About six weeks ago, an email sent to Satia Nadila complaining of homogeneous architecture, which seems to be a silicon Valley, is imagined of artificial intelligence, as it contradicts the “structure of participation” that prompted the previous technological revolutions, most notably the Internet and open source programs. I thought that Satia may be sympathetic because of The previous talks we had When he wrote Update It was published in 2017.
I have been inserted to need the structure of the artificial intelligence industry that allows artificial intelligence cooperation, and this is not a winning market, and this does not make the current companies in every industry simply the colonial areas of the minerals of artificial intelligence, which seems to be the vision of Silicon Valley.
I didn’t know that Microsoft has already had something in the business and is evidence of what he hoped. It is called NLWB (Natural Language Network), and it is announced today. Satia O’Reilly offered the chance to be part of the operation, and we jumped on it.
Embrace the early stage of innovation
My thoughts are rooted in an idea on how technology markets develop. We lived through three times in computing. Each of them began the distribution innovation, and passed through a period of fierce competition, and ended with the guards of the monopolistic gate. In the first era, IBM was in Microsoft (PCS), and in the third (Internet and mobile) Google, Amazon, Meta and Apple.
The mistake that everyone makes is a rush to crowned the new player at the beginning of what is mainly an open field at the beginning of a new introductory market. They imagine that monopoly greatly as a substitute for what happened before, rather than realizing that the model has changed. When the personal computer challenged the IBM -based IBM monopoly, companies are racing to become the dominant personal computers company. Microsoft won because she realized that the program, not the devices, was the new source of the competitive feature.
The same story was repeated at the beginning of the Internet era. Netscape from Marc Andressen sought to replace Microsoft as a dominant software platform, except for the Internet instead of the computer. AOL realized that the content and society, not software, will be a source of competitive advantage on the Internet, but they made the same mistake in assuming the final game of unified monopoly instead of adopting the early stage of distributed innovation.
So we are here at the beginning of the fourth era, the era of artificial intelligence, and again, everyone rushes to the culmination of the new king. A lot of gossip is whether Openai or one of its competitors will be the next Google, when it seems to me that it is likely that the next Netscape or the following. Deepseek threw a bomb in the coronation procession, but we have not yet realized the depth of re -appointment, or perceived what comes after that. This is usually discovered during a period of distributed innovation.
We need the structure of participation for artificial intelligence
the term “Participation structure“It originally came to me as an explanation for the reason for the success of UNIX as a cooperative project despite its royal license while other projects failed despite the presence of open source licenses. UNIX was designed as a small operating parade in the Kernel operating system that supports small layers of auxiliary tools and applications that can come from anyone, as they follow the same rules. And how they can develop cooperatively.
This concept was subject to my web invitation in the early nineties, open source calls in the late 1990s, and Web 2.0 in A efforts. Participants are innovative markets. Unified markets prematurely, not much. The barriers that prevented entering the early computer market were very low, and entrepreneurship was high. Ditto web, Ditto for open source programs and sites 2.0. Late Silicon Valley, Installed on early monopoly via “Blitzscaling” (Think of Uber, Lyft and Wenge as examples, now Openai and HotHROPIC), not much. It has become A type of central planning. Small cadres of investors who enjoy deep spices choose the winners early and try to flood the competition with huge quantities of capital rather than allowing the experience and competition that allows the discovery of the real market’s suitability.
I don’t think we have the appropriate product market for Amnesty International so far. It is not suitable for the product market only. He also finds business models that pay the costs of these services, which create value for more than the central platform. The early unification problem is that it narrows the focus on the platform’s business model, and often at the expense of its ecosystem of developers.
Bill Gates is also famous Shamath said Balbaitia When he was running the emerging Facebook developer platform (and eventually failed), “This is not a platform. To clarify, this is not only the value of the final users. It is a value for developers and businessmen. This means that the opportunity to benefit from their innovations, and not to obtain this value that is harvested immediately by the dominant gatekeeper.
Now of course, Sam German talks about creating value for developers. In the appearance of a modern appearance at the Sequoia Capital’s Ai Ascent event, He said His hope is to create “just like an incredible amount of creating wealth in the world and other people to build on it.” But it uses the “operating system” language that others are above (and pushing Openai to use) instead of the joint infrastructure in which an ecosystem for developers participate.
For this reason I was rooting for something different. A world can provide specialized content providers for Amnesty International’s interfaces for their own content instead of absorbing it by AI’s builders who provide services based on their users. A world can developers can provide new types of services that enable others in a cooperative series.
We just started
The Hothropic’s model context protocol, is an open standard for the connection of artificial intelligence agents and their aides to data sources, the first step towards a vision centered on the protocol based on the collaborative AIS. I was born a lot of enthusiasm. Google takes this with a vision on how artificial intelligence agents cooperate. NLWB adds to this easy way to internet content sites to join the party, providing both the interface of the conversation to its content and the MCP server so that it is available to the agents.
All this will take years to correct. But since it revolves around the protocol rather than a vision centered around the basic system, solutions can come from everywhere, not only from a dominant monopoly.
Each new wave of computing also has a new user interface model. In the main computer era, it was the Telegype station; For personal computer, graphical user interface; For the Internet, the interface that focuses on the documents on the web; For mobile, touch screens. For artificial intelligence (at least at least), it appears to be conversation interfaces.
Companies such as Salesforce and Bret TayLor’s Sierra are betting on conversation agents that are front ends of companies, services and business operations, in the same way that their website or mobile phone application is today. Others are betting on agents on the part of the customer who will reach remote sites, but often by contacting application facades of application programming or even the screening of the screening of the screen. MCP, A2A and other agent protocols refer to a richer -a -cooperative reaction layer, capable of calling Any site Providing artificial intelligence services, not only via API calls to the dominant AI platform.
All companies need at least start on the front of Amnesty International today. There is a great line from the CS Lewis novel Even we have faces: “We cannot see the gods face to face until we have faces.” Currently, some companies can provide Amnesty International’s face to their users, but most of them do not do so. NLWB is an opportunity for every company to obtain the interface of Amnesty International (or simply “face”) not only for its human users but any robot chooses to visit it.
NLWB is fully compatible with MCP and current web sites provide a simple mechanism to add Amnesty International and other services to an existing web interface. We have collected our experimental search interface for O’reilly within a few days. We will rotate it to the audience soon.
