3 ms·

The blog articulates a real pain point that I am also trying to solve: "Anonymity, security, verifiable communication, consent, and portability. These are the p

by yonz 5y ago

The blog articulates a real pain point that I am also trying to solve: "Anonymity, security, verifiable communication, consent, and portability. These are the pillars of a truly open internet. A full end-to-end technology stack requires identity, authentication, messaging, and personal data storage" Drop a line if you know any other players in the space, so far I am tracking https://fission.codes/ https://fission.codes/ https://inrupt.com/ https://inrupt.com/ https://0data.app/ https://0data.app/ https://ceramic.network/ https://ceramic.network/ https://www.usevana.com/ https://www.usevana.com/ All having to do with creating a user-centric instead of product centric internet storage and authentication model.

nl 5y ago
Oh hey. Feels kind of weird having our whitepaper shared. Happy to answer any questions!
- yonz 5y ago
  I was surprised it wasn't here already. My biggest question is around doing personalization at the edge, it is a complicated idea but I would love to hear if you have some thoughts on it.
  - nl 5y ago
    The short answer is that yes, it is complicated but there is plenty of room to do more than is done now. Edge capability is improving significantly. My background is in AI, and I've been pretty impressed with what is possible now in desktop web browsers (eg https://twitter.com/nlothian/status/1507239300404228098 https://twitter.com/nlothian/status/1507239300404228098). Tensorflow.js is pretty good now, and importantly has a set of accelerated primitives on par with numpy in Python which makes it possible to build your own models (including non-NN based models). I think there are good prospects of being able to get decent results using techniques like sending trained models to the client and the client running that model against local data to generate recommendations. This could work well for things like ecommerce.

by yonz 5y ago

nl 5y ago

Oh hey. Feels kind of weird having our whitepaper shared. Happy to answer any questions!

yonz 5y ago
I was surprised it wasn't here already. My biggest question is around doing personalization at the edge, it is a complicated idea but I would love to hear if you have some thoughts on it.
- nl 5y ago
  The short answer is that yes, it is complicated but there is plenty of room to do more than is done now. Edge capability is improving significantly. My background is in AI, and I've been pretty impressed with what is possible now in desktop web browsers (eg https://twitter.com/nlothian/status/1507239300404228098 https://twitter.com/nlothian/status/1507239300404228098). Tensorflow.js is pretty good now, and importantly has a set of accelerated primitives on par with numpy in Python which makes it possible to build your own models (including non-NN based models). I think there are good prospects of being able to get decent results using techniques like sending trained models to the client and the client running that model against local data to generate recommendations. This could work well for things like ecommerce.