Live data from Hacker News

How did Roomba-recorded photos end up on Facebook?

technologyreview.com

31–40 of 148 posts

Re: How did Roomba-recorded photos end up on Facebook?

#31
post #22

Earlier quoted context omitted.

The part that's missing is that it isn't just "a Roomba", it's "a Roomba labelled with “video recording in progress”". Saying 'a Roomba' implies that it could've been done from any Roomba.

The article addresses this head on. Even though those users once consented to share data “It’s not expected that human beings are going to be reviewing the raw footage.” The meat of the article is that what technology and tech companies are doing is divorced from the expectations that we have as a society. It couldn’t have been done from any roomba, but it could happen to almost everyone who didn’t understand the exa…

how did we get to the point where people have Internet-connected cameras and microphones everywhere in their houses yet implicitly trust that anything recorded will never under any circumstances be viewed by another human being

Re: How did Roomba-recorded photos end up on Facebook?

#32

Key paragraph for our friends who don’t RTFA > iRobot … confirmed that these images were captured by its Roombas in 2020. All of them came from “special development robots with hardware and software modifications that are not and never were present on iRobot consumer products for purchase,” the company said in a statement. They were given to “paid collectors and employees” who signed written agreements acknowledging…

> our friends who don’t RTFA

I opened the article (on a phone) and no fewer than 3 separate popovers appeared over the content. “Hey! This is our cookie policy” “Happy holidays! We have a special subscription price!” And something else that was covered by the first two before I had a chance to read it.

Thankyou for summarising. I noped right out of there out of disgust.

Re: How did Roomba-recorded photos end up on Facebook?

#33
post #25

The question I want to ask is: how did Roomba-recorded photos end up in a major publication on the Internet? And the sequence seems to be: 1. iRobot hires people to use special development versions of the Roomba in their homes to collect training data. These are clearly labeled, and the participants are informed that the images are being sent to iRobot for training. This seems fine - if you want to exchange some degr…

> A contractor posts some of these photos to a private Facebook group used by other contractors on the project.

That's where it went wrong. Everything else seems reasonable for a visual AI training project, well signalled to the participating users and the data securely communicated.

Thereafter, the data was mismanaged.

There is clearly no such things as a "private" Facebook group. So called "contractors" [1] using a disservice like Facebook to communicate beggars belief.

[1] people with the unremarkable skill of being able to spot ordinary household objects and label them - so someone probably had the bright idea of creating a CAPTCHA "Find all the women on toilets".

Re: How did Roomba-recorded photos end up on Facebook?

#35
post #22

Earlier quoted context omitted.

The article addresses this head on. Even though those users once consented to share data “It’s not expected that human beings are going to be reviewing the raw footage.” The meat of the article is that what technology and tech companies are doing is divorced from the expectations that we have as a society. It couldn’t have been done from any roomba, but it could happen to almost everyone who didn’t understand the exa…

how did we get to the point where people have Internet-connected cameras and microphones everywhere in their houses yet implicitly trust that anything recorded will never under any circumstances be viewed by another human being

I think the second part of your sentence is a least a partial answer to your first part.

Re: How did Roomba-recorded photos end up on Facebook?

#36

Key paragraph for our friends who don’t RTFA > iRobot … confirmed that these images were captured by its Roombas in 2020. All of them came from “special development robots with hardware and software modifications that are not and never were present on iRobot consumer products for purchase,” the company said in a statement. They were given to “paid collectors and employees” who signed written agreements acknowledging…

Here's a funny bit.

Roomba iOS app refuses to go past its welcome screen unless its granted access to the location info.

This is unreasonable, they don't need this info for their app to function.

However their devices are all but unusable without an app, so they ultimately blackmail people into giving location data to them.

Meaning they don't really give a sh#t about users' privacy, so it's not that "they are watching you", but that they won't think twice about hooking up to a random Roomba and shooting a video with it. Consent or not.

Re: How did Roomba-recorded photos end up on Facebook?

#37

Yeah, I’m not sure I buy their explanation about special development roombas since they offered zero proof. I have an ancient Roomba that’s still going and dread having to replace it one day.

>Yeah, I’m not sure I buy their explanation about special development roombas since they offered zero proof.

It is simple for any person with even basic knowledge of networking to independently come to the conclusion that Roombas are not uploading video streams (or photographs) to the internet.

I know the IP (10.0.0.11) and MAC (50:14:79:1E:AB:6B) address of my Roomba and using the Insight Netflow Analyzer for OPNsense I can see how much data it has sent to the internet. In the last six months it has sent approximately 72MB of data outside my network. That's about 600KB per day.

It has received much more, presumably firmware downloads.

This is consistent with firmware update checks, notification traffic, and me periodically adjusting its schedule remotely.

That's just me clicking on some tabs in my router's web UI. Hundreds if not thousands of people globally are constantly reviewing and monitoring Roomba network traffic in fine detail in order to understand and/or reverse engineer it for research and other purposes.

So one of three things is happening:

1. All Roombas send photo and video streams to iRobot and they have thus far managed to hide this from the public and the thousands of eyeballs constantly monitoring the network traffic of their products, or

2. A subset of Roombas send photo and video streams to iRobot and they have thus far managed to hide this from the public and the subset of eyeballs monitoring the network traffic of their products

3. These are development devices like they claim.

Based on my own experience we can eliminate 1, based on the images accompanying the article option 3 is highly likely.

Re: How did Roomba-recorded photos end up on Facebook?

#38

Yeah, I’m not sure I buy their explanation about special development roombas since they offered zero proof. I have an ancient Roomba that’s still going and dread having to replace it one day.

You're operating in a framework where you already think they are lying. What proof would you even deem acceptable?

Re: How did Roomba-recorded photos end up on Facebook?

#39
post #21

Key paragraph for our friends who don’t RTFA > iRobot … confirmed that these images were captured by its Roombas in 2020. All of them came from “special development robots with hardware and software modifications that are not and never were present on iRobot consumer products for purchase,” the company said in a statement. They were given to “paid collectors and employees” who signed written agreements acknowledging…

Everyone RTFA in this case! I did not expect to see actual photos of the woman sitting on the toilet in this article. But damn, they're real and published dead center. It's awful and voyeuristic to feature, but in a way it brings to life the freakishly perverse Orwellian horror of all of this. This piece hits hard, as it should. How did neither Roomba nor ScaleAI have safeguards against PII of this nature? This is in…

If you read the article you'll also find the owners were specifically aware that these units are special in that they upload all data - commercial devices do not share any images or video without the users consent [0].

It's also absurd to think they didn't face safeguards. We can only speculate if the individual was fired, or if stronger policies were put in place since 2020, but it's naive to expect that whatever policy is put in place will stop a human data labeler from smuggling PPI for personal reasons.

[0]: https://homesupport.irobot.com/s/article/964

Re: How did Roomba-recorded photos end up on Facebook?

#40
post #21

Earlier quoted context omitted.

Everyone RTFA in this case! I did not expect to see actual photos of the woman sitting on the toilet in this article. But damn, they're real and published dead center. It's awful and voyeuristic to feature, but in a way it brings to life the freakishly perverse Orwellian horror of all of this. This piece hits hard, as it should. How did neither Roomba nor ScaleAI have safeguards against PII of this nature? This is in…

If you read the article you'll also find the owners were specifically aware that these units are special in that they upload all data - commercial devices do not share any images or video without the users consent [0]. It's also absurd to think they didn't face safeguards. We can only speculate if the individual was fired, or if stronger policies were put in place since 2020, but it's naive to expect that whatever po…

The owner who said ok probably was not the girl with her pants down.

But as a guy, I don't know if girls pull them down farther if they think they're not being watched. She might be the one who agreed to it.

Still, the owner of a space is making the decision for all people who come into that space.

Post reply on HN