What "Open" Means

March 1, 2024 | John Shaughnessy

An image of a person wearing headphones, leaning forward over a desk with various electronics.

What does it mean to be "open" in the era of AI?

The Mozilla Manifesto lays out Mozilla's principles and mission. It articulates a vision for an internet that continues to benefit the public good and commercial aspects of life. It clarifies the common cause behind Mozilla's expanding family of software products and the Foundation's role as advocates, community organizers, and policy advisors.

The word "open" appears several times in the manifesto, the mission, and Mozilla's other publications. What does it really mean? And how is the meaning of "open" changing in this era of AI?

"Open" as in "Libre"

In English, "free" is ambiguous. The Free Software Foundation (FSF) used to highlight the difference with the phrase, "Free as in 'freedom', not 'free beer'." In French we'd distinguish "libre" or "liberté" (liberty) from "gratis" (without charge).

The first meaning of "open" refers to this concept: Liberty. Freedom.

It is freedom from: The freedom from surveillance, control, persecution.
And it is freedom to: The freedom to do what you wish with your hardware.

The FSF defines four essential freedoms that qualify "free software":

- The freedom to run the program as you wish, for any purpose (freedom 0).
- The freedom to study how the program works, and change it so it does your computing as you wish (freedom 1). Access to the source code is a precondition for this.
- The freedom to redistribute copies so you can help others (freedom 2).
- The freedom to distribute copies of your modified versions to others (freedom 3). By doing this you can give the whole community a chance to benefit from your changes. Access to the source code is a precondition for this.

"Openness" refers in some way to the spirit of these freedoms, but does not adhere strictly to this definition.

"Open" as in "Open Source"

The "official" definition of open source comes from the Open Source Initiative (OSI), which defines "open source software" as software that complies with its Open Source Definition (OSD). This document outlines 10 specific requirements, mostly focused on the software license.

This definition of "open source" has much in common with the FSF's definition of "free software". They're so similar, in fact, that many people use the acronyms "FOSS" (Free and Open-Source Software) and "FLOSS" (Free/Libre and Open-Source Software) to refer to both.

The difference is mostly historical and philosophical: The FSF believes that the OSI focuses too much on the practical benefits of non-proprietary software, rather than on the ethical issues behind restricting users' rights to change and improve code on their own terms (source).

"Open" as in "Source Available"

Meanwhile, there's an unofficial, colloquial understanding of "open source" in widespread use today, which roughly translates to "source available".

For example, when people say that Unreal Engine is "open source", they really mean that you can read the source code. However, Unreal Engine's license introduces restrictions on commercial use and royalty payments.

Although people familiar with the specific definitions of "open source" try to correct this misunderstanding, the conflation of "open source" with "source available" remains pervasive.

"Open" as in "Open Weights", "Open Science", and "Open Data"

With the dramatic rise of AI models, more variations of "open":

"Open Weights" roughly translates to "You can download and use the model".
"Open Science" roughly translates to "You can read publications without paying exorbitant fees from scientific journals." The publications explain model architectures, details about training processes, and benchmarks used to measure model performance and limitations.
"Open Data" roughly translates to "you can download the data set used to train this model".

When Meta refers to Llama as an "open platform", others rightly point out that although the weights are available for download and their publications qualify as "open science", they did not publish the dataset used to train the model and Llama's community license specifically prohibit commercial use by products with large user bases:

2. Additional Commercial Terms. If, on the Llama 2 version release date, the 
monthly active users of the products or services made available by or for Licensee, 
or Licensee's affiliates, is greater than 700 million monthly active users in the 
preceding calendar month, you must request a license from Meta, which Meta may 
grant to you in its sole discretion, and you are not authorized to exercise any of the 
rights under this Agreement unless or until Meta otherwise expressly grants you 
such rights.

Commentary

Allow me to editorialize for a moment. These are my personal opinions:

I am grateful for Yann LeCun and the Meta AI team's work to advance AI research and development, their public advocacy for "open source" foundation models, and arguments against some of the fear-mongering rhetoric around AI.

That is not to say they're perfect, but the world is better for having Llama in it.

Mozilla will rightfully continue to be vocal about the need for Trustworthy AI, its philosophy of open and accessible systems and its criticism of Facebook's pervasive tracking.

In a cultural climate inundated with short-form content and click-bait headlines, we should seek and promote nuance.

"Open" as in "OpenAI"

Let's get this one out of the way. OpenAI publishes research papers, which qualifies it as "Open Science". But OpenAI is far less open than their name suggests. Their publications sometimes lack critical details that would allow people to replicate the work easily. Their model weights and data sources are proprietary, and their APIs are behind proprietary licensing. Changes they make to their models or products happen behind closed doors.

"Open" as in "participatory"

The Mozilla Manifesto emphasizes the need to involve and engage with a diverse, global community. This notion of openness extends beyond access to source code or data; it means creating an environment where everyone can contribute, participate, and affect outcomes. Participatory openness aims to democratize development processes and ensure that a wide array of perspectives and expertise are represented and valued.

"Open" as in "open debate"

The world is a rapidly changing and impossibly complex. None of us has it all figured out. Openness in this context means being transparent, open to criticism, and supportive public debate. It means encouraging honest discussion about the implications, ethics, and impact that rapidly evolving technology has on our lives.

"Open" as in "permissionless", "open platform", and "open standards"

The Internet is a collection of protocols and standards that allow any organizations to build on top of an equal playing field. Mozilla promotes the use and development of protocol networks that, by their very nature, are resistant to corporate capture, unjust censorship, or centralized control.

A "protocol network" is a network built on an open standards. The word "protocol" refers to the process by which a participant can engage with the network. Protocol networks allow anyone able to build, participate, or extend the network without permission. Examples include: - Email (SMTP) - World Wide Web (HTTP) - Web Feeds (RSS)

Open platforms and protocol networks stand in stark contrast to "corporate networks" like Facebook and Twitter. While these networks promoted third-party app development when they started, they eventually cut off access, stifling innovation and forcing smaller participants to shut down or scale back. In the tech world, this dynamic is called "Platform Risk", and it's inherent to all centrally-managed corporate networks.

One of the major differences between protocol networks and corporate networks are where power aggregates. In protocol networks, power aggregates at the "leaf nodes": The users of the network. By contrast, corporate networks naturally aggregate power to the center.

Mozilla's commitment to an open Internet refers to its belief in the importance of protocol networks, permissionless innovation, healthy competition, and putting people over profits.