3 ms·
The data you enter in Google Search is much more limited (at least it is for me). None of the other Google services are essential. (Google Search isn’t essenti
by layer8 4y ago
The data you enter in Google Search is much more limited (at least it is for me). None of the other Google services are essential. (Google Search isn’t essential as well, but it’s a distinctive kind of service in that it isn’t practical to self-host.) With “WA” I guess you mean WhatsApp. For messaging, E2E encryption is now standard, which is not possible for AI services (barring huge breakthroughs in homomorphic encryption). In addition, it’s easy to self-host messaging solutions without losing out on features much.
- sebzim4500 4y agoSorry, I should have specified WA means Wolfram Alpha. >barring huge breakthroughs in homomorphic encryption Actually this seems plausible to me without having to use those general purpose algorithms that cause a 1000000000x slowdown. For example, you can obscure the input to each matrix multiplication by shifting the input and outputs by a predetermined (user specific) constant. Maybe there are similar tricks for cross attention, although that might be a bit optimistic. Certainly you can do some amount of obfuscation by permuting the inputs, but I doubt that would achieve much.
- layer8 4y agoI’m pretty sure your query will effectively still be derivable on the server side, as it has to syntactically/semantically interact with the model. To truly disguise it, at least part of the model/weights would have to be transferred to the client side (so that the server doesn’t learn the adjustment), which would in turn leak the model.