The People’s AI

Prompt: “A vast field on which the ground spells the letters A and I, with people on it, having a good time.” Via Copilot | Designer

People need their own AIs. Personally and collectively.

We won’t get them from Anthropic, Apple, Google, OpenAI, Meta, or Microsoft. Not even from Apple.

All those companies will want to provide AIaaS: AI as a Service, rather than AI that’s yours alone. Or ours, collectively.

The People’s AI can only come from people. Since it will be made of code, it will come from open-source developers working for all of us, and not just for their employers—even if those employers are companies listed above.*

That’s how we got Linux, Apache, MySQL, Python, and countless other open-source code bases on which the digital world is now built from the ground up. Our common ground is open-source code, standards, and protocols.

The sum of business that happens atop that common ground is incalculably vast. It also owes to what we first started calling because effects twenty years ago at Bloggercon. That was when people were making a lot more money because of blogging than with blogging.

Right after that it also became clear that most of the money being made in the whole tech world was because of open-source code, standards, and protocols, rather than with them. (I wrote more about it here, here, and here.)

So, thanks to because effects, the most leveraged investments anyone can make today will be in developing open source code for The People’s AI.

That’s the AI each of us will have for our own, and that we can use both by ourselves and together as communities.

Those because investments will pay off on the with side as lavishly as investments in TCP/IP, HTTP, Linux, and countless other open-source efforts have delivered across the last three decades.

Only now they’ll pay off a lot faster. For all of us.


*See what I wrote for Linux Journal in 2006 about how IBM got clueful about paying kernel developers to work for the whole world and not just one company.



4 responses to “The People’s AI”

  1. […] (version 4o, which I pay for) to give me an illustration to use in my last post here, titled The People’s AI. But don’t go there yet (if you haven’t already). What I ended up using there is a […]

  2. There already is personal AI: GPT4All (https://gpt4all.io/index.html). There may be others; this one is relatively easy to use. Through my company, Overlogix, I am developing this ecosystem first for personal use; practical business use will come next. See https://www.linkedin.com/pulse/overlogix-master-index-overlogix-69ivf/ for details. I found it delightful and surprising that your article called for exactly what I’m doing. You’ll need a HOT CPU/GPU and plenty (min 32GB) of memory. The WizardLM model is recommended.

  3. The problem will come when you need to decide if any of the closed source tools that you use, like the browser, will be able to access your personal AI. See this https://github.com/TomCJones/threat-modeling/blob/main/models/ai-in-browser.md

    1. Thanks, Thomas. And yes, that problem will come. As will many others. I just spent the last two hours on the topic of the OSI and the Open Source Definition for AI. (Our side—Kwaai’s side) is Sam Johnston’s, here: https://samjohnston.org/articles/

Leave a Reply to Phil Otken Cancel reply

Your email address will not be published. Required fields are marked *