The Hugging Face ecosystem extends in many directions and is characterized by hosting the largest collection of models and datasets.
When artificial intelligence (AI) is discussed in the media landscape, tech giants and their massive developments are often highlighted. However, when thinking about applying AI in news organizations, it is essential to look further and recognize how this technology—advancing exponentially in every direction—can be used in concrete processes ethically and effectively.
That is why this year the media innovation conference, Media Party, has as one of its main goals exploring applied AI, including its ethical, security, and privacy challenges. Along these lines, one of the most prominent presences this year is Hugging Face, the platform that hosts one of the world’s largest open-source machine learning communities. We spoke with Brigitte Tousignant, Communications Lead, about how Hugging Face is collaborating with the media industry.
The “GitHub” of Machine Learning
Founded in 2017, Hugging Face originated from the creation of a chatbot for which they developed a natural language processing model. Due to its positive impact, they decided to release this model to the community. This milestone allowed them to diversify their focus toward open-source machine learning, with Bloom being a prominent example. Additionally, they expanded their specialization into AI-driven image and audio generation and processing.
The Hugging Face ecosystem extends in multiple directions and is characterized by hosting the most extensive collection of models and datasets. Furthermore, it provides a space where a diverse community of developers, researchers, and enthusiasts from around the world can publish their own AI applications so they can be used without needing massive resources.
Hugging Face in numbers:
- 300,000 open-source machine learning models.
- 250,000 available datasets.
- 250,000 free AI applications.
- Over 50,000 organizations on the platform.
- 1 million repositories and projects.
3 Questions for Brigitte Tousignant on Hugging Face and AI Applications in Journalism
1. What is Hugging Face’s vision regarding the application of artificial intelligence in the media?
Brigitte: Helping creators and complementing—not replacing—their work is one of our main objectives. On Hugging Face, users can find tools to both ideate and create impactful projects. With our ethics group, we can help media organizations understand the limitations, harms, and biases of different AI tools before they spread. Ethical considerations regarding AI authorship, accountability, and transparency must be considered and incorporated into editorial guidelines. The key to this is consent, credit, and compensation, which is made possible through dataset transparency—one of the foundational pillars of Hugging Face’s vision.
2. What advice can you give journalists interested in leveraging Hugging Face’s capabilities in their work? Do you have specific examples of how Hugging Face has collaborated with newsrooms to address industry-specific challenges?
Brigitte: The first step is to create an account and try it out. Hugging Face is currently collaborating with Hacks/Hackers to develop a library of Hugging Face apps that could be useful for journalists: DIY for journalists collections.
Additionally, our work on bias in generative AI models was influential for this excellent Bloomberg article on the same topic and cited in research conducted by MIT Technology Review. Furthermore, our collaboration on dataset transparency also aided The Washington Post in its investigation into datasets used to train chatbots like ChatGPT.
3. What are the benefits of open source in the creation and use of artificial intelligence and machine learning?
Brigitte: The rapid advancement of AI over the past decade can be attributed to open source: the quick proliferation of ideas allows researchers to build on each other’s strengths and learn from mistakes. It also lowers entry barriers, allowing anyone to use AI models (regardless of their expertise level) and making it more accessible and inclusive. We firmly believe that open source leads to greater innovation, accountability, and transparency, whereas closed-source systems are often centralized, controlled, and opaque.
If you want to learn more, don’t miss Brigitte Tousignant’s keynote and the workshop “Building Charming Machine Learning Applications with Gradio” by Freddy Alfonso Boulton Gonzalez, software developer at Hugging Face.
We look forward to seeing you from October 5 to 7 at Konex — register at Mediaparty.org
