Trends in IT that Open Source Advocates Should Address, Part 1

The world of computing used to be rich with opportunities for mavericks and the marginalized, as celebrated in books such as Hackers: Heroes of the Computer Revolution and Broad Band: The Untold Story of the Women Who Made the Internet. But over the past decade, several trends have threatened this freedom:

• The centralization of advanced computing, which includes AI, powerful new chips, and quantum computing
• Proprietary control over technology, notably through Software as a Service
• National fragmentation of computing and the Internet

What does this mean for the future of innovation and the freedom that individuals, communities, and countries have in our digital environment? This two-part series looks at these trends and how the open source movement can act in their wake.

Centralization of Advanced Computing

Intense concentration in major industries tends to lead to abuse, as can be seen in railroads during the nineteenth century, and later in the energy, telecom, and agribusiness sectors. Various companies in the computer field are now the largest in the world, and the public is concerned about what they do next.

In the golden age of computing recorded in books like Hackers, amateurs and tinkerers could perform the same basic functions as programmers at the most advanced software houses, albeit perhaps on systems with less memory and processing power. But now, the most cutting-edge computing technologies require huge corporate investment: AI, powerful chips, and quantum computing.

Hacker Don Marti, in his review of this article, pointed out that most programmers in the 1960s and 1970s used time-sharing to get periods of programming time on remote computers. That trend is echoed by IaaS and PaaS (cloud computing) today, and might have comparable markets in the future for the technologies highlighted in this series.

Artificial Intelligence

Once researchers discovered the modern path to artificial intelligence through neural networks, AI has thrived on bigger and bigger data sets, along with almost unfathomable quantities of processing power.

It’s inevitable, then, that the market for LLMs is highly concentrated. Only two LLM services have more than 10% of the consumer-level market, and more than half of requests go to ChatGPT. The market is also skewed by country, with U.S. companies having 80% to 90% of requests and most others going to China or (surprisingly) France. Enterprise LLM usage is less concentrated, with Anthropic at 32%, OpenAI at 25%, and Google at 20%.

But we need AI, and its value is undeniable. Some trends hint at limits to its growth, including a famous call for a slowdown by leaders of the major companies, bipartisan opposition to building data centers, and the ominous compute-efficient frontier.

Still, AI won’t go away. We will learn over time how to manage the models’ frequent errors, their biases, and their dangerous vulnerabilities. Users will be trained in what AI is good for, and how to use it properly.

Many tools for developing AI models are open source, and open source is making strong advances in the marketplace, especially in developing areas such as Africa. AI companies champion open-source AI tools as a force for innovation, democratization, and even (according to Meta CEO Mark Zuckerberg) “individual empowerment“.

However, malicious actors of all types have gotten access to AI anyhow, either by building their own models or by circumventing the guard rails on popular public services. Most of the tools used in chatbots are open source software, so maybe it’s time to release the code for the bots as well. The computational power and data storage required to handle a query will still limit who can offer an AI service.

AI has a fundamental drawback: It’s opaque by definition. Even though AI developers like to throw around the words “explainability” and “transparency,” there is still a machine determining how to interpret a request, what resources to retrieve, and how to organize them. LLMs’ output surprises the developers themselves.

Thus, open source developers can point to AI to demonstrate the contrast with the true transparency of free and open source software, developing support for its products. Marti points out that open source developers can also pick and choose the best LLM for their task and use tools that interact with LLMs in a standardized way that makes switching convenient.

Chips

Chip development has been concentrated for a long time in a few companies such as Intel. There is more variety in the newer AI-specific chips, but development is still dominated by familiar companies, led by NVIDIA. Stats list the big players in computing: besides NVIDIA, there are AMD, Apple, Google, and Intel, joined by Amazon, Microsoft, Meta, Qualcomm, Broadcom, and some lesser-known players.

DeepSeek was notable for releasing its software as open source, but an article with the somewhat provocative title “DeepSeek’s Chip Push Shows AI Freedom is Ending” claims that they are becoming more proprietary.

Luckily, everyday chips are becoming more powerful too. The open RISC-V standard and other open source hardware could be the basis for cheap computer systems running free software to do more and more for everyday users.

Quantum Computing

Finally, we have to consider the potential for quantum computing. With costs ranging from $10,000 to $50,000 per qubit, a complete quantum computer costs millions of dollars. Small organizations can lease time on quantum computers just as they do on conventional cloud services, but the hardware itself is centralized.

But quantum computing is predicted to be appropriate for a limited set of applications. It will be used for certain complex tasks that can’t be solved in a reasonable time frame by conventional digital computers. Open source software can still address the common needs of everyday computer users.

The second and final part of this article examines the implications of proprietary control and national fragmentation.