logo
Menu
Let's Build a Startup - S2E2 - Anatomy of a Unicorn: Hugging Face - Generative AI, the importance of Data, and the Open Source community.

Let's Build a Startup - S2E2 - Anatomy of a Unicorn: Hugging Face - Generative AI, the importance of Data, and the Open Source community.

Julien Simon, Chief Evangelist at Hugging Face, shares his views on LLMs, Generative AI, the Open Source Community and building an AI fueled startup

Giuseppe Battista
Amazon Employee
Published Jun 17, 2024
Oh what an episode this week! We had the unquantifiable honour of chatting with Julien Simon, Chief Evangelist at Hugging Face, the Unicorn who won the hearts and mind of the AI Open Source Community.
Here are a few highlights from our conversation.

AI: "First and Foremost, it is a business conversation"

Julien emphasizes the importance of focusing on business needs when discussing AI and large language models (LLMs) with a diverse range of customers, from startups to enterprises. He stresses that technology, including AI, must address real-life business problems and improve organizational productivity or customer experience. He advocates for a neutral, data-driven approach and the use of the right tool for the job, whether it's a closed model, open source, or traditional machine learning. He highlights the importance of maintaining focus on cost, performance, and return on investment. "The goal is to provide clarity and ensure that businesses make informed, pragmatic decisions about their AI and machine learning implementations"

"Data is how you create competitive advantage, not models"

Julien underscores how crucial it is to have a robust data strategy before jumping onto AI projects. He stresses that data operability is the most critical aspect of an AI project's success and more valuable in the long term compared to models, which are expendable and rapidly become outdated. He then highlights that while models evolve quickly, the data, especially historical data, retains its value and provides a competitive advantage. Focus on data and make sure you have the widest range of models available to you.

"LLMs are not intelligent, there is no reasoning"

Julien highlights that while large language models (LLMs) are not capable of human-like reasoning, they excel in their designed function: predicting the next word in a sequence. This capability allows them to generate coherent and contextually relevant text, which often feels remarkably human-like.
This predictive power, although it might seem magical, is rooted in sophisticated statistical analysis rather than understanding.

"Closed model builders have decided for you"

Julien discusses the complexities associated with using closed models and the importance of localization in AI. He highlights that while some closed models exhibit impressive capabilities, their creators make critical decisions about data curation, alignment processes, and system prompts. This can save time but might also introduce biases or limitations.
He encourages local users, open source communities, and governments to take charge of building their models, ensuring they are tailored to their specific cultural and linguistic needs. Partnerships with tech companies can help speed up these efforts, but local ownership and control over data and evaluation processes are essential for success.
In conclusion, Julien advocates for a decentralized approach to AI development, where local expertise and context are prioritized, leading to more relevant, accurate, and culturally sensitive AI solutions.

Bias, risk management, cultural differences

Julien concludes the interview providing a thoughtful perspective on the challenges and opportunities of using AI responsibly. He acknowledges that while AI can be misused by bad actors, these individuals were likely problematic even before AI's advent. The focus should be on dealing with these actors rather than solely blaming AI.
Simon emphasizes the importance of a proactive approach in business and enterprise settings when it comes to AI. He discourages excessive bureaucracy and overthinking, advocating instead for a hands-on, experimental approach to understand and manage risks associated with AI. By experimenting and testing AI models in various scenarios, businesses can identify specific risks and address them effectively.

Watch the full episode on Twitch

Loading...

About Julien Simon

If you want to stay up to date with Julien's insights, please consider using the following resources:

About Hugging Face on AWS

Get started with Hugging Face models on SageMaker!

About Let's Build a Startup!

AWS Let’s Build a Startup on Twitch is a one hour weekly show that will accompany you throughout summer. This is our second season, and we have 14 more episodes ahead of us, with a line-up bringing together founders, industry leaders, investors, and subject matter experts from across the globe. This season’s episodes are packed with tens of sessions on topics including generative AI, raising venture capital, scaling to unicorn, MVP monetization, and so much more.
Read more about Let's Build a Startup on our Livestream Page on Community.aws

Episode Engagement Metrics

metrics from S2E2
We started with around 500 viewers, then we averaged 110 live viewers. Not bad compared to last week!
We had a total of 50 messages from 11 unique chatters.
CTA clicks as of 15/06: 65

Authors

Giuseppe Battista is a Senior Solutions Architect at Amazon Web Services. He leads soultions architecture for Early Stage Startups in UK and Ireland. He hosts the Twitch Show "Let's Build a Startup" on twitch.tv/aws and he's head of Unicorn's Den accelerator. Follow Giuseppe on LinkedIn
Nicolas David is a Senior Solutions Architect at Amazon Web Services. With 20+ years of experience in IT within major Industry defining players in PLM & Security Software Solutions, Banking and Insurance, Nicolas gained his experience in USA, Europe and the Middle East. His professional commitment through innovation led him to manage the Digital Innovation technical team for AWS Cloud Innovation Centers in the Kingdom of Bahrain (Bahrain Polytechnic & the University of Bahrain). After pivoting early 2022, Nicolas now focuses on Engaged Startups in the Middle East. Follow Nicolas on LinkedIn
 

Any opinions in this post are those of the individual author and may not reflect the opinions of AWS.

Comments