Hacker News Reader

Comment by 01100011 | Hacker News Reader

If you want people like me to buy it, you've got to cut the cloud out of the picture. Give me something that does inference on the device. Sure, give me an easy way to send selected videos back to you for training data, and feel free to push optional updates to the camera with updated models, but I'm not giving you a raw video feed of my home or business.

View on HN

The reason we went with the cloud is not just to use the data for training data. 1- we wanted to be hardware agnostic so our solution works with any hardware without any dependency on hardware specifications. 2- Doing deep learning on the device would require GPU which would significantly add to the initial cost for consumers. 3- as I mentioned in my response to another comment, regardless of where the inference is done, the recorded clips will have to be stored on the cloud so if a bad actor comes in and take the camera with them, the user can still access the recorded video clips after the fact to know what happened. Even though we do inference in the cloud, we do not store the video clips unless they correspond to an event that the user is interested in--we discard the other clips. So I'm not clear on how doing inference on the device would have any advantage from a privacy stand point.

Not to say you shouldn't move forward with your grand plans, but there are lot of people, myself included, who won't touch any of this for those reasons.

I don't need more of my life stored on someone elses network.

I hear you and understand the concern and appreciate the feedback. What do you use the cameras for if you don't mind me asking?

I get it. Moving things to the cloud keeps the devices cheap and keeps them from becoming out of date. It also makes you identical to other large players and makes me think you don't have much of an advantage. Make the devices expensive. Put privacy first. It might just pay off with higher sales. Your alternative is racing to the bottom in a competition with Amazon, Ring, etc. You will lose.

Sure, you can push the video to the cloud, but encrypt it on the device. Let me control my data.

Based on the feedback from people's comments here, I realize we should do more to alleviate the privacy concerns. Curious on to know your thoughts about the following aspect: As I mentioned, we only store short video clips corresponding to events that the user created (we already discard the other motion clips that are deemed as irrelevant by our models.) We also allow users to delete the the alerts they have received and when they delete each alert, we permanently delete the corresponding video clips from our dbs... does that help alleviate your concerns?

I just want to say that it's good that you're listening to people here, but at the end of the day you have to go with what's pragmatic for your business.

Not really. Having a device send video frames to the cloud basically kills the deal for me. If you did something to the video first... maybe. Let's say you train a model to process video. You come up with a network where the first few layers are fixed and they perform a non-reversible transform of the video into some sort of symbolic representation. I might sign up for that. I want to know that my video/audio feed isn't being used by a 3rd party, and that it isn't being used by you for something other than the intended purpose(i.e. to model my behavior to sell me things, to create profiles of my activity, etc). I understand that my first proposal doesn't actually address all of those concerns, but I am willing to trade privacy for convenience to a very limited degree.

I'm basically working on something like this for myself. I have a Nvidia Jetson Nano that I'm trying to train to tell me when my garage door is open without my wife or I present, when my laundry is done, and whether or not the lights are on.

That is pretty interesting but non-trivial. We'll have to think about it. Thanks for the suggestion. I should mention our business model is not based on monetizing users data. I'm interested to know more about your personal project. Feel free to shoot me an email if you are interested in having offline discussions and bouncing ideas off each other: rafiee at visualone dot tech

Just because that’s your model today, that doesn’t mean you won’t get bought by another company whose sole purpose is to use all assets for gathering as much private data as possible and using it for their own purposes.

If you want to sell to people who actually care about privacy, then you have to build your systems so that it is impossible to abuse them or use them in any other way. Or, at the very least, it is extremely difficult to abuse them.

That is a valid point. We will think more deeply about these. Thanks for the feedback.

Reply on news.ycombinator.com