Jul 28, 2017 · 20m · a16z

Jason Mars

Jason Mars · 15m spoken Johann Hauswald · 37s 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 the Andreessen Horowitz Academic Roundtable, Professor Jason Mars presents Sirius (Lucida), an open-source intelligent personal assistant designed to solve severe cloud infrastructure scalability challenges. He further demonstrates how Clarity Lab translated these academic AI breakthroughs into enterprise commercial solutions through their spin-out company, Clinc.

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 4.0 Guest disagreement 1.0 The host pushing back 0.0
05100:0010:0020:000:46–3:08 · The host as informed peer 0/10 Academic Track Record and Presentation Overview In this solo presentation segment, Jason Mars introduces his academic credentials at the University of Michigan and gives background stats on intelligent personal assistants. Because there is no host present in the recording, host-side scores are zero.3:08–5:57 · The host as informed peer 0/10 The Sirius Project: Open-Source Intelligent Personal Assistant Mars frames big tech companies as monopolizing AI infrastructure, explaining why his team decided to tackle the hard problem of building open-source Sirius. The guest shows mild pushback against corporate dominance while explaining technical query pipelines.5:57–9:40 · The host as informed peer 0/10 Sirius Technical Stack and Rebranding to Lucida Mars walks through the technical components of Sirius, including GMMs, deep neural networks, SERF algorithms, and OpenEphyra, before playing a live demo. He briefly mentions Apple's request to rebrand the software from Sirius to Lucida.9:40–13:13 · The host as informed peer 0/10 The Data Center Scalability Gap Mars delivers a data-heavy breakdown of the scalability gap in data centers, explaining that moving to voice/vision queries requires a 100x expansion in data center capacity. He outlines how porting compute-intensive ML algorithms to GPUs/FPGAs provides a 10x improvement.13:13–15:28 · The host as informed peer 0/10 Deep Learning Ecosystem and Clarity Lab Inc. Mars discusses deep learning industry trends and introduces Clarity Lab Inc. as a commercial vehicle to translate academic papers into real-world applications. Host scores remain zero as this is a continuous presentation.15:28–17:56 · The host as informed peer 0/10 Stage Transition to Commercial Video A promotional video clip plays showcasing Clarity Lab team members and a partner from Notion explaining how sentiment analysis is integrated into email processing. This is a broadcast video clip with no host interactions.17:56–19:51 · The host as informed peer 0/10 Partner Spotlight: BankJoy Conversational Banking The video continues with a spotlight on BankJoy's conversational banking features before Mars concludes the presentation with details on their intellectual property licensing model. Scores reflect standard technical presentation dynamics.0:46–3:08 · Guest teaching 3/10 Academic Track Record and Presentation Overview In this solo presentation segment, Jason Mars introduces his academic credentials at the University of Michigan and gives background stats on intelligent personal assistants. Because there is no host present in the recording, host-side scores are zero.3:08–5:57 · Guest teaching 4/10 The Sirius Project: Open-Source Intelligent Personal Assistant Mars frames big tech companies as monopolizing AI infrastructure, explaining why his team decided to tackle the hard problem of building open-source Sirius. The guest shows mild pushback against corporate dominance while explaining technical query pipelines.5:57–9:40 · Guest teaching 5/10 Sirius Technical Stack and Rebranding to Lucida Mars walks through the technical components of Sirius, including GMMs, deep neural networks, SERF algorithms, and OpenEphyra, before playing a live demo. He briefly mentions Apple's request to rebrand the software from Sirius to Lucida.9:40–13:13 · Guest teaching 6/10 The Data Center Scalability Gap Mars delivers a data-heavy breakdown of the scalability gap in data centers, explaining that moving to voice/vision queries requires a 100x expansion in data center capacity. He outlines how porting compute-intensive ML algorithms to GPUs/FPGAs provides a 10x improvement.13:13–15:28 · Guest teaching 4/10 Deep Learning Ecosystem and Clarity Lab Inc. Mars discusses deep learning industry trends and introduces Clarity Lab Inc. as a commercial vehicle to translate academic papers into real-world applications. Host scores remain zero as this is a continuous presentation.15:28–17:56 · Guest teaching 3/10 Stage Transition to Commercial Video A promotional video clip plays showcasing Clarity Lab team members and a partner from Notion explaining how sentiment analysis is integrated into email processing. This is a broadcast video clip with no host interactions.17:56–19:51 · Guest teaching 3/10 Partner Spotlight: BankJoy Conversational Banking The video continues with a spotlight on BankJoy's conversational banking features before Mars concludes the presentation with details on their intellectual property licensing model. Scores reflect standard technical presentation dynamics.0:46–3:08 · Guest disagreement 1/10 Academic Track Record and Presentation Overview In this solo presentation segment, Jason Mars introduces his academic credentials at the University of Michigan and gives background stats on intelligent personal assistants. Because there is no host present in the recording, host-side scores are zero.3:08–5:57 · Guest disagreement 2/10 The Sirius Project: Open-Source Intelligent Personal Assistant Mars frames big tech companies as monopolizing AI infrastructure, explaining why his team decided to tackle the hard problem of building open-source Sirius. The guest shows mild pushback against corporate dominance while explaining technical query pipelines.5:57–9:40 · Guest disagreement 1/10 Sirius Technical Stack and Rebranding to Lucida Mars walks through the technical components of Sirius, including GMMs, deep neural networks, SERF algorithms, and OpenEphyra, before playing a live demo. He briefly mentions Apple's request to rebrand the software from Sirius to Lucida.9:40–13:13 · Guest disagreement 2/10 The Data Center Scalability Gap Mars delivers a data-heavy breakdown of the scalability gap in data centers, explaining that moving to voice/vision queries requires a 100x expansion in data center capacity. He outlines how porting compute-intensive ML algorithms to GPUs/FPGAs provides a 10x improvement.13:13–15:28 · Guest disagreement 1/10 Deep Learning Ecosystem and Clarity Lab Inc. Mars discusses deep learning industry trends and introduces Clarity Lab Inc. as a commercial vehicle to translate academic papers into real-world applications. Host scores remain zero as this is a continuous presentation.15:28–17:56 · Guest disagreement 0/10 Stage Transition to Commercial Video A promotional video clip plays showcasing Clarity Lab team members and a partner from Notion explaining how sentiment analysis is integrated into email processing. This is a broadcast video clip with no host interactions.17:56–19:51 · Guest disagreement 0/10 Partner Spotlight: BankJoy Conversational Banking The video continues with a spotlight on BankJoy's conversational banking features before Mars concludes the presentation with details on their intellectual property licensing model. Scores reflect standard technical presentation dynamics.0:46–3:08 · The host pushing back 0/10 Academic Track Record and Presentation Overview In this solo presentation segment, Jason Mars introduces his academic credentials at the University of Michigan and gives background stats on intelligent personal assistants. Because there is no host present in the recording, host-side scores are zero.3:08–5:57 · The host pushing back 0/10 The Sirius Project: Open-Source Intelligent Personal Assistant Mars frames big tech companies as monopolizing AI infrastructure, explaining why his team decided to tackle the hard problem of building open-source Sirius. The guest shows mild pushback against corporate dominance while explaining technical query pipelines.5:57–9:40 · The host pushing back 0/10 Sirius Technical Stack and Rebranding to Lucida Mars walks through the technical components of Sirius, including GMMs, deep neural networks, SERF algorithms, and OpenEphyra, before playing a live demo. He briefly mentions Apple's request to rebrand the software from Sirius to Lucida.9:40–13:13 · The host pushing back 0/10 The Data Center Scalability Gap Mars delivers a data-heavy breakdown of the scalability gap in data centers, explaining that moving to voice/vision queries requires a 100x expansion in data center capacity. He outlines how porting compute-intensive ML algorithms to GPUs/FPGAs provides a 10x improvement.13:13–15:28 · The host pushing back 0/10 Deep Learning Ecosystem and Clarity Lab Inc. Mars discusses deep learning industry trends and introduces Clarity Lab Inc. as a commercial vehicle to translate academic papers into real-world applications. Host scores remain zero as this is a continuous presentation.15:28–17:56 · The host pushing back 0/10 Stage Transition to Commercial Video A promotional video clip plays showcasing Clarity Lab team members and a partner from Notion explaining how sentiment analysis is integrated into email processing. This is a broadcast video clip with no host interactions.17:56–19:51 · The host pushing back 0/10 Partner Spotlight: BankJoy Conversational Banking The video continues with a spotlight on BankJoy's conversational banking features before Mars concludes the presentation with details on their intellectual property licensing model. Scores reflect standard technical presentation dynamics.

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

0:00 · the host 0% · guest 100%0:00 · the host 0% · guest 100%3:00 · the host 0% · guest 100%3:00 · the host 0% · guest 100%6:00 · the host 0% · guest 100%6:00 · the host 0% · guest 100%9:00 · the host 0% · guest 100%9:00 · the host 0% · guest 100%12:00 · the host 0% · guest 100%12:00 · the host 0% · guest 100%15:00 · the host 0% · guest 100%15:00 · the host 0% · guest 100%18:00 · the host 0% · guest 100%18:00 · the host 0% · guest 100%
Sharpest disagreement ▶ 3:25 Rejection of Big Tech Infrastructure Monopoly

Mars criticizes how companies like Apple, Google, and Microsoft monopolize intelligent assistant infrastructure, arguing that academic researchers must overcome this barrier rather than taking the easy route.

Hardest push from the host ▶ 7:25 Response to Apple's Trademark Name Change Request

Mars notes that Apple contacted his lab after Sirius gained traction to request a name change, prompting them to rebrand the open-source platform as Lucida.

Biggest teaching moment ▶ 10:10 Exposing the 100x Data Center Scalability Gap

Mars educates the audience with quantitative analysis showing that replacing text search queries with intelligent personal assistants would require data centers to expand by two orders of magnitude.

The host holds their own ▶ 11:15 Technical Breakdown of ML Acceleration Platforms

Mars demonstrates deep systems expertise by identifying that 92% of execution time is spent on specific ML algorithms and detailing 10x speedups gained by porting workload algorithms to GPUs and FPGAs.

the scores for every segment, with the reasoning behind each
ChapterTopicThe host as informed peerGuest teachingGuest disagreementThe host pushing backWhy
Academic Track Record and Presentation Overview 0310 In this solo presentation segment, Jason Mars introduces his academic credentials at the University of Michigan and gives background stats on intelligent personal assistants. Because there is no host present in the recording, host-side scores are zero.
The Sirius Project: Open-Source Intelligent Personal Assistant 0420 Mars frames big tech companies as monopolizing AI infrastructure, explaining why his team decided to tackle the hard problem of building open-source Sirius. The guest shows mild pushback against corporate dominance while explaining technical query pipelines.
Sirius Technical Stack and Rebranding to Lucida 0510 Mars walks through the technical components of Sirius, including GMMs, deep neural networks, SERF algorithms, and OpenEphyra, before playing a live demo. He briefly mentions Apple's request to rebrand the software from Sirius to Lucida.
The Data Center Scalability Gap 0620 Mars delivers a data-heavy breakdown of the scalability gap in data centers, explaining that moving to voice/vision queries requires a 100x expansion in data center capacity. He outlines how porting compute-intensive ML algorithms to GPUs/FPGAs provides a 10x improvement.
Deep Learning Ecosystem and Clarity Lab Inc. 0410 Mars discusses deep learning industry trends and introduces Clarity Lab Inc. as a commercial vehicle to translate academic papers into real-world applications. Host scores remain zero as this is a continuous presentation.
Stage Transition to Commercial Video 0300 A promotional video clip plays showcasing Clarity Lab team members and a partner from Notion explaining how sentiment analysis is integrated into email processing. This is a broadcast video clip with no host interactions.
Partner Spotlight: BankJoy Conversational Banking 0300 The video continues with a spotlight on BankJoy's conversational banking features before Mars concludes the presentation with details on their intellectual property licensing model. Scores reflect standard technical presentation dynamics.

Statements from this episode (13)

Assertion Supported
Mars: Clarity Lab used candle wax to cool servers longer
“There's a work where we put candle wax in servers to keep them cooler longer.”
Jason Mars Jul 28, 2017 ▶ 1:01
Assertion Supported
Mars: Apple built a $1 billion data center for Siri's launch
“Apple built a one billion dollar data center at the time that Siri was released.”
Jason Mars Jul 28, 2017 ▶ 2:39
Assertion Not checkable as stated
Mars: AI assistant infrastructures are monopolized by major tech companies
“These kinds of infrastructures are monopolized essentially by companies like Apple, Google, Microsoft, and so forth.”
Jason Mars Jul 28, 2017 ▶ 3:14
Disclosure
Mars: Sirius QA system uses the entire Wikipedia corpus for answers
“We have a QA system, a question and answer system that has at its disposal the entire corpus of Wikipedia to answer your question, and it hits that system, and the answer is plucked from the knowledge base of Wikipedia and sent to the user.”
Jason Mars Jul 28, 2017 ▶ 5:31
Assertion Supported
Mars: Speech recognition shifted from Gaussian mixture models to deep neural networks
“Google used to use Gaussian mixture models in the early days, and now everyone has moved to deep neural networks. To do the speech recognition, ah, task.”
Jason Mars Jul 28, 2017 ▶ 6:39
Disclosure
Mars: Apple requested Sirius AI project rename, resulting in Lucida
“After Sirius gained a good bit of traction, Apple reached out to us and said, you probably should change the name. And so we're now calling it Lucida.”
Jason Mars Jul 28, 2017 ▶ 7:41
What-if
Mars: Converting search to AI assistants requires 100x data center scaling
“If you moved all of the queries that were search queries and turn them into intelligent personal assistant queries, we would have to grow our data centers by two orders of magnitude.”
Jason Mars Jul 28, 2017 ▶ 10:19
Assertion Supported
Mars: Compute-intensive ML algorithms drive 92% of Sirius execution workload
“And so across the entire workload, 92% of the execution of series could be described by these compute-intensive algorithms.”
Jason Mars Jul 28, 2017 ▶ 11:23
Assertion Supported
Mars: Porting Sirius to GPUs and FPGAs yields 10x speedup
“We can get significant speedups when we port these algorithms To GPU, when we leverage GPUs and FPGAs to build future servers, and what that means is we can bring that scalability gap down, and so we can get about 10 X back if we move by taking these, ah, by t…”
Jason Mars Jul 28, 2017 ▶ 11:57
Assertion Not publicly verifiable
Mars: Open-source Sirius AI system ranked number one on GitHub
“It was number one on GitHub for a couple weeks.”
Jason Mars Jul 28, 2017 ▶ 12:35
Assertion Supported
Mars: Tech giants are building future data centers around deep learning
“Deep learning is absolutely the bet that companies are making, right? So Google has put a lot of resources into deep learning you know, Facebook. We're designing future data centers and future infrastructures around this one technique, and it's primarily becau…”
Jason Mars Jul 28, 2017 ▶ 13:15
Assertion Supported
Mars: Clarity Lab's custom GPU appliance outperforms off-the-shelf deep learning tools
“Jin is a deep learning engine. It's an appliance. We co-designed a GPU box with a software stack for deep learning and it can, it beats the pants off of out-of-the-box stuff. That you can find.”
Jason Mars Jul 28, 2017 ▶ 15:13
Assertion Not checkable as stated
Mars: Clinc's license covers four years of future university lab research
“We have an unprecedented license agreement where in addition to all the tech we currently have, four years of future tech is also covered in our license. So as we improve it from our research lab, It's exclusively owned by Clarity Lab Inc, the improvements.”
Jason Mars Jul 28, 2017 ▶ 20:09
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 1,000 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.