Spaces:
Running
Running
Update README.md
Browse files
README.md
CHANGED
@@ -7,4 +7,52 @@ sdk: static
|
|
7 |
pinned: false
|
8 |
---
|
9 |
|
10 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
7 |
pinned: false
|
8 |
---
|
9 |
|
10 |
+
# Building up Open Source AI with the Hugging Face Community
|
11 |
+
|
12 |
+
As the field of AI advances, it's becoming increasingly important to understand what "Open Source" means in the context of artificial intelligence.
|
13 |
+
While Open Source software has long had established standards, AI models and datasets bring new challenges around transparency, accessibility and collaboration.
|
14 |
+
|
15 |
+
To address this, the Open Source Initiative (OSI) has been developing an **Open Source AI Definition** (OSAID) which will outline the criteria that AI components—such as models,
|
16 |
+
datasets and code—must meet to be considered Open Source.
|
17 |
+
|
18 |
+
## Co-designing the Open Source AI Definition v1.0
|
19 |
+
|
20 |
+
Understanding that AI impacts a global community with diverse perspectives, we embarked on the quest to define Open Source AI by adopting a co-design methodology—a
|
21 |
+
participatory approach that shares knowledge and power among all stakeholders. This method allows us to integrate diverging viewpoints into a cohesive and consensual
|
22 |
+
standard.
|
23 |
+
|
24 |
+
The 1.0 version of the definition is set to be unveiled on October 28, 2024, at the [All Things Open](https://2024.allthingsopen.org/osi) conference.
|
25 |
+
The definition lays out the main requirements for what we hope will become a robust standard for OSAI across the ecosystem.
|
26 |
+
|
27 |
+
## Open Source AI in practice at Hugging Face
|
28 |
+
|
29 |
+
To support the Open Source AI Definition, we need to see how it will be operationalized in practice by the people building and using Open Source AI systems.
|
30 |
+
What Open Source licenses and legal frameworks are favored for sharing model weights? How is training data re-used and documented?
|
31 |
+
Given the role of the Hugging Face platform in showcasing those decisions, we invite the community to use this resource to share that information and discuss
|
32 |
+
new related questions as they come up; so we can keep building up Open Source AI with a shared vocabulary and understanding.
|
33 |
+
|
34 |
+
## What is the purpose of the Open Source AI Definition (OSAID)?
|
35 |
+
|
36 |
+
The OSAID outlines the core freedoms that should apply to AI systems to be considered truly Open Source. These include the right to:
|
37 |
+
- **Use** the system for any purpose and without having to ask for permission.
|
38 |
+
- **Study** how the system works and inspect its components.
|
39 |
+
- **Modify** the system for any purpose, including to change its output.
|
40 |
+
- **Share** the system for others to use with or without modifications, for any purpose.
|
41 |
+
|
42 |
+
## What are the benefits of Open Source AI?
|
43 |
+
Open Source AI offers significant benefits, including enhanced transparency and safety by enabling system audits, mitigating bias, and ensuring accountability
|
44 |
+
in data sources, which accelerates AI safety research. It also fosters competition and innovation by making more models available, challenging AI monoculture,
|
45 |
+
and providing access to foundational technology for a wider range of stakeholders. Additionally, Open Source AI supports the development of diverse, localized
|
46 |
+
applications that reflect different cultural and linguistic contexts, allowing models to be aligned with various value systems.
|
47 |
+
|
48 |
+
# How can the Hugging Face community participate?
|
49 |
+
As the largest platform for the AI community, Hugging Face plays a key role in promoting the Open Source principles in the context of AI.
|
50 |
+
We would like to invite both individuals and organizations of the Hugging Face community to [review](https://opensource.org/deepdive/drafts/the-open-source-ai-definition-1-0-rc1),
|
51 |
+
[endorse](https://opensource.org/osaid-endorse), and adopt the OSAID, and to [discuss OSAI artifacts](https://huggingface.co/datasets/opensourceorg/osaid/discussions/1) on this page.
|
52 |
+
This is your chance to take part in a key moment for AI’s future.
|
53 |
+
|
54 |
+
As we approach the release of the Open Source AI Definition v. 1.0, your support is fundamental.
|
55 |
+
We encourage you to explore the definition, contribute models that adhere to the OSAID, and engage in discussions around Open Source AI.
|
56 |
+
Together, we can ensure that Open Source AI continues to evolve in ways that are fair, transparent, and accessible to all.
|
57 |
+
|
58 |
+
|