Inside a radical new project to democratize AI: the case of BLOOM

Inside a radical new project to democratize AI: the case of BLOOM

In the world of artificial intelligence (AI), there has been a growing concern about the lack of transparency and accessibility of large language models (LLMs). These models, such as OpenAI's GPT-3 and Google's LaMDA, have been hailed for their impressive capabilities, but their inner workings have been closely guarded secrets. However, a new project called BigScience is aiming to change this status quo by developing a large language model called BLOOM (BigScience Large Open-science Open-access Multilingual Language Model) that is designed to be transparent and accessible to all.

BLOOM is the result of a year-long effort by over 1,000 volunteer researchers coordinated by AI startup Hugging Face and funded by the French government. Unlike other LLMs, BLOOM is designed to be as transparent as possible, with researchers sharing details about the data it was trained on, the challenges in its development, and the way they evaluated its performance. This level of transparency is unprecedented in the AI community and represents a radical departure from the norm.

One of the key selling points of BLOOM is its ease of access. Now that it is live, anyone can download and tinker with it free of charge on Hugging Face's website. This opens up opportunities for AI developers to use BLOOM as a foundation to build their own applications. With 176 billion parameters, BLOOM is even bigger than OpenAI's GPT-3, offering similar levels of accuracy and toxicity. This makes BLOOM a powerful tool for researchers and developers looking to explore the potential of AI.

The release of BLOOM is a significant step towards democratizing AI. Most big tech companies restrict the use of their LLMs by outsiders and have not released detailed information about their models. Meta, for example, released its own large language model called Open Pretrained Transformer (OPT-175B) in 2022, but it is only available upon request and has a limited research-only license. Hugging Face, on the other hand, goes a step further by launching a new Responsible AI License. This license acts as a deterrent from using BLOOM in high-risk sectors or for harmful purposes, ensuring that the model is used responsibly.

Ethics has been a guiding principle throughout the development of BLOOM. Hugging Face had ethical guidelines in place from the very beginning, and Giada Pistilli, Hugging Face's ethicist, drafted BLOOM's ethical charter. These guidelines ensure that the model is developed and used in a way that respects human values and promotes fairness and inclusivity.

The development of BLOOM was not without its challenges. One of the main obstacles was the lack of suitable data sets in languages that are not as well represented online. This is where the power of collaboration came into play. The BigScience project rallied volunteers from around the world to build data sets in different languages, ensuring that BLOOM is a truly multilingual model. This collaborative effort highlights the potential of open science and the impact it can have on AI research.

In conclusion, the release of BLOOM represents a significant milestone in the democratization of AI. Its transparency and accessibility set it apart from other large language models, opening up new possibilities for researchers and developers. However, it is essential to use AI responsibly and adhere to ethical guidelines to ensure that these powerful tools are used for the benefit of humanity. Here are three actionable pieces of advice for AI developers:

1. Embrace transparency: Share details about the data your models are trained on and the evaluation process. This will foster trust and allow the AI community to learn from your work.

2. Collaborate for inclusivity: Seek out collaborations with researchers and volunteers from diverse backgrounds to ensure that your models are representative and inclusive.

3. Develop ethical guidelines: Prioritize ethical considerations from the beginning of your projects. Establish clear guidelines to ensure that your models are used responsibly and do not cause harm.

By following these advice, we can collectively work towards a future where AI is accessible, transparent, and beneficial to all.