I was thinking about this the other day.
I can’t see the point, desktop search is a solved problem.
Yes Windows find is shit, but it doesn’t have to be.
I know a bunch of the conspiracies, but what is the legitimate use case?
I was thinking about this the other day.
I can’t see the point, desktop search is a solved problem.
Yes Windows find is shit, but it doesn’t have to be.
I know a bunch of the conspiracies, but what is the legitimate use case?
Because it’s not just desktop search in the conventional sense, it parses and indexes EVERYTHING that is shown on screen regardless of app.
Normal desktop search is fine for ‘where is file X?’ or ‘which document contains Y?’ but the point of Recall is to be able to deal with far more wide reaching queries, like ‘what website did I order a cake from last week?’ or ‘where did I see a great idea for a pizza recently?’, where it can find sources for natural-language queries across anything that was shown onscreen, whether it was a website, an email, a slack message, or something within any other app.
Ok, so my next question would be, is this a problem that needs solving?
At a long time Linux user, I’m wondering if it is useful enough to look into an open and trustworthy solution to give this a go.
I could see the usefulness of saying, “hey I saw this on my phone a few days ago, which site was it?”
Its a double-edged sword. There is a ton of use in something like “help me display the data in this excel workbook for a presentation” but the tradeoff is that data is now indexed in the recall database along with everything else you do.
Ultimately, using recall is a security concern that massively outweighs any benefits it offers, but to say it offers no benefit at all is incorrect.
By default, the security concerns do massively outweigh the benefits, to me. On my work machine the benefits outweigh the security issues, because it’s tightly locked down.
But if we can figure out how to safely implement such a system with very tight controls, it will be a huge game changer for individuals.
You’re kidding, right?
MS did some research in the 90’s (related to My Life Bits) and determined how much data a person engaged with during an average day, and even then it was more than anyone could hope to manage.
This is a common question on new tech things lately. Unfortunately the answer has been “no” almost invariably.
To answer my own pondering. We could feed your browsing history to a local LLM, it could fetch a synopsis of the site and be able to answer that question very easily.
This wouldn’t require a really powerful AI model, combine this with desktop search and you have close to what MS is offering.