Centralizing and Decentralizing Forces - The Innovation of Enterprise Technology
Enterprise technology has long been defined by a series of centralizing advancements and decentralizing innovations. Large firms, driven by capital growth, develop new devices and frameworks to create a more universal technological landscape. Inversely, disruptors work to extract value from those advancements, seeking to improve the availability and utility of these technologies to better serve individual users. This push and pull dynamic of innovation has led to massive leaps in technological advancement throughout the past few decades, creating systems that continue to influence our use of technology to this day.
Mainframes to PCs: Decentralizing Hardware
The first major centralizing innovation came in 1964, with IBM’s creation of the first modern mainframe. Prior to this, computing was fragmented and highly localized — every machine operated as a standalone system requiring its own bespoke programming. Because there were no commercial software development companies, there was a massive entry barrier, preventing many from accessing computer technology. Mainframes were a revolutionary step forward, enabling firms to store all their data and software on a centralized piece of hardware, accessible to any terminal connected to it. Enterprises could now have multiple users work with the same data simultaneously and had a simple path to deploying their proprietary software at scale.
Mainframes were not without their flaws, however. Their massive size and high cost made them impractical for use outside of large corporations, and their centralized nature posed massive concentration risks — if the mainframe went down, every terminal it hosted went down. As computing technology improved however, smaller processors were created that could run individual devices, leading to the first PCs. This innovation allowed individuals and small businesses to make use of computing power without a mainframe, creating entirely new use cases outside of commercial enterprises. As PCs grew in popularity, the hardware necessary to access computing power became increasingly decentralized, spreading the technology beyond its commercial roots.
Cloud Computing and the Democratization of Software
The advent of cloud computing helped recentralize the control of technology back into the hands of enterprises. It offered a flexible infrastructure, allowing computers to access data and software remotely over the internet for the first time. This opened new opportunities for enterprises to sell digital services and host computing resources on their own servers. As cloud computing infrastructures like Amazon Web Services (AWS) grew, so did the availability of computing technology. This consolidated control back into the hands of a select few tech giants, as individuals and small businesses were forced to rely on these platforms. This structure exposes cloud computing infrastructures to the same concentration risks as mainframes, exemplified by AWS’ outage just this week — a single configuration issue caused outages for thousands of services.
Despite these downsides, cloud computing was crucial in improving the accessibility of new software, leading to a myriad of products that prioritized the cloud as a means of accessing applications. Devices like the Palm Treo and the Blackberry offered users new tools in a convenient package. Apple’s launch of the App Store, alongside their iPhone3G, was the most influential of these developments. For the first time, users were no longer restricted to software produced by their phone manufacturer, gaining access to an expansive library of mobile-specific, third-party apps. This new decentralized ecosystem not only expanded the utility of the iPhone but also helped monetize third-party developers, further democratizing the availability and use of software as competitors were forced to implement similar frameworks to compete with the Apple.
AI: Flexibility vs. Specificity
This brings us to today, with the rise of AI defining technological innovation over the last few years. Currently, the market is dominated by established tech companies creating massive LLMs and generative AI models. These models have been used for every use case imaginable, as developers seek to centralize the use and output of AI software to their own environments. The massive amounts of data used to train these models help it perform a wide variety of tasks, but it rarely provides the context needed to produce effective, personalized solutions to users’ problems. They too are vulnerable to centralization risks, with countless organizations relying on the stability of a single firm’s infrastructure. Like the previously discussed innovations, these advancements offer great opportunity, but in their current form fall short of providing real value on an individual basis.
In reaction to this, many firms have pivoted to implementing smaller, highly specialized models built for more explicit functions. They are designed to understand and meet the specific needs of the user, trained on their data to avoid the generic output and hallucination issues large models face. Custom models apply the power of AI on a user-centric scale, opting for specificity over an idealized goal of limitless functionality. In the coming years, large corporations will realize the benefits of these solutions and be pushed into a more granular information economy — one based on buying or renting an individual’s private data to train specialist, local AI models that prioritize user interest above all else.
From Localization to Personalization
The advancement of technology is defined by the ebb and flow of these opposing forces of innovation. So long as there are attempts to centralize technology and people’s use of it, disruptors will find ways to extract even greater individual value. Just as the mainframe led to PCs, and the cloud to the endless stream of applications we use daily, large artificial intelligence models will lead to local, custom models focused on providing personalized value for each user.
How soon the big players will begin creating this type of model is unknown, as they continue to hone their efforts on achieving AGI (artificial general intelligence). However, if past innovations provide any indication, the most influential developments are those that seek to adapt technology to meet the needs of each user. Controlling your own destiny has become increasingly difficult in the complex technological landscape of today, but it is not impossible. The path forward is paved by your own custom models running on your own infrastructure, not shackling yourself to the interdependent web of frameworks that large AI companies rely on.
This article was written by Sultan Meghji, CEO of Frontier Foundry. Visit his LinkedIn here.
To stay up to date with our work, visit our website, or follow us on LinkedIn, X, and Bluesky. To learn more about the services we offer, please visit our product page.