Big Idea: Solid Architecture to Overcome The Glaringly Bad AI Privacy Issues

While we fight about whether AI is helping workers or destroying jobs, not many of us consider a major hurdle AI has yet to overcome: privacy. This article dives into what the issue is, and what we can do about it.

First I want to talk about the problem.

If you’re more than aware of the problem, feel free to skip down to the solution (sorry, Medium doesn’t have in-page links, otherwise I’d provide one).

And a note, I’ll default to talking about Large Language Models for this article for the sake of brevity, but these issues (and the solutions) can be applied to other AI models as well: text-to-image, image-to-text, text-to-audio, and so on.

The Internet and Privacy

It’s now a running joke (rather, a reminder we tell friends to which they nervously laugh) that nothing we do online is private.

Believe it or not, there was once a time when this was a common belief.