Expression is Compression - Democratizing AI

Expression is Compression - Democratizing AI

In the world of creativity and communication, expression is compression. Just as tree sap becomes maple syrup through a process of collection and refinement, experiences become shareable creations. This concept holds true for full-time authors, who spend most of their time collecting experiences rather than constantly typing away at their computers. By sitting down to make sense of their existing ideas, writers mold the wet clay of experience into shape, creating well-compressed stories that captivate their audience.

The art of compression is not limited to storytelling alone. It extends to the field of artificial intelligence (AI) as well. In a radical new project to democratize AI, a large language model (LLM) called BLOOM has been developed with the aim of making AI more transparent and accessible. Unlike other famous LLMs like OpenAI's GPT-3 and Google's LaMDA, BLOOM is designed to be as transparent as possible, with researchers openly sharing details about its development process and performance evaluation.

BLOOM was the result of a year-long collaborative effort by over 1,000 volunteer researchers in a project called BigScience. Coordinated by AI startup Hugging Face and funded by the French government, this project aimed to break away from the traditional closed-door approach to AI development. The model's biggest selling point is its ease of access. Anyone can now download BLOOM and experiment with it for free on Hugging Face's website, using it as a foundation to build their own AI applications.

With 176 billion parameters, BLOOM is larger than OpenAI's GPT-3 and offers similar levels of accuracy and toxicity control. While most big tech companies restrict access to their LLMs and keep the inner workings of their models under wraps, Hugging Face takes a different approach. Not only do they make BLOOM available to the public, but they are also launching a new Responsible AI License, acting as a deterrent for the model's use in high-risk sectors or for malicious purposes.

Ethics played a crucial role in the development of BLOOM. From the very beginning, the project had its own ethical guidelines that served as guiding principles. Hugging Face's ethicist, Giada Pistilli, drafted BLOOM's ethical charter, ensuring that the model's development aligns with responsible AI practices. By placing ethics at the forefront, Hugging Face aims to mitigate potential risks and ensure that BLOOM is used for the greater good.

One of the reasons BLOOM was able to improve upon existing models is its diverse range of data sets. The team behind the project rallied volunteers from around the world to build suitable data sets in languages that were not well represented online. This inclusivity and global collaboration allowed BLOOM to overcome the limitations of existing models and provide a more comprehensive and accurate representation of different languages and cultures.

While concerns about the sexist and racist language embedded in AI models have led companies like OpenAI to keep their models and code private, Hugging Face believes that having an open language model like BLOOM for research purposes has a strong long-term impact. By providing researchers and developers with the opportunity to study and experiment with a transparent and accessible AI model, new insights and innovations can emerge.

In conclusion, the concept of expression as compression applies not only to storytelling but also to the field of AI. BLOOM, a radical new project aiming to democratize AI, is a prime example of this. By making AI more transparent and accessible, Hugging Face is paving the way for responsible and inclusive AI development. As we move forward, it is crucial to prioritize ethics and ensure that AI technologies are used for the greater good. Here are three actionable pieces of advice for those venturing into the world of AI:

1. Embrace transparency: Share details about the development process, data sets used, and evaluation methods to foster trust and collaboration.

2. Prioritize ethics: Develop clear ethical guidelines and ensure that AI models are used responsibly, with a focus on avoiding harm and exploitation.

3. Foster inclusivity: Encourage global collaboration to build diverse data sets that accurately represent different languages, cultures, and perspectives.

By following these principles, we can work towards a future where AI is not only powerful but also beneficial to society as a whole.