StarZero All shows Try the podcast clipper →

Apr 6, 2017 · 24m · mad

Rethinking Predictive Analytics // Yaniv Altshuler, Endor (FirstMark's Data Driven)

Yaniv Altshuler · 18m 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 Data Driven NYC, Yaniv Altshuler introduces Endor's Social Physics platform, demonstrating how mathematical behavioral invariants can automate predictive analytics without manual feature engineering or data cleaning. He illustrates how this MIT spin-off technology outperforms traditional machine learning models across real-world enterprise deployments and competitive benchmarks.

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 2.9% of the talking time here. How this is scored →

Matt as informed peer 0.5 Guest teaching 0.5 Guest disagreement 0.5 Matt pushing back 0.5
05100:0010:0020:000:44–7:06 · Matt as informed peer 0/10 The Fundamental Challenge in Machine Learning In this opening presentation segment, Yaniv gives a solo monologue breaking down the limitations of standard machine learning models when applied to human behavior. The host is silent throughout the presentation, so all host scores are 0.7:06–11:52 · Matt as informed peer 0/10 Endor's Solution: Social Physics and Technological Edge Yaniv continues his monologue, presenting Endor's solution rooted in social physics mathematical invariances developed at MIT. The host does not participate or interject in this presentation portion.11:52–15:51 · Matt as informed peer 0/10 Real-World Case Studies and Kaggle Benchmark Yaniv presents concrete case studies including credit card predictive modeling and a Kaggle challenge victory. Host involvement remains non-existent during the talk.15:51–24:18 · Matt as informed peer 2/10 Conclusion of Presentation Matt Turck opens the Q&A session by asking basic foundational questions about why human data is unique and how many rules govern social physics. Yaniv corrects Matt's assumption that there are millions of rules, clarifying there are only a few, before taking questions from the audience.0:44–7:06 · Guest teaching 0/10 The Fundamental Challenge in Machine Learning In this opening presentation segment, Yaniv gives a solo monologue breaking down the limitations of standard machine learning models when applied to human behavior. The host is silent throughout the presentation, so all host scores are 0.7:06–11:52 · Guest teaching 0/10 Endor's Solution: Social Physics and Technological Edge Yaniv continues his monologue, presenting Endor's solution rooted in social physics mathematical invariances developed at MIT. The host does not participate or interject in this presentation portion.11:52–15:51 · Guest teaching 0/10 Real-World Case Studies and Kaggle Benchmark Yaniv presents concrete case studies including credit card predictive modeling and a Kaggle challenge victory. Host involvement remains non-existent during the talk.15:51–24:18 · Guest teaching 2/10 Conclusion of Presentation Matt Turck opens the Q&A session by asking basic foundational questions about why human data is unique and how many rules govern social physics. Yaniv corrects Matt's assumption that there are millions of rules, clarifying there are only a few, before taking questions from the audience.0:44–7:06 · Guest disagreement 1/10 The Fundamental Challenge in Machine Learning In this opening presentation segment, Yaniv gives a solo monologue breaking down the limitations of standard machine learning models when applied to human behavior. The host is silent throughout the presentation, so all host scores are 0.7:06–11:52 · Guest disagreement 0/10 Endor's Solution: Social Physics and Technological Edge Yaniv continues his monologue, presenting Endor's solution rooted in social physics mathematical invariances developed at MIT. The host does not participate or interject in this presentation portion.11:52–15:51 · Guest disagreement 0/10 Real-World Case Studies and Kaggle Benchmark Yaniv presents concrete case studies including credit card predictive modeling and a Kaggle challenge victory. Host involvement remains non-existent during the talk.15:51–24:18 · Guest disagreement 1/10 Conclusion of Presentation Matt Turck opens the Q&A session by asking basic foundational questions about why human data is unique and how many rules govern social physics. Yaniv corrects Matt's assumption that there are millions of rules, clarifying there are only a few, before taking questions from the audience.0:44–7:06 · Matt pushing back 0/10 The Fundamental Challenge in Machine Learning In this opening presentation segment, Yaniv gives a solo monologue breaking down the limitations of standard machine learning models when applied to human behavior. The host is silent throughout the presentation, so all host scores are 0.7:06–11:52 · Matt pushing back 0/10 Endor's Solution: Social Physics and Technological Edge Yaniv continues his monologue, presenting Endor's solution rooted in social physics mathematical invariances developed at MIT. The host does not participate or interject in this presentation portion.11:52–15:51 · Matt pushing back 0/10 Real-World Case Studies and Kaggle Benchmark Yaniv presents concrete case studies including credit card predictive modeling and a Kaggle challenge victory. Host involvement remains non-existent during the talk.15:51–24:18 · Matt pushing back 2/10 Conclusion of Presentation Matt Turck opens the Q&A session by asking basic foundational questions about why human data is unique and how many rules govern social physics. Yaniv corrects Matt's assumption that there are millions of rules, clarifying there are only a few, before taking questions from the audience.

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 0% · guest 100%12:00 · Matt 0% · guest 100%15:00 · Matt 19.1% · guest 80.9%15:00 · Matt 19.1% · guest 80.9%18:00 · Matt 0% · guest 100%18:00 · Matt 0% · guest 100%21:00 · Matt 3.6% · guest 96.4%21:00 · Matt 3.6% · guest 96.4%24:00 · Matt 24.2% · guest 75.8%24:00 · Matt 24.2% · guest 75.8%
Sharpest disagreement ▶ 22:18 Yaniv challenges audience concern about corporate risk

When an audience member raises concerns about legacy corporate blame culture, Yaniv bluntly interrupts asking 'what's the alternative?' before explaining why internal data science teams aren't fully replaced.

 Hardest push from Matt ▶ 16:40 Matt questions the premise of human activity uniqueness

Matt opens the Q&A by probing the foundational premise of social physics, asking why patterns exist in human behavior that are absent in non-human data.

 Biggest teaching moment ▶ 17:32 Yaniv corrects host on the rule set complexity

Matt asks if social physics relies on thousands or millions of underlying rules, and Yaniv immediately corrects him by revealing there are only 'a few', surprising the host.

 Matt holds his own ▶ 16:40 Matt targets the foundational premise of the platform

Matt demonstrates sharp host intuition by immediately zeroing in on the theoretical distinction between human behavior datasets and general data streams.
the scores for every segment, with the reasoning behind each
ChapterTopicMatt as informed peerGuest teachingGuest disagreementMatt pushing backWhy
The Fundamental Challenge in Machine Learning 0010 In this opening presentation segment, Yaniv gives a solo monologue breaking down the limitations of standard machine learning models when applied to human behavior. The host is silent throughout the presentation, so all host scores are 0.
Endor's Solution: Social Physics and Technological Edge 0000 Yaniv continues his monologue, presenting Endor's solution rooted in social physics mathematical invariances developed at MIT. The host does not participate or interject in this presentation portion.
Real-World Case Studies and Kaggle Benchmark 0000 Yaniv presents concrete case studies including credit card predictive modeling and a Kaggle challenge victory. Host involvement remains non-existent during the talk.
Conclusion of Presentation 2212 Matt Turck opens the Q&A session by asking basic foundational questions about why human data is unique and how many rules govern social physics. Yaniv corrects Matt's assumption that there are millions of rules, clarifying there are only a few, before taking questions from the audience.

Statements from this episode (9)

Assertion Partly supported
Altshuler: Modern machine learning technologies were developed roughly 50 years ago
“All the technologies that we have today were developed around 50 years ago, plus minus a decade. Even deep learning that is so trendy today, it's basically very large neural networks. Concepts from the fifties matured around the seventies.”
Yaniv Altshuler Apr 6, 2017 ▶ 2:19
Insight
Altshuler: Deep learning is inadequate for predicting dynamic human behavior
“Deep learning requires a lot of data, relatively speaking, and therefore, by definition, cannot detect recent changes in the data, emerging patterns, ah, and therefore, it's, ah, really, really, ah, inadequate to predict human behavior that is, ah, actually do…”
Yaniv Altshuler Apr 6, 2017 ▶ 6:12
Assertion Not checkable as stated
Altshuler: Mathematical invariances guaranteed in all structured human data
“There is a set of mathematical invariances that Are guaranteed to happen in any structured human data.”
Yaniv Altshuler Apr 6, 2017 ▶ 7:21
Assertion Not checkable as stated
Altshuler: Endor can run predictive analytics on one week of data
“In addition, because we use social physics as a prior to the data, we can analyze data from very short periods of time. We don't need to analyze five years worth of data. In some cases, even one week is enough”
Yaniv Altshuler Apr 6, 2017 ▶ 10:47
Assertion Not checkable as stated
Endor tripled loan sales for a credit card company client
“The result is tripling the amount of loans they actually sell. Three times more loans actually sell in production with our prediction.”
Yaniv Altshuler Apr 6, 2017 ▶ 12:39
Assertion Contradicted
Endor outperformed nearly 1,000 teams on a Kaggle benchmark
“Almost a thousand teams participated. And they were given three months, and you can see that the team who won the first place after us, they submitted almost 200 models, so they actually worked pretty hard, and as you can see here, our accuracy is significantl…”
Yaniv Altshuler Apr 6, 2017 ▶ 15:08
Assertion Not checkable as stated
Altshuler: Domain and ML expertise can be replaced with computation
“Can we replace domain expertise and machine learning expertise with computation? Our answer, yes we can, and it's not even so computationally intense, but it would require us to take a very, very different approach about data.”
Yaniv Altshuler Apr 6, 2017 ▶ 15:55
Opinion
Altshuler: Internal data science teams cannot be replaced by automated platforms
“I actually don't think this is the case because I think that nothing and no one can replace internal teams of data scientists and of analytics people because they are the only ones who are familiar with data and their business, and therefore they are the only …”
Yaniv Altshuler Apr 6, 2017 ▶ 22:55
Assertion Not checkable as stated
Altshuler: Endor can build predictive models with 20 to 30 positive labels
“We can also work with very, very, ah, smallest, so, even 20, 30 positive, ah, Ah, samples, a sample of, ah, 20, 30 positive labels could be enough.”
Yaniv Altshuler Apr 6, 2017 ▶ 23:53
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