StarZero All shows Try the podcast clipper →

Nov 20, 2014 · 20m · mad

Matthew Zeiler, Clarifai // Data Driven #31 // Nov 2014 (Hosted by FirstMark Capital)

Matthew Zeiler · 14m spoken Matt Turck · 1m spoken Larry Smith · 21s 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

Matthew Zeiler, founder of Clarifai, presents his company's deep learning visual recognition platform at Data Driven NYC. Through live demonstrations and technical explanations, he showcases real-time image auto-tagging, visual similarity search, enterprise applications, and developer API capabilities.

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 6.3% of the talking time here. How this is scored →

Matt as informed peer 1.3 Guest teaching 1.3 Guest disagreement 0.5 Matt pushing back 1.3
05100:0010:0020:000:52–7:48 · Matt as informed peer 0/10 Live Homepage Demo - Image Auto-Tagging & Similarity Search In this solo presentation segment, the host does not speak at all. Matthew Zeiler demonstrates Clarifai's auto-tagging and image search demo while detailing the underlying deep learning architecture and historical GPU speedups.7:48–12:11 · Matt as informed peer 0/10 Real-World Applications & Stock Photo / Visual Search Demo Zeiler continues his presentation uninterrupted by the host, outlining real-world applications across consumer photos, e-commerce, stock photography, brand safety, satellite imagery, and medical imaging, alongside a live visual similarity search demo.12:11–14:12 · Matt as informed peer 0/10 Company Background, Future Research & Developer API Zeiler delivers the final section of his presentation introducing Clarifai's team background, investors, API performance metrics, and future R&D directions in video, speech, and text recognition without host involvement.14:12–20:20 · Matt as informed peer 5/10 Audience Q&A Session Host Matt Turck opens the Q&A by challenging Zeiler's claim of easy horizontal deployment across verticals like medical imaging, noting human experts are still needed to label cancerous data. Zeiler acknowledges the necessity of labeled data, while answering subsequent technical and competitive questions from the audience in a collaborative manner.0:52–7:48 · Guest teaching 0/10 Live Homepage Demo - Image Auto-Tagging & Similarity Search In this solo presentation segment, the host does not speak at all. Matthew Zeiler demonstrates Clarifai's auto-tagging and image search demo while detailing the underlying deep learning architecture and historical GPU speedups.7:48–12:11 · Guest teaching 0/10 Real-World Applications & Stock Photo / Visual Search Demo Zeiler continues his presentation uninterrupted by the host, outlining real-world applications across consumer photos, e-commerce, stock photography, brand safety, satellite imagery, and medical imaging, alongside a live visual similarity search demo.12:11–14:12 · Guest teaching 0/10 Company Background, Future Research & Developer API Zeiler delivers the final section of his presentation introducing Clarifai's team background, investors, API performance metrics, and future R&D directions in video, speech, and text recognition without host involvement.14:12–20:20 · Guest teaching 5/10 Audience Q&A Session Host Matt Turck opens the Q&A by challenging Zeiler's claim of easy horizontal deployment across verticals like medical imaging, noting human experts are still needed to label cancerous data. Zeiler acknowledges the necessity of labeled data, while answering subsequent technical and competitive questions from the audience in a collaborative manner.0:52–7:48 · Guest disagreement 0/10 Live Homepage Demo - Image Auto-Tagging & Similarity Search In this solo presentation segment, the host does not speak at all. Matthew Zeiler demonstrates Clarifai's auto-tagging and image search demo while detailing the underlying deep learning architecture and historical GPU speedups.7:48–12:11 · Guest disagreement 0/10 Real-World Applications & Stock Photo / Visual Search Demo Zeiler continues his presentation uninterrupted by the host, outlining real-world applications across consumer photos, e-commerce, stock photography, brand safety, satellite imagery, and medical imaging, alongside a live visual similarity search demo.12:11–14:12 · Guest disagreement 0/10 Company Background, Future Research & Developer API Zeiler delivers the final section of his presentation introducing Clarifai's team background, investors, API performance metrics, and future R&D directions in video, speech, and text recognition without host involvement.14:12–20:20 · Guest disagreement 2/10 Audience Q&A Session Host Matt Turck opens the Q&A by challenging Zeiler's claim of easy horizontal deployment across verticals like medical imaging, noting human experts are still needed to label cancerous data. Zeiler acknowledges the necessity of labeled data, while answering subsequent technical and competitive questions from the audience in a collaborative manner.0:52–7:48 · Matt pushing back 0/10 Live Homepage Demo - Image Auto-Tagging & Similarity Search In this solo presentation segment, the host does not speak at all. Matthew Zeiler demonstrates Clarifai's auto-tagging and image search demo while detailing the underlying deep learning architecture and historical GPU speedups.7:48–12:11 · Matt pushing back 0/10 Real-World Applications & Stock Photo / Visual Search Demo Zeiler continues his presentation uninterrupted by the host, outlining real-world applications across consumer photos, e-commerce, stock photography, brand safety, satellite imagery, and medical imaging, alongside a live visual similarity search demo.12:11–14:12 · Matt pushing back 0/10 Company Background, Future Research & Developer API Zeiler delivers the final section of his presentation introducing Clarifai's team background, investors, API performance metrics, and future R&D directions in video, speech, and text recognition without host involvement.14:12–20:20 · Matt pushing back 5/10 Audience Q&A Session Host Matt Turck opens the Q&A by challenging Zeiler's claim of easy horizontal deployment across verticals like medical imaging, noting human experts are still needed to label cancerous data. Zeiler acknowledges the necessity of labeled data, while answering subsequent technical and competitive questions from the audience in a collaborative manner.

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

0:00 · Matt 0% · guest 100%0:00 · Matt 0% · guest 100%3:00 · Matt 0% · guest 100%3:00 · Matt 0% · guest 100%6:00 · Matt 0% · guest 100%6:00 · Matt 0% · guest 100%9:00 · Matt 0% · guest 100%9:00 · Matt 0% · guest 100%12:00 · Matt 22.5% · guest 77.5%12:00 · Matt 22.5% · guest 77.5%15:00 · Matt 16.1% · guest 83.9%15:00 · Matt 16.1% · guest 83.9%18:00 · Matt 8% · guest 92%18:00 · Matt 8% · guest 92%
Sharpest disagreement ▶ 19:07 Rejection of Google as a direct competitor

Zeiler politely reframes an audience question about tech titan competition, contending that Google focuses on its own internal products and users rather than directly competing with Clarifai's B2B visual recognition API.

 Hardest push from Matt ▶ 15:36 Pressing on domain expert training requirements

Turck refuses to fully accept Zeiler's assertion that neural networks trivially adapt to complex domain verticals like healthcare without specialized domain input, pressing him on who identifies cancerous versus non-cancerous tumors.

 Biggest teaching moment ▶ 14:58 Explaining agnostic paired input-output mappings in neural networks

Zeiler educates the host on how deep learning treats arbitrary domain data identically by converting paired input pixels to target outputs regardless of the underlying vertical.

 Matt holds his own ▶ 14:18 Framing horizontal platforms versus vertical expertise

Turck demonstrates industry foresight by framing his question around a key macroeconomic software debate: whether general horizontal AI platforms can easily disrupt specialized domain verticals.
the scores for every segment, with the reasoning behind each
ChapterTopicMatt as informed peerGuest teachingGuest disagreementMatt pushing backWhy
Live Homepage Demo - Image Auto-Tagging & Similarity Search 0000 In this solo presentation segment, the host does not speak at all. Matthew Zeiler demonstrates Clarifai's auto-tagging and image search demo while detailing the underlying deep learning architecture and historical GPU speedups.
Real-World Applications & Stock Photo / Visual Search Demo 0000 Zeiler continues his presentation uninterrupted by the host, outlining real-world applications across consumer photos, e-commerce, stock photography, brand safety, satellite imagery, and medical imaging, alongside a live visual similarity search demo.
Company Background, Future Research & Developer API 0000 Zeiler delivers the final section of his presentation introducing Clarifai's team background, investors, API performance metrics, and future R&D directions in video, speech, and text recognition without host involvement.
Audience Q&A Session 5525 Host Matt Turck opens the Q&A by challenging Zeiler's claim of easy horizontal deployment across verticals like medical imaging, noting human experts are still needed to label cancerous data. Zeiler acknowledges the necessity of labeled data, while answering subsequent technical and competitive questions from the audience in a collaborative manner.

Statements from this episode (9)

Assertion Supported
Siri and Android voice recognition already rely entirely on neural networks
“Speech recognition, basically all the Siri processing or Android voice recognition is done with neural networks these days.”
Matthew Zeiler Nov 20, 2014 ▶ 5:07
Opinion
Competitiveness in computer vision now strictly requires using neural networks
“And you can see now, to even be competitive, you have to use neural networks.”
Matthew Zeiler Nov 20, 2014 ▶ 6:48
Assertion Supported
Clarifai won the seminal 2013 ImageNet computer vision competition
“Clarify happened to be this result at the bottom. We won the 2013 competition.”
Matthew Zeiler Nov 20, 2014 ▶ 6:53
Assertion Not checkable as stated
Stock photography platforms still rely entirely on manual labeling in 2014
“Same story with stock photography. They are all manually labeled at this point.”
Matthew Zeiler Nov 20, 2014 ▶ 8:19
Assertion Not checkable as stated
Clarifai automatically tagged 1.3 million stock images in a few minutes
“And again, this is done automatically on, this is 1.3 million images, it takes a matter of minutes to do this, so you can scale this up to billions of images, no problem.”
Matthew Zeiler Nov 20, 2014 ▶ 10:46
Disclosure
Google, Qualcomm, and Nvidia are strategic investors in Clarifai
“We got some great investors like Google, Qualcomm, and NVIDIA who help power this technology for us.”
Matthew Zeiler Nov 20, 2014 ▶ 12:23
Assertion Not publicly verifiable
Clarifai's video recognition runs ten times faster than real time
“It can run about 10 times faster than real time,”
Matthew Zeiler Nov 20, 2014 ▶ 13:08
Insight
Unsupervised neural networks will never match models trained on well-labeled data
“These models work really well with labeled data. There are approaches where you can train without any labels but the performance is never as good as if you have well-labeled data.”
Matthew Zeiler Nov 20, 2014 ▶ 15:48
Opinion
Clarifai does not view Google as a direct competitor
“So Google, we don't consider them a competitor. They have great research teams and really big research teams and great resources but they're working on their own problems. They have their own users, and they don't they don't really compete with us in terms of …”
Matthew Zeiler Nov 20, 2014 ▶ 19:07
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