I have tens of thousands of photos, it’d be nice to search “[daughter’s name] holding blue teddy bear” and have it come back with the exact picture I have in mind but would struggle to find.
Is it worth the privacy and potential security issues, though? I’d say hell no.
I look forward to the future where we have powerful, energy efficient neural processing hardware, and a robust open source software ecosystem to do this in a more trustable way offline and on-device, or on a Nextcloud home server or something like that.
I don’t understand why Google is pushing cloud AI so much after pushing their Tensor AI chips for so long. iPhones have indexed pictures offline for years but Google insists on doing this tagging in the cloud.
I have a copy of Photoprism running that does some (not great) indexing of my pictures, and that runs on half a CPU core on an old laptop. Surely all of these fancy Tensor processors and thousand dollar phones can do that stuff while they’re charging overnight?
That’s not entirely true. They may have the same processor but they don’t have the same amount of RAM, which is actually super critical for on device AI tasks. They recently brought all the features to the Pixel 8, but it took additional time to optimize for 8 GB vs the 12 GB of the Pro.
I can see the utility in this.
I have tens of thousands of photos, it’d be nice to search “[daughter’s name] holding blue teddy bear” and have it come back with the exact picture I have in mind but would struggle to find.
Is it worth the privacy and potential security issues, though? I’d say hell no.
I look forward to the future where we have powerful, energy efficient neural processing hardware, and a robust open source software ecosystem to do this in a more trustable way offline and on-device, or on a Nextcloud home server or something like that.
I don’t understand why Google is pushing cloud AI so much after pushing their Tensor AI chips for so long. iPhones have indexed pictures offline for years but Google insists on doing this tagging in the cloud.
I have a copy of Photoprism running that does some (not great) indexing of my pictures, and that runs on half a CPU core on an old laptop. Surely all of these fancy Tensor processors and thousand dollar phones can do that stuff while they’re charging overnight?
Money. The latest Pixel devices all have the same processors although the features are locked behind software depending on which model you got.
That’s not entirely true. They may have the same processor but they don’t have the same amount of RAM, which is actually super critical for on device AI tasks. They recently brought all the features to the Pixel 8, but it took additional time to optimize for 8 GB vs the 12 GB of the Pro.