BLUF: This question refers specifically to Wyze Cam videos recorded and stored on Wyze servers when designated for Cam Plus services.
If users submit feedback correcting an inaccurate designation made by the AI recognition algorithms (detected Pet when no animal was present), does that feedback actually get used to refine the accuracy of the algorithms?
If so, how quickly should we expect to start seeing improvements for those reported errors we see frequently made (7-9 times per night), & were caused by the same algorithmic inaccuracies?
WHY I ASKED THOSE QUESTIONS - THE CIRCUITOUS JOURNEY THROUGH THE MIND OF A VERBOSE LITERARY EXTROVERT
INTRODUCTION
Once I finished this post, I read what had just flowed out of my mind unfettered to see if had been able to type fast enough to keep up with the deluge of poetry below. I realized that while I had set out to ask a pretty simple question, one which most would have been capable of keeping concise, dry, and a bore to read, as is regularly the case in all my musings, I was not capable because I was not willing to do that.
Now, before you allow the thought that this unwillingness was driven by some sort of self-serving motives or a lack of consideration I have for others, hear me out. The reality is, I’ve sat front row and watched in very gruesome clarity and detail the utter catastrophic destruction and brutal dismantling of an individual’s self worth, successful & beloved career, financial security, and social & professional network, none of which they deserved or could control. While there were a confluence of other factors which contributed to this incredible betrayal of someone by those who had benefited from this person’s work, the first dominos were unbelievably allowed to fall and trigger a cascade beyond anyone’s ability to put Humpty Dumpty back together again.
What seemed like a minor and irrelevant factor, but would become one of the few key factors which may have been enough to prevent at least some, but potentially all of the damage inflicted related to a number of critically important elements of relevant context related to the narrative had been haphazardly removed from the conversation from the very start. This was not likely done with deliberate intent, rather by someone who felt that the brevity they had been told to use in other unrelated aspects of their job somehow translated into being given the authority to filter out what they felt was extraneous and not germane to the topic. I can not say it any clearer than this person was wrong and someone lost their career over it.
I don’t presume to know what details will be considered important or relevant by someone I’m talking or writing to, and will be damned if I don’t include every bit of contextual detail and minutia. To do otherwise would be disrespectful to those who take the time to read by ramblings and is actually an effective means to both eliminate a significant amount of underlying biases or at least make them glaringly obvious to the reader.
With that out of the way, read on for all the supporting context which lead to my questions in the BLUF (defined below), just in case some little detail will have implications for the answers or guidance.
HOW THE REPORTED ISSUES MANIFESTED
So, I can get all “Zuckerberged” (defined below) into a repetitive task pretty easily, especially if I know that whatever that task is accomplishing is having some positive impact on someone or something I feel is worth the effort.
I have a WyzeCam v2 mounted outside displaying my back deck and sliding door into the house. To comply with the current version of house layered security program directives, something I talk about as if I was head of security at the Pentagon, the WyzeCams designated as and mounted to cover any of the house’s exterior entrance doors from the outside, it is Cam Plus configured to ensure video records of all activity that occurred at those entrances is captured and saved in it’s entirety for future review if needed.
During the spring and summer months of the year, the need to periodically remove the cobwebs and debris which accumulates throughout the nights when the wildlife decides to use the front of the cameras as a handy landing spot or anchor point for that night’s web. Because I travel so frequently, if I forget to evict these squatters before the trip or something gets caught in a web obstructing it’s view after I leave, areas of the camera’s video display at night show a blurry wiggling something.
I was looking at this week’s recordings to see if any evil-doers were trying to place paper bags filled with a collection of feces piles prosuces by the many neighborhood dogs near the entrances, ignite the paper bags while ringing the doorbell, and fleeing the area to ensure they can avoid being held accountable. I’m thrilled to report that as with all of the footage I’ve reviewed over the last few years, this threat has yet to manifest in that way. However, I did realize that the blurry wiggling spider web had been present, obstructing a large, yet no critical area of the coverage, but there was one odd thing that didn’t seem right, and that’s what I’ve called you all here to discuss.
HOUSTON, WE HAVE A PROBLEM
The Wyze Cam motion sensing logic was being triggered by the blurry wiggling, which is part of why I like to keep them brushed off, and is to be expected. However, the Cam Plus AI recognition system was erroneously interpreting the blurry wiggling as a “Pet”, and labeling it as such. I did ponder as to the possibility of the AI really being so on point that it figures some people do keep large spiders as pets, and who is it to discriminate against certain pets just because most people have cats, dogs, and naked mole rats. But further analysis of these videos, many 5:00 minutes long, and there are actually no creatures or wild beast visible, just blurry wiggling.
THE TIME & EFFORTS IN QUESTION
As one who prefers to be a part of the solution and ridicule who is being marginalized as the problem, I set forth to go down every nights Cam Plus recordings captured by this specific camera from the previous week which had the blurry wiggling, load the video, wait for the question to pop up asking if I wanted to help improve the AI’s accuracy, press “Heck Yeah, Buddy”, deselect the erroneous “Pet” botton, select the “Nothing” button (not sure why “Blurry Wiggling” wasn’t an option, but whatever’s), and selecting “Send” to kick it over to whoever screwed up at Wyze HQ that week and was being assigned to review all these shared user videos as a punishment.
Whilst this is a simple, mindless task which doesn’t take a whole lot of cognitive horsepower to complete, the time it takes from first selecting the video, running through that process, and closing the video and be sent back to the chronological list can add up quickly, especially when one becomes Zuckerberged by it. That said, if the time and efforts I’m spending to correct the AI’s recognition are actually being ingested by some machine learning Terminator T-1000 or there actually is some poor intern who is doing it by hand, and the AI logic algorithms are being perpetually or periodically refined for more accuracy, these investments of my time and efforts are being put to good use and I’m more than happy to continue doing so.
THE UGLY ALTERNATE UNIVERSE
However, there is a worry that these corrections I’m submitting are being received by a server which directs them to a “To Be Reviewed” database filled with 500,000+ other correction submissions recieved by a few thousand other customers. If this is happening, you guys (gender nonspecific) are very likely running the same line as my horder dad by insisting that you’ll eventually get to them but you’re just busy right now. This is an intervention, you may genuinely want and intend to get to them, but you won’t and we need you come to terms with that, be honest with us, and just come clean. We won’t be mad, we don’t have to be bothered by distracting emotions because we drink whenever they start to bubble up. Don’t worry, we’ll just drink!
THE ORIGIN OF THE BLUF CONCEPT
I’m going to do the right thing and provide what’s called a BLUF way up at the top of this post. That’s a tool used by the intelligence community which stands for Button Line Up Front.
Not everyone is a fan of the long form story arc style with building narratives and exciting finales at the end to pull all the story lines together and making for an interesting read that I LOVE to use (if you didn’t figure that out by this point, we should have a reading comprehension clinic soon). One of those groups is the US intelligence community (IC) who almost always use a BLUF to provide the reader with a summarized conclusion or assessment with just key elements included as the first few lines, usually no longer than a couple paragraphs.
Then, the rest of the report will provide the granular details discussing the information and sources which were used, details of those resources, and essentially provide a narrative outlining and demonstrating how the analyst arrived at the conclusions they did (in the BLUF), how confident the assessment was, and if any intelligence gaps existed regarding the matter.
In some of these intelligence reports, generally what’s called “predictive analysis”, the analyst will provide a matrix of all other potential alternate outcomes which they judge as less likely to occur, what the indications and warnings that they are occurring would be, what would they assess would cause them to occur instead of what they assess was the most likely, and quantified as a stated measure of estimated probability.
THE ADMISSION OF GUILT & BLUF PROMISE
Lucky for you, this isn’t predictive analysis, this is a simple question which was unnecessarily turned into a context heavy, tangent heavy, brevity shirking literary work of art. The BLUF will be wordsmithed very carefully and scrutinized at a granular level of detail, all to ensure that those who don’t love themselves enough to receive the cognitive gifts and enjoyment of reading every word will still know what compelled me to write it, what my question was, and what the answers are referring to.
EXTRA CREDIT - ZUCKERBERED DEFINED
For the uninitiated, the term “Zuckerberged” is used as a colloquialism to describe someone who is in a zombie-like state staring at a laptop or cell phone screen usually typing at a higher than normal rate while so intensely focused on their task (often coding, writing impassioned prose, or researching) that they are oblivious to any and all activities, noise, or people which are in their vicinity.
The term is an obscure reference to a series of scenes featured in the 2010 motion picture titled “The Social Network”, an unauthorized dramatized story about the social media platform Facebook. While disputed by some, the movie told a version of this story which focussed on the unflattering role of long-time CEO, Mark Zuckerberg, suggesting he was not as involved in developing the original concepts for and creation of the widely used service than he has claimed.
The specific scenes depicted a young Zuckerberg in a college dorm room with headsets on, laptop open, and typing feverishly away non-stop for hours at a time while the comings and goings of his roommates and partners didn’t register, and only stopped when someone physically shook him because he had gone 15 hours without food and only drinking Red Bull.
Ergo, if one is considered to be “Zuckerberged”, this individual is in a state of hyper-focused and unwavering commitment to a time-consuming repetitive task being accomplished using a computer or cell phone.
Copyright 2021 by Pete-TheGinger
Gingers have no rights, thus none can be reserved. Unauthorized use of this dissertation, in whole or any portion, is strictly allowed, and punishable by having to read more of them.
