Sep 12, 2024 · 26m · no-priors

No Priors Ep. 81 | With Sarah Guo & Elad Gil

Elad Gil · 13m spoken Sarah Guo · 11m 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

In this episode of No Priors, venture capitalists Sarah Guo and Elad Gil break down the shifting macro dynamics of the AI ecosystem, analyzing foundation model consolidation, rapid multimodal breakthroughs, legal and regulatory scraping risks, and semiconductor infrastructure strategies.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 100% of the talking time here. How this is scored →

The hosts as informed peer 6.2 Guest teaching 0.0 Guest disagreement 0.2 The hosts pushing back 0.4
05100:0010:0020:000:23–7:13 · The hosts as informed peer 7/10 LLM Market Consolidation and Capital Dynamics Sarah and Elad examine foundation model consolidation, capital requirements, and 200x token pricing drops. Sarah offers a slight nuance, arguing the market is simultaneously more competitive on performance benchmarks.7:14–14:21 · The hosts as informed peer 7/10 Domain-Specific Models and Multimodal Advancements The co-hosts explore specialized vs. multimodal architectures, drawing analogies to brain modularity and analyzing how rapidly smaller teams caught up to Sora in video generation.14:21–21:42 · The hosts as informed peer 8/10 Managing Risk, Scraping, and Moderation in AI Elad provides an in-depth breakdown of Google's historical copyright battles, fair use snippets, and robots.txt as a framework for current AI scraping and liability risks.21:43–25:37 · The hosts as informed peer 8/10 Semiconductor Innovation and AMD's System Strategy Sarah details AMD's acquisition strategy, breaking down the specific roles of Silo AI for CUDA software alternatives and ZT Systems for data center rack-scale architecture.25:37–26:24 · The hosts as informed peer 1/10 Robotics PSA and Episode Conclusion Brief outro segment featuring Elad's lighthearted PSA requesting leads on purchasing a humanoid robot.0:23–7:13 · Guest teaching 0/10 LLM Market Consolidation and Capital Dynamics Sarah and Elad examine foundation model consolidation, capital requirements, and 200x token pricing drops. Sarah offers a slight nuance, arguing the market is simultaneously more competitive on performance benchmarks.7:14–14:21 · Guest teaching 0/10 Domain-Specific Models and Multimodal Advancements The co-hosts explore specialized vs. multimodal architectures, drawing analogies to brain modularity and analyzing how rapidly smaller teams caught up to Sora in video generation.14:21–21:42 · Guest teaching 0/10 Managing Risk, Scraping, and Moderation in AI Elad provides an in-depth breakdown of Google's historical copyright battles, fair use snippets, and robots.txt as a framework for current AI scraping and liability risks.21:43–25:37 · Guest teaching 0/10 Semiconductor Innovation and AMD's System Strategy Sarah details AMD's acquisition strategy, breaking down the specific roles of Silo AI for CUDA software alternatives and ZT Systems for data center rack-scale architecture.25:37–26:24 · Guest teaching 0/10 Robotics PSA and Episode Conclusion Brief outro segment featuring Elad's lighthearted PSA requesting leads on purchasing a humanoid robot.0:23–7:13 · Guest disagreement 1/10 LLM Market Consolidation and Capital Dynamics Sarah and Elad examine foundation model consolidation, capital requirements, and 200x token pricing drops. Sarah offers a slight nuance, arguing the market is simultaneously more competitive on performance benchmarks.7:14–14:21 · Guest disagreement 0/10 Domain-Specific Models and Multimodal Advancements The co-hosts explore specialized vs. multimodal architectures, drawing analogies to brain modularity and analyzing how rapidly smaller teams caught up to Sora in video generation.14:21–21:42 · Guest disagreement 0/10 Managing Risk, Scraping, and Moderation in AI Elad provides an in-depth breakdown of Google's historical copyright battles, fair use snippets, and robots.txt as a framework for current AI scraping and liability risks.21:43–25:37 · Guest disagreement 0/10 Semiconductor Innovation and AMD's System Strategy Sarah details AMD's acquisition strategy, breaking down the specific roles of Silo AI for CUDA software alternatives and ZT Systems for data center rack-scale architecture.25:37–26:24 · Guest disagreement 0/10 Robotics PSA and Episode Conclusion Brief outro segment featuring Elad's lighthearted PSA requesting leads on purchasing a humanoid robot.0:23–7:13 · The hosts pushing back 2/10 LLM Market Consolidation and Capital Dynamics Sarah and Elad examine foundation model consolidation, capital requirements, and 200x token pricing drops. Sarah offers a slight nuance, arguing the market is simultaneously more competitive on performance benchmarks.7:14–14:21 · The hosts pushing back 0/10 Domain-Specific Models and Multimodal Advancements The co-hosts explore specialized vs. multimodal architectures, drawing analogies to brain modularity and analyzing how rapidly smaller teams caught up to Sora in video generation.14:21–21:42 · The hosts pushing back 0/10 Managing Risk, Scraping, and Moderation in AI Elad provides an in-depth breakdown of Google's historical copyright battles, fair use snippets, and robots.txt as a framework for current AI scraping and liability risks.21:43–25:37 · The hosts pushing back 0/10 Semiconductor Innovation and AMD's System Strategy Sarah details AMD's acquisition strategy, breaking down the specific roles of Silo AI for CUDA software alternatives and ZT Systems for data center rack-scale architecture.25:37–26:24 · The hosts pushing back 0/10 Robotics PSA and Episode Conclusion Brief outro segment featuring Elad's lighthearted PSA requesting leads on purchasing a humanoid robot.

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

0:00 · the hosts 99.9% · guest 0.1%0:00 · the hosts 99.9% · guest 0.1%3:00 · the hosts 100% · guest 0%3:00 · the hosts 100% · guest 0%6:00 · the hosts 100% · guest 0%6:00 · the hosts 100% · guest 0%9:00 · the hosts 100% · guest 0%9:00 · the hosts 100% · guest 0%12:00 · the hosts 100% · guest 0%12:00 · the hosts 100% · guest 0%15:00 · the hosts 100% · guest 0%15:00 · the hosts 100% · guest 0%18:00 · the hosts 100% · guest 0%18:00 · the hosts 100% · guest 0%21:00 · the hosts 100% · guest 0%21:00 · the hosts 100% · guest 0%24:00 · the hosts 99.8% · guest 0.2%24:00 · the hosts 99.8% · guest 0.2%
Sharpest disagreement ▶ 2:17 Pushing back on the market consolidation narrative

Sarah counters Elad's thesis of consolidation by pointing out that the market has gotten noticeably more competitive on price and benchmarks.

Hardest push from the hosts ▶ 2:17 Reframing LLM market competitiveness

Sarah directly reframes the premise by distinguishing between capital moats and aggressive consumption-side benchmark competition.

Biggest teaching moment ▶ 15:20 Historical analysis of web scraping and copyright precedents

Elad educates on the historical legal evolution of Google search snippets, robots.txt, and fair use doctrines to contextualize modern AI data scraping.

The host holds their own ▶ 24:25 Deep breakdown of AMD system strategy

Sarah demonstrates authoritative technical expertise by mapping out AMD's software, networking, and rack-scale hardware stack components.

the scores for every segment, with the reasoning behind each
ChapterTopicThe hosts as informed peerGuest teachingGuest disagreementThe hosts pushing backWhy
LLM Market Consolidation and Capital Dynamics 7012 Sarah and Elad examine foundation model consolidation, capital requirements, and 200x token pricing drops. Sarah offers a slight nuance, arguing the market is simultaneously more competitive on performance benchmarks.
Domain-Specific Models and Multimodal Advancements 7000 The co-hosts explore specialized vs. multimodal architectures, drawing analogies to brain modularity and analyzing how rapidly smaller teams caught up to Sora in video generation.
Managing Risk, Scraping, and Moderation in AI 8000 Elad provides an in-depth breakdown of Google's historical copyright battles, fair use snippets, and robots.txt as a framework for current AI scraping and liability risks.
Semiconductor Innovation and AMD's System Strategy 8000 Sarah details AMD's acquisition strategy, breaking down the specific roles of Silo AI for CUDA software alternatives and ZT Systems for data center rack-scale architecture.
Robotics PSA and Episode Conclusion 1000 Brief outro segment featuring Elad's lighthearted PSA requesting leads on purchasing a humanoid robot.

Statements from this episode (10)

Opinion
Gil: Venture capital cannot fund the next foundation model rounds
“The venture capital industry is just too small to actually be able to support the next round for these companies.”
Elad Gil Sep 12, 2024 ▶ 1:12
Assertion Supported
Gil: LLM API costs dropped roughly 200x in 18 to 24 months
“The API costs have dropped something like 200 X. In the last 18 to 24 months or something along the, along that order of magnitude, David on my team actually pulled together a chart of all the pricing for all the various models and what that looks like over ti…”
Elad Gil Sep 12, 2024 ▶ 2:52
Prediction Not checkable as stated
Guo: Model providers will build proprietary API features to avoid commoditization
“You're going to see providers try and push customers down a more locked in path in terms of like APIs, right? We saw this with, you know, AWS. You're going to have sophistication around like, don't just do, it's a storage bucket. You're, you'll have prompt cac…”
Sarah Guo Sep 12, 2024 ▶ 5:08
Prediction Not checkable as stated
Gil: AI model market will fragment into niche winners like social networks
“I think the same thing is likely to be true here, where there'll be certain people who start to grab parts of the market. You know, LinkedIn became the sort of enterprise identity social network, or whatever you want to call it, right? Your resume social netwo…”
Elad Gil Sep 12, 2024 ▶ 8:46
Assertion Not checkable as stated
Guo: Small video AI startups are producing models competitive with Sora
“You could argue now that you have a handful of companies that are showing really amazing video generation capabilities, where it's not actually, like, a bunch of smaller players are step function behind there between runway and Pika. And even you know, if you …”
Sarah Guo Sep 12, 2024 ▶ 11:38
Prediction Held up
Guo: Real-time generative audio and video apps will emerge within months
“I think we're going to see really, really cool experiences on the image video audio side, because as you say, if the models get smaller and they get better they also get, you know, and there's different architectures, like what Cartesia is working on, you're g…”
Sarah Guo Sep 12, 2024 ▶ 13:41
Opinion
Guo: Scaling video AI models requires using YouTube data
“I'd be shocked to find out if there are ways to get to scale of video data that don't involve some YouTube data.”
Sarah Guo Sep 12, 2024 ▶ 18:24
Opinion
Gil: AI labs restrict model outputs beyond societal norms
“A lot of the companies have actually curtailed it more than the norms for much of society, right? There may be a slim part of society that feels a certain way, but for most of the society, you know, there tends to be, or it looks like there's broader tolerance…”
Elad Gil Sep 12, 2024 ▶ 20:51
Assertion Supported
Guo: AMD's $600M Silo AI acquisition was essentially a massive acqui-hire
“They bought this company silo a little while back, which is essentially like a six hundred million dollar aqua hire of hundreds of AI engineers and researchers who've done a lot of work on AMD.”
Sarah Guo Sep 12, 2024 ▶ 24:35
Opinion
Guo: AMD bought ZT Systems as a billion-dollar data center acqui-hire
“The theoretical like missing component that ZT fills is if you think of it as like a one to two billion dollar aqua hire of a thousand systems designs folks to support the like rack and data center scale AI business instead of like individual chips or componen…”
Sarah Guo Sep 12, 2024 ▶ 24:56
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 100 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.