May 19, 2023 · 37m · no-priors
No Priors Ep. 5 | With Huggingface’s Clem Delangue
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
Hugging Face co-founder and CEO Clem Delangue joins Sarah Guo and Elad Gil to discuss the evolution of Hugging Face, the necessity of open-source AI, startup execution strategies, and emerging frontiers across machine learning.
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 17.4% of the talking time here. How this is scored →
The hosts as informed peer 4.3 Guest teaching 2.5 Guest disagreement 1.2 The hosts pushing back 0.4
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 31:45 Pushing back against the GitHub missed opportunities framing
Clem immediately challenges Elad's premise that GitHub missed key commercial opportunities, asserting GitHub is an incredible $1B ARR business whose only mistake was selling too early to Microsoft.
Hardest push from the hosts ▶ 31:58 Host clarifies and restates the platform expansion premiseElad refuses to let his question be interpreted as disparaging GitHub, directly restating that the platform had immense uncaptured opportunities in security and on-prem enterprise tooling.
Biggest teaching moment ▶ 12:42 Dismantling the slow research-to-enterprise adoption mythClem reframes Sarah's concern about the enterprise research gap by showing that ML translates from scientific papers to production in days or weeks compared to multi-decade lag in traditional sciences.
The host holds their own ▶ 30:33 Elad details developer platform monetization vectorsElad demonstrates substantial venture and ecosystem expertise by listing concrete lines of business like Snyk, Socket, and GitLab enterprise workflows as untapped developer hub opportunities.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Clem Delangue's Journey from eBay to Machine Learning | 3 | 0 | 0 | 0 | Elad opens with a warm biographical prompt regarding Clem's transition from eBay to machine learning. Clem shares an entertaining origin story without any friction. | |
| The Origins and Open-Source Pivot of Hugging Face | 3 | 1 | 0 | 0 | Elad asks about the pivot from an AI Tamagotchi chatbot to an open-source hub. Clem explains how community traction around BERT steered the team towards their current focus. | |
| Balancing Exploration and Exploitation in Startups | 5 | 1 | 0 | 0 | Elad demonstrates domain knowledge by drawing parallels between Hugging Face's pivot and Stewart Butterfield's trajectory with Flickr and Slack. Clem outlines his 30-40% exploration rule. | |
| Organic Growth and Direct Technical Community Engagement | 4 | 1 | 0 | 0 | Elad asks about organic distribution tactics versus targeted community outreach. Clem details their non-traditional stance against hiring dedicated PR or community managers. | |
| Cultivating an Authentic Open Source Culture | 4 | 4 | 2 | 0 | Sarah asks about closing the enterprise adoption gap with cutting-edge ML research. Clem gently reframes the premise by contrasting ML's rapid cycle of days and weeks against traditional science's multi-decade latency. | |
| Open Source Plurality Versus Proprietary AI Monopolies | 5 | 5 | 3 | 1 | Sarah raises industry concerns about proprietary labs with massive compute moats dominating AI. Clem pushes back against the winner-take-all narrative, citing historical software parallels and Hugging Face's quarter-million uploaded models. | |
| Model Modalities, Architectures, and Size Trade-Offs | 4 | 3 | 0 | 0 | Sarah inquires about the distribution of modalities and model sizes on the hub. Clem provides a thorough breakdown from NLP and vision to real-time latency use cases like Bloomberg terminal models. | |
| Infrastructure Efficiency, Online Learning, and Data Consent | 4 | 3 | 1 | 0 | Elad prompts Clem on infrastructure frontiers and non-competitive tooling wishlists. Clem criticizes ecosystem 'cloud money laundering' and advocates for online learning and data consent architectures. | |
| Project BLOOM, BigScience, and Democratizing AI | 4 | 4 | 2 | 0 | Sarah asks about Hugging Face's direct involvement in foundational training like BLOOM. Clem explains BigScience's collaborative structure and passionately argues that open science prevents concentrated demographic and geographic bias. | |
| Adapting to RLHF and Rapid Paradigm Shifts | 5 | 2 | 0 | 0 | Sarah asks about RLHF adoption and Elad asks about commercialization models. Clem explains their rapid integration of new ML paradigms and previews enterprise freemium tiers. | |
| The GitHub Comparison and Monetizing the Compute Gateway | 6 | 4 | 4 | 4 | Elad cites specific developer platform business lines (Snyk, Socket, GitLab) to ask about GitHub's missed opportunities. Clem pushes back in defense of GitHub's $1B revenue scale, prompting Elad to clarify his framing while Clem highlights compute monetization. | |
| Future Frontiers: Biology, Chemistry, and ML-Native Companies | 5 | 2 | 2 | 0 | Sarah asks about future application frontiers in biology and chemistry. Clem explicitly refuses to make narrow predictions, emphasizing full-stack ML-native startups instead. |
Statements from this episode (22)
Delangue: Hugging Face's early AI companion saw billions of messages exchanged
“And we did that for almost, almost three years got some level of traction billions of messages exchanged between, between users and the chatbots.”
Delangue: Hugging Face dedicates 30% to 40% of effort to exploration
“What we've always done and I think we'll always do with hugging face is to make sure that, you know, Whenever, always kind of like make sure to spend at least like 30 or 40% of the company's efforts on explorating new things and kind of like finding the long-t…”
Delangue: Hugging Face Spaces reached 50,000 ML demos in ~18 months
“We just crossed 50,000 of machine learning demos in the past year, year and a half on that.”
Gil: Companies that do not innovate early never innovate later
“It seems like in general companies that iterate or launch new things early, keep launching things later in the life of the company and companies that never innovate early don't ever innovate again in their lives is kind of like the difference between eBay and …”
Hugging Face never hired dedicated PR or community managers, per Delangue
“And then something that we did that I think worked really, really well for us is that we never hired any
Kind of like community manager, any kind of like communication PR kind of team members because we wanted it to be part of every single team members work.”
Every Hugging Face employee has access to its official Twitter account
“Everyone, everyone in the team has access to the Twitter account and are tweeting from the Twitter account.”
Machine learning models move from research to production in days
“What we're seeing in, in machine learning is that it's actually making its way into production after, you know, a year, a few months, a few weeks sometimes a few days now. So this is, in my opinion, this is amazing. And that's what's driving Most of the speeds…”
Delangue: AI Would Be Decades Behind Without Five Years of Open Source
“If you know, remove the open source from that equation, if they wouldn't have been as much open source at As there's been in the five past five years, we would be like decades away from, you know, where, where we are now.”
Delangue: Open-source and proprietary AI will both permanently coexist
“The truth is that there's always going to be both, right? I think if you see most, most technologies, you know, if you look at search, you always have the elastic search, the Algolia, or like, if you look at databases, you have that MongoDB and the proverter a…”
Delangue: Hugging Face crossed 250,000 models from nearly 15,000 companies
“On Hugging Face we just crossed 250,000 models, right? A quarter of a million models uploaded by almost 15,000 companies now.”
Delangue: Specialized AI models are cheaper, faster, and usually more accurate
“When you look at why are companies using so many models on Hugging Face, you usually realize that you know, a more specialized model is more efficient. It's cheaper to run. It's usually faster to run. And most of the time actually more accurate for the specifi…”
Delangue: NLP, vision, and audio are the top three Hugging Face tasks
“The three main tasks right now are NLP, so text, right?
From like information extraction, text generation, text classification.
the second one is text to image and computer vision, right?
So object detection, text to image, text image generation.
the third o…”
Uber calculates ETAs using a time series transformer model, says Delangue
“The ETA from Uber, when you get your Uber is like a transformer time series”
Delangue: Hugging Face sees rising adoption of opt-in training datasets
“We're starting to see on Hugging Face more and more opt-in data sets, meaning like data sets that have been trained only on data that the creators of the data have consented to having a model trained on.”
BigScience is the largest ML collaboration to date, says Delangue
“Big science was like the largest collaboration in, in machine learning to date with like a thousand researchers from 200 organizations, kind of like coming together in order to build and train a large language model completely in the open.”
Open-source AI prevents bias better than closed labs, argues Delangue
“Building in the open with open source is actually
more part of the solution than part of the problem, because obviously control of power as you democratize it much more and biases you actually include in the process.
People who are impacted by these biases,…”
Hugging Face is building an open-source library for RLHF, says Delangue
“We're leading the development of an open source library that is helping companies integrate that into their models and into their workflows.”
Hugging Face has 15,000 company users and 3,000 paying customers
“Right now we have like 15,000 companies using, using the platform and we have 3000 companies paying us, right?”
Delangue: Bloomberg and Meta are Hugging Face customers
“When we're talking about Bloomberg, for example, using us or Meta using us both being kind of like customers.”
GitHub probably sold to Microsoft too early, argues Delangue
“So when you talk about, you know, money in GitHub for me, like the, if there was a mistake, it's probably to have sold it too, too early to Microsoft.”
Developer platforms like Hugging Face become gateways for compute and infrastructure
“When you get so much usage, so much network effects, and you actually like for a lot of these projects, you are like the starting point. Of projects like the same way. I think probably companies are starting on GitHub to see the open source projects before sta…”
ML-native companies like Runway are poised to challenge tech incumbents
“And when you look at the capabilities of some of these companies, when they're like translated into product building I mean, runway, you've all seen like the videos of runway. I think that's amazing. And I think they're going to be able to really challenge, ch…”