Can Open Source Software Stand in the AI Age?

How will software like LibreOffice, GIMP and Inkscape compete with Microsoft's and Google's AI tools?

The last few months have been groundbreaking for AI technologies.

Almost every major company in the world started pushing for new AI tools and features to be included in their services to some extent. The most remarkable of these were:

• Microsoft adding AI to their Office 365 service, allowing the user to interact with Word & Excel documents in much more powerful ways than before. A user can ask the integrated AI bot to create a chart or analyze data, and the bot will do it automatically.

• Microsoft announcing “Bing AI”; which is powered by OpenAI’s latest GPT4 model, bringing the power of large language models (LLMs) to interactive search results.

• Microsoft Designer being slowly released; an AI-powered Canva-alternative that can generate images on the fly using DALL-E 2.

• GitHub announcing Copilot X; a new generation of Copilot that now can help developers in solving issues, writing documentation or managing pull requests in GitHub repositories. Which allows a much more rapid software development process overall.

• Google announcing PaLM API; a new Google Cloud service that allows developers to access their “PaLM” LLM via an API to develop AI-powered software. They also announced the integration of Google Workspaces with this new LLM, so that users can effectively make documents and sheets with the help of AI.

• OpenAI itself announcing the availability of “plugins” to ChatGPT, which are additional integrations to the bot to allow it to access 3rd-party services and APIs, including online URLs. ChatGPT is now more powerful than ever before.

All of these advances and new services are good and all… However, one critical issue remains: They are all proprietary and closed-source.

Thanks to their new capabilities, these proprietary solutions are expected to continue to gain more users and expand their monopolies in different software sectors.

This is not a surprise. AI-powered software indeed saves huge time and effort in many areas, so most people will prefer that over anything that does not have the same capabilities.

But this leaves old-fashioned open source software in a tough spot: They have none of the AI capabilities provided by these giant corporations, so how will they be able to compete and stay relevant?

Why should a digital artist, for example, use GIMP while Adobe Photoshop has tons of AI-powered features that much more effortlessly enable image editing and creation? Or why should anyone keep using LibreOffice if you have Google Workspaces/Microsoft Office 365 AI capable of doing a better job for a fraction of the effort?

Open source software need to adapt to the latest AI trends and start providing useful features for their users. Otherwise, they will be out of the race.

However, there are many issues that could prevent open source software developers from doing this:

• Most of these AI models are vendor locked-in, meaning that they can only be used via an API and an Internet connection. It is not yet possible to run these AI models without having to let your data go through the vendor, leaving an advantage to these proprietary cloud solutions over the self-hosted open source ones.

• Huge resources are required to run the AI models anyways. Even if the models become available and fully open, it will take a long time before they can be bundled and distributed smoothly with the open source software to run locally on users’ machines.

• Many open source software developers are already struggling because of burnout and other financial issues related to their projects. It is not simply feasible to ask them to “just add more AI to your software, bro”.

GIMP for example didn’t even migrate to GTK4, and GTK5 is around the corner, so how will it focus its efforts on AI-related features when it has other core issues to solve?

On the contrary of that, some open source software succeeded in using the AI trends to their benefit. Blender, for example, has an addon for Stable Diffusion to generate textures or other objects from inside the software. But not all open source programs were lucky enough to adapt easily to the new changes.

There have been small attempts to create an open source version of ChatGPT and a LibreOffice integration of it, but these are still toy adventures at the moment:

It seems that if your open source program is already successful, then it can be more successful with AI. But if you were struggling, then you will have to struggle more now.

The rich get richer and the poor get poorer, all over again.

Large corporations like Microsoft, Google and Meta are definitely more advantaged in deploying AI models than various independent open source developers.

Perhaps this pushes the need for a large community discussion involving various open source stakeholders on how to approach this problem. The entire open source ecosystem needs to figure out a way to find a feasible method to utilize AI models for their needs in various sectors, and this can’t happen with various individuals here and there only.

Organizations such as the Linux Foundation, Open Source Initivate (OSI) and Free Software Foundation (FSF) need to take action as soon as possible on this matter.

Share with us your thoughts on this matter in the comments below.

Last Week Open Source News

We follow open source news from all over the Internet so you don’t have to.

• GNOME 44 was released with many new features. The most interesting of these is that the file chooser can now display thumbnails instead of the old list view. Various areas in the Control Center were enhanced as well.

• Curl 8.0.0 was released, celebrating 25 years of the project. Version 8.0.1 followed a few days later to fix a small regression.

• Docker Company has apologized for its last announcement about removing free open source organizations from the Docker registry. They have revoked the decision partially or completely in some areas.

Interesting Stuff from the Web

These discussions and articles might be interesting for you, all related to either Linux or open source software:

• The Framework laptop is now available in 13” Intel/AMD CPUs options as well as a 16” configuration coming soon.

• Twitter CEO says they will open source the recommendation algorithm code for Twitter on 31st of March, 2023. Could be interesting to see how Twitter actually recommends content for people and what factors it takes into account.

Hope you liked our newsletter! Kindly don’t forget to support our efforts on Patreon for just one cup of coffee per month ($5).