Jul 15, 2017 · 14m · a16z

Supernovas and Novel Insight: Where Machine Learning is Headed Next

0:00 / 0:00

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

UC Berkeley professor and Wise.io CTO Josh Bloom explores how machine learning automates massive astronomical image processing and real-time cosmic discovery. He further demonstrates how these scientific methodologies directly inform production-grade enterprise AI applications.

Chapters · click to jump

2 statements ▾

Astronomical Image Subtraction Produces 1,000 False Alarms Per Discovery open →

Machine Learning Tools Have Neglected Time Series Data open →

1 statement ▾

Bloom: Astronomy requirements drove early CCD development at Bell Labs open →

2 statements ▾

Bloom: Machine learning must focus on real-time future data over history open →

Bloom: ML model detected supernova in 11 hours, driving Nature publications open →

1 statement ▾

Bloom: Machine learning software acts like virtualized graduate students at scale open →

2 statements ▾

Josh Bloom: Data-driven approaches do not require prior theory to predict outcomes open →

Josh Bloom: Real-time astronomy ML know-how directly transfers to enterprise problems open →

2 statements ▾

Bloom: Machine learning's destiny is personalized models for everyone open →

Bloom: Future software buyers will purchase products with embedded AI open →

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →

The host as informed peer 0.0 Guest teaching 5.3 Guest disagreement 1.5 The host pushing back 0.0

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

Sharpest disagreement ▶ 5:40 Challenging Superficial ML Adoption

Bloom forcefully rejects using machine learning merely because it is cool or fun, laying down a gauntlet that it must enable capabilities otherwise impossible.

Hardest push from the host ▶ 0:09 No Host Pushback Present

The provided transcript is a continuous monologue by the guest with zero host utterances, so no host pushback occurs.

Biggest teaching moment ▶ 9:00 Reframing Data-Driven Models vs Ex Ante Theories

Bloom educates on how high-dimensional streaming data allows algorithms to generate actionable insights like churn prediction without requiring ex ante theoretical models.

The host holds their own ▶ 0:09 No Host Expertise Demonstration Present

Because the transcript contains no host dialogue, the host has no opportunity to demonstrate expertise or challenge the guest.

the scores for every segment, with the reasoning behind each

Topic	Guest teaching	Guest disagreement	Why
Machine Learning in Physical Sciences and Production Realities	5	1	Josh Bloom delivers an uninterrupted monologue detailing the shift from hiring graduate students to using machine learning classifiers for high-volume astronomical image processing. Because the host does not speak during this segment, host-side metrics remain at zero.
Historical Precedents of Astronomy Driving Technology	5	2	Bloom provides historical context on how astronomy drives hardware and software breakthroughs, citing Galileo's telescope and CCD development at Bell Labs. He lays down a mild challenge against using ML just because it is cool rather than scientifically necessary, while host scores remain zero.
The Synergy of Hardware Capabilities and Virtual Graduate Students	6	2	Bloom connects domain science engineering to enterprise software, demonstrating how handling streaming real-time sky data translates directly into commercial customer churn prediction. The host remains silent throughout, keeping host pushback and expertise at zero.
The Future of Embedded Machine Intelligence	5	1	Bloom explains the evolution of ML from descriptive analytics to embedded, personalized intelligence built into software products. As with previous segments, the complete lack of host dialogue results in zero host expertise and pushback.

Statements from this episode (10)

Assertion Not checkable as stated

Astronomical Image Subtraction Produces 1,000 False Alarms Per Discovery

“It's about a thousand bad candidates or bogus candidates, every one real one”

Josh Bloom Jul 15, 2017 ▶ 2:35

Assertion Not checkable as stated

Machine Learning Tools Have Neglected Time Series Data

“Now you're getting into some interesting, ah, places where machine learning hasn't spent a lot of time, which is on time series data. And what we wound up realizing in our own, sort of, domain specific research is that there weren't a lot of tools for us from …”

Josh Bloom Jul 15, 2017 ▶ 3:43

Assertion Contradicted

Bloom: Astronomy requirements drove early CCD development at Bell Labs

“Doing that directly with charged couple devices was one of the main, sort of, use cases of what drove CCD development, you know, at Bell Labs you know, sort of, 40 years ago.”

Josh Bloom Jul 15, 2017 ▶ 6:07

Insight

Bloom: Machine learning must focus on real-time future data over history

“In some sense that's the greatest imperative and like the gauntlet that I lay down in front of anyone is that you're not doing machine learning because it's cool and it's fun and you can learn something about the data from the past. You're trying to really use…”

Josh Bloom Jul 15, 2017 ▶ 6:43

Assertion Supported

Bloom: ML model detected supernova in 11 hours, driving Nature publications

“So one of the great things is our, ah, machine learning algorithm and framework wound up finding a new supernova that was in a very nearby galaxy. And because it was found about 11 hours after explosion, which were days earlier than had ever been found for tha…”

Josh Bloom Jul 15, 2017 ▶ 6:58

Insight

Bloom: Machine learning software acts like virtualized graduate students at scale

“Because the software exists that can actually sift through and look at that data as if it's, you know, essentially virtualized graduate students with a huge amount of domain knowledge, and do this at scale it allows you to take more and more data.”

Josh Bloom Jul 15, 2017 ▶ 8:33

Insight

Josh Bloom: Data-driven approaches do not require prior theory to predict outcomes

“Well, the point is, and this is sort of the conceit of all data-driven approaches, is that you don't need to have a theory about why something's gonna happen. The idea is that you've got enough data, both in, in terms of the number of examples, and then also s…”

Josh Bloom Jul 15, 2017 ▶ 10:03

Assertion Not checkable as stated

Josh Bloom: Real-time astronomy ML know-how directly transfers to enterprise problems

“The actual insight that we wind up sort of learning how to do with astronomy data in real time on noisy streaming data is exactly that sort of same know-how that we wind up applying to more of these conventional problems.”

Josh Bloom Jul 15, 2017 ▶ 10:56

Prediction Not checkable as stated

Bloom: Machine learning's destiny is personalized models for everyone

“The manifest destiny is everybody has their own machine learning models built upon their own past behavior, perhaps leveraging some of the insights that you wind up getting from the whole system.”

Josh Bloom Jul 15, 2017 ▶ 11:54

Prediction Not checkable as stated

Bloom: Future software buyers will purchase products with embedded AI

“In the future people are just going to be buying products where machine learning and machine intelligence are baked in.”

Josh Bloom Jul 15, 2017 ▶ 14:06

Supernovas and Novel Insight: Where Machine Learning is Headed Next

How this conversation actually went

Statements from this episode (10)

Turn any episode into a week of clips.