO Bradford Cross critiques common AI industry hype while outlining actionable strategies for building defensible, full-stack vertical AI startups.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 9.5% of the talking time here. How this is scored →

Matt as informed peer 3.1 Guest teaching 4.3 Guest disagreement 2.4 Matt pushing back 1.3
05100:0010:0020:0030:000:09–4:20 · Matt as informed peer 1/10 Bradford Cross Background and Data Collective Overview Matt opens with standard background prompts, allowing Bradford to detail his career across hedge funds, Data Collective, and Merlon Intelligence. Matt plays a purely conversational host role without challenging or adding technical detail.4:20–9:33 · Matt as informed peer 1/10 Prediction 1: Why Chatbots Fail ('Bots Go Bust') Bradford delivers a strong contrarian opinion on the failure of chatbots and Silicon Valley consumer tech detachment. Matt offers minimal intervention while Bradford dismisses mainstream hype around conversational interfaces.9:33–13:39 · Matt as informed peer 7/10 Prediction 2: Deep Learning Commoditization Matt actively shapes the discussion by asking Bradford to clarify that machine learning as a whole is not commoditized, then contributes informed industry context regarding TensorFlow open-sourcing and VC signal confusion.13:39–19:06 · Matt as informed peer 3/10 Prediction 3: AI as CleanTech 2.0 for Venture Capital Bradford rails against venture capital trends and cloud providers pushing machine learning APIs. Matt introduces the topic from Bradford's writings but steps back as Bradford aggressively critiques big tech strategy.19:06–24:08 · Matt as informed peer 4/10 Building Full-Stack Vertical AI Applications Matt prompts Bradford on full-stack vertical AI integration, demonstrating familiarity with architectural paradigms. Bradford outlines how taking state-of-the-art internet search ranking patterns to banking compliance creates high application value.24:08–26:45 · Matt as informed peer 5/10 Defensibility and Data Compounding Effects Matt brings up industry terminology on network effects versus compounding data effects. Bradford agrees with Matt's framing and explains how feedback loops reinforce proprietary advantage.26:45–31:03 · Matt as informed peer 4/10 Ideal Team Structure for Vertical AI Startups Bradford caricatures extreme failure modes in startup founding teams. Matt shows domain knowledge of Bradford's company by citing specific executive hires on the regulatory side.31:03–34:25 · Matt as informed peer 0/10 Audience Q&A: Federated Learning and Regulated Data An audience member asks about regulated banking data sharing. Bradford educates the room on parameter obfuscation and federated learning instead of raw data pooling, while host acts as facilitator.0:09–4:20 · Guest teaching 3/10 Bradford Cross Background and Data Collective Overview Matt opens with standard background prompts, allowing Bradford to detail his career across hedge funds, Data Collective, and Merlon Intelligence. Matt plays a purely conversational host role without challenging or adding technical detail.4:20–9:33 · Guest teaching 5/10 Prediction 1: Why Chatbots Fail ('Bots Go Bust') Bradford delivers a strong contrarian opinion on the failure of chatbots and Silicon Valley consumer tech detachment. Matt offers minimal intervention while Bradford dismisses mainstream hype around conversational interfaces.9:33–13:39 · Guest teaching 4/10 Prediction 2: Deep Learning Commoditization Matt actively shapes the discussion by asking Bradford to clarify that machine learning as a whole is not commoditized, then contributes informed industry context regarding TensorFlow open-sourcing and VC signal confusion.13:39–19:06 · Guest teaching 4/10 Prediction 3: AI as CleanTech 2.0 for Venture Capital Bradford rails against venture capital trends and cloud providers pushing machine learning APIs. Matt introduces the topic from Bradford's writings but steps back as Bradford aggressively critiques big tech strategy.19:06–24:08 · Guest teaching 5/10 Building Full-Stack Vertical AI Applications Matt prompts Bradford on full-stack vertical AI integration, demonstrating familiarity with architectural paradigms. Bradford outlines how taking state-of-the-art internet search ranking patterns to banking compliance creates high application value.24:08–26:45 · Guest teaching 4/10 Defensibility and Data Compounding Effects Matt brings up industry terminology on network effects versus compounding data effects. Bradford agrees with Matt's framing and explains how feedback loops reinforce proprietary advantage.26:45–31:03 · Guest teaching 4/10 Ideal Team Structure for Vertical AI Startups Bradford caricatures extreme failure modes in startup founding teams. Matt shows domain knowledge of Bradford's company by citing specific executive hires on the regulatory side.31:03–34:25 · Guest teaching 5/10 Audience Q&A: Federated Learning and Regulated Data An audience member asks about regulated banking data sharing. Bradford educates the room on parameter obfuscation and federated learning instead of raw data pooling, while host acts as facilitator.0:09–4:20 · Guest disagreement 1/10 Bradford Cross Background and Data Collective Overview Matt opens with standard background prompts, allowing Bradford to detail his career across hedge funds, Data Collective, and Merlon Intelligence. Matt plays a purely conversational host role without challenging or adding technical detail.4:20–9:33 · Guest disagreement 5/10 Prediction 1: Why Chatbots Fail ('Bots Go Bust') Bradford delivers a strong contrarian opinion on the failure of chatbots and Silicon Valley consumer tech detachment. Matt offers minimal intervention while Bradford dismisses mainstream hype around conversational interfaces.9:33–13:39 · Guest disagreement 2/10 Prediction 2: Deep Learning Commoditization Matt actively shapes the discussion by asking Bradford to clarify that machine learning as a whole is not commoditized, then contributes informed industry context regarding TensorFlow open-sourcing and VC signal confusion.13:39–19:06 · Guest disagreement 5/10 Prediction 3: AI as CleanTech 2.0 for Venture Capital Bradford rails against venture capital trends and cloud providers pushing machine learning APIs. Matt introduces the topic from Bradford's writings but steps back as Bradford aggressively critiques big tech strategy.19:06–24:08 · Guest disagreement 1/10 Building Full-Stack Vertical AI Applications Matt prompts Bradford on full-stack vertical AI integration, demonstrating familiarity with architectural paradigms. Bradford outlines how taking state-of-the-art internet search ranking patterns to banking compliance creates high application value.24:08–26:45 · Guest disagreement 1/10 Defensibility and Data Compounding Effects Matt brings up industry terminology on network effects versus compounding data effects. Bradford agrees with Matt's framing and explains how feedback loops reinforce proprietary advantage.26:45–31:03 · Guest disagreement 3/10 Ideal Team Structure for Vertical AI Startups Bradford caricatures extreme failure modes in startup founding teams. Matt shows domain knowledge of Bradford's company by citing specific executive hires on the regulatory side.31:03–34:25 · Guest disagreement 1/10 Audience Q&A: Federated Learning and Regulated Data An audience member asks about regulated banking data sharing. Bradford educates the room on parameter obfuscation and federated learning instead of raw data pooling, while host acts as facilitator.0:09–4:20 · Matt pushing back 0/10 Bradford Cross Background and Data Collective Overview Matt opens with standard background prompts, allowing Bradford to detail his career across hedge funds, Data Collective, and Merlon Intelligence. Matt plays a purely conversational host role without challenging or adding technical detail.4:20–9:33 · Matt pushing back 1/10 Prediction 1: Why Chatbots Fail ('Bots Go Bust') Bradford delivers a strong contrarian opinion on the failure of chatbots and Silicon Valley consumer tech detachment. Matt offers minimal intervention while Bradford dismisses mainstream hype around conversational interfaces.9:33–13:39 · Matt pushing back 5/10 Prediction 2: Deep Learning Commoditization Matt actively shapes the discussion by asking Bradford to clarify that machine learning as a whole is not commoditized, then contributes informed industry context regarding TensorFlow open-sourcing and VC signal confusion.13:39–19:06 · Matt pushing back 1/10 Prediction 3: AI as CleanTech 2.0 for Venture Capital Bradford rails against venture capital trends and cloud providers pushing machine learning APIs. Matt introduces the topic from Bradford's writings but steps back as Bradford aggressively critiques big tech strategy.19:06–24:08 · Matt pushing back 1/10 Building Full-Stack Vertical AI Applications Matt prompts Bradford on full-stack vertical AI integration, demonstrating familiarity with architectural paradigms. Bradford outlines how taking state-of-the-art internet search ranking patterns to banking compliance creates high application value.24:08–26:45 · Matt pushing back 1/10 Defensibility and Data Compounding Effects Matt brings up industry terminology on network effects versus compounding data effects. Bradford agrees with Matt's framing and explains how feedback loops reinforce proprietary advantage.26:45–31:03 · Matt pushing back 1/10 Ideal Team Structure for Vertical AI Startups Bradford caricatures extreme failure modes in startup founding teams. Matt shows domain knowledge of Bradford's company by citing specific executive hires on the regulatory side.31:03–34:25 · Matt pushing back 0/10 Audience Q&A: Federated Learning and Regulated Data An audience member asks about regulated banking data sharing. Bradford educates the room on parameter obfuscation and federated learning instead of raw data pooling, while host acts as facilitator.

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

0:00 · Matt 11.1% · guest 88.9%0:00 · Matt 11.1% · guest 88.9%3:00 · Matt 22.2% · guest 77.8%3:00 · Matt 22.2% · guest 77.8%6:00 · Matt 0.8% · guest 99.2%6:00 · Matt 0.8% · guest 99.2%9:00 · Matt 8.3% · guest 91.7%9:00 · Matt 8.3% · guest 91.7%12:00 · Matt 18.2% · guest 81.8%12:00 · Matt 18.2% · guest 81.8%15:00 · Matt 17.3% · guest 82.7%15:00 · Matt 17.3% · guest 82.7%18:00 · Matt 0.5% · guest 99.5%18:00 · Matt 0.5% · guest 99.5%21:00 · Matt 4% · guest 96%21:00 · Matt 4% · guest 96%24:00 · Matt 15.2% · guest 84.8%24:00 · Matt 15.2% · guest 84.8%27:00 · Matt 3% · guest 97%27:00 · Matt 3% · guest 97%30:00 · Matt 4.6% · guest 95.4%30:00 · Matt 4.6% · guest 95.4%33:00 · Matt 8.7% · guest 91.3%33:00 · Matt 8.7% · guest 91.3%
Sharpest disagreement ▶ 17:35 Bradford mocks big tech cloud ML APIs

Bradford forcefully dismisses cloud APIs from IBM, Google, Amazon, and Microsoft, claiming they are burning hundreds of millions of dollars copying AWS and failing.

 Hardest push from Matt ▶ 11:50 Matt interrupts to clarify commoditization scope

Matt directly interrupts Bradford's narrative to insist on a precise distinction between deep learning commoditization and broader machine learning commoditization.

 Biggest teaching moment ▶ 32:45 Bradford reframes data sharing via parameterization

Bradford reframes an audience question by explaining why raw data lakes fail in banking, detailing how parameter export in federated learning avoids PII regulatory liabilities.

 Matt holds his own ▶ 12:50 Matt analyzes market distortion caused by TensorFlow

Matt displays clear expertise by detailing how TensorFlow open-sourcing led non-technical MBAs to overestimate early AI development speed.
the scores for every segment, with the reasoning behind each
ChapterTopicMatt as informed peerGuest teachingGuest disagreementMatt pushing backWhy
Bradford Cross Background and Data Collective Overview 1310 Matt opens with standard background prompts, allowing Bradford to detail his career across hedge funds, Data Collective, and Merlon Intelligence. Matt plays a purely conversational host role without challenging or adding technical detail.
Prediction 1: Why Chatbots Fail ('Bots Go Bust') 1551 Bradford delivers a strong contrarian opinion on the failure of chatbots and Silicon Valley consumer tech detachment. Matt offers minimal intervention while Bradford dismisses mainstream hype around conversational interfaces.
Prediction 2: Deep Learning Commoditization 7425 Matt actively shapes the discussion by asking Bradford to clarify that machine learning as a whole is not commoditized, then contributes informed industry context regarding TensorFlow open-sourcing and VC signal confusion.
Prediction 3: AI as CleanTech 2.0 for Venture Capital 3451 Bradford rails against venture capital trends and cloud providers pushing machine learning APIs. Matt introduces the topic from Bradford's writings but steps back as Bradford aggressively critiques big tech strategy.
Building Full-Stack Vertical AI Applications 4511 Matt prompts Bradford on full-stack vertical AI integration, demonstrating familiarity with architectural paradigms. Bradford outlines how taking state-of-the-art internet search ranking patterns to banking compliance creates high application value.
Defensibility and Data Compounding Effects 5411 Matt brings up industry terminology on network effects versus compounding data effects. Bradford agrees with Matt's framing and explains how feedback loops reinforce proprietary advantage.
Ideal Team Structure for Vertical AI Startups 4431 Bradford caricatures extreme failure modes in startup founding teams. Matt shows domain knowledge of Bradford's company by citing specific executive hires on the regulatory side.
Audience Q&A: Federated Learning and Regulated Data 0510 An audience member asks about regulated banking data sharing. Bradford educates the room on parameter obfuscation and federated learning instead of raw data pooling, while host acts as facilitator.

Statements from this episode (11)

Assertion Partly supported
JPMorgan Chase and Citi each spend $2B to $4B annually on compliance
“The big ones, like JPMorgan Chase, Citi, et cetera, have 10,000 plus analysts working on this every year. They're spending two to four billion dollars a year on compliance.”
Bradford Cross Sep 28, 2017 ▶ 3:10
Prediction Not checkable as stated
Silicon Valley will see few consumer tech wins before a major correction
“I don't expect to see very many consumer success stories coming from Silicon Valley right now, until we have a major correction.”
Bradford Cross Sep 28, 2017 ▶ 8:27
Prediction Not checkable as stated
Standalone chatbots will eventually be absorbed into broader software platforms
“What'll happen is they end up getting folded into systems that really meet the more basic human needs.”
Bradford Cross Sep 28, 2017 ▶ 9:08
Opinion
Twitter's Magic Pony acquisition marked the peak of AI talent premiums
“But I think that's, you know, that was the sort of tailing off of the, for me, the Twitter Magic Pony deal was kind of the, ah, the tail end of this absurd, extraordinary premium on, on deep learning, and then now, more and more, it's becoming just part of the…”
Bradford Cross Sep 28, 2017 ▶ 11:32
Prediction Not checkable as stated
Most companies cannot adopt machine learning due to limited talent pools
“I think actually most companies will not be able to use, ah, machine learning. Even though deep learning may be commodity within the machine learning community is still way too small to be commodity overall.”
Bradford Cross Sep 28, 2017 ▶ 12:35
Prediction Not checkable as stated
Investors funding AI startups without scientific competency will lose money
“So I think as, insofar as that continues to happen, Those people are gonna lose a lot of money.”
Bradford Cross Sep 28, 2017 ▶ 15:01
Insight
Capable engineering teams prefer open-source tools over buying machine learning APIs
“If a team knows what they're doing, then they tend to use open source and cobble it together themselves.”
Bradford Cross Sep 28, 2017 ▶ 16:16
Prediction Not checkable as stated
Big tech companies will lose massive amounts of money on ML APIs
“I do not think that these numbers are going to be good for these machine learning APIs. And I think those bigger companies are going to lose a massive amount of money over the next several years.”
Bradford Cross Sep 28, 2017 ▶ 17:44
Insight
Startups building low-level tagging APIs lack defensibility and ability to scale
“If you're doing a low-level tagging API, I really, really worry about your defensibility and the ability to scale that business.”
Bradford Cross Sep 28, 2017 ▶ 18:36
Assertion Not checkable as stated
Top machine learning talent is heavily concentrated at Google and Facebook
“Almost all of the really smart machine learning people are in less than 10 companies in the world. They're actually in less than five, right? In fact, they're mostly all at Google and Facebook.”
Bradford Cross Sep 28, 2017 ▶ 19:21
Insight
AI startups targeting traditional industries must build full-stack applications for adoption
“So you've got to kind of come all the way up to spoon feeding them the solution, or else I worry a lot about getting adoption for machine learning applications.”
Bradford Cross Sep 28, 2017 ▶ 23:57
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 400 conversations 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