Interesting read if you are wanting practical ways to confuse automated recognition cameras.

[W54/XM-388]

Here is an interesting article from some researchers about practical and simple ways to beat the automated recognition / identification cameras that are going up everywhere thanks to our Technocrat Overlords.

Even if you don't actually put any of this into practice, it's still an interesting short read on how things work in case you need it.

 

[ArmyJerry]

I avoided posting this subject to not incent them from targeting fixes.

 

[Maggot]

Ive seen some other info on this.

Where can you buy this type of stuff. It would be fun to wear the next time I visit DC just to see if it set off anything or got me stopped.

 

[Crang]

To evade notice my plan is to wear Christmas sweaters with MC Hammer pants, and those giant vaginas on my head the protesting SJWs love so much......

In all seriousness though what AJ said. I work in a field with some overlap. I recently spent a good amount of time on a project working with the company that creates the training data for most AIs of note- from defense contractors, to aviation, to self driving cars to all the people you cant talk about and everything in between. AI is ALL about that training data and there is one group who creates it afor everyone else. Their library continues to grow and be refined. Anytime an anomaly comes many images and variations of it gets added to the core library. From there the images are presented to thousands of workers who I believe are in Mumbai who look at it and basically click yes or no. It goes back into the system for live work and then errors back to the people in a constant feedback loop.

Long way of saying once anything that beats it becomes known it will work for a bit but less and less with each pass of refinement. Gotta keep it new.

 

[W54/XM-388]

From there the images are presented to thousands of workers who I believe are in Mumbai who look at it and basically click yes or no. It goes back into the system for live work and then errors back to the people in a constant feedback loop.

Pretty much the state of all this great "AI" that all the technocrats are trumpeting as the great new thing.
A bunch of Indians looking at your data and listening to you sing in the shower (from recent revelations about Alexa).

Then you just work on making sure that their "AI" classifies something as what you want them to think of it as for one off things.
Also that is where you can start having fun with specific regional cultural things that look exactly one way to Indians but those in the know understand it is something different (also in one off not publicized ways).

 

[Crang]

Pretty much the state of all this great "AI" that all the technocrats are trumpeting as the great new thing.
A bunch of Indians looking at your data and listening to you sing in the shower (from recent revelations about Alexa).

Then you just work on making sure that their "AI" classifies something as what you want them to think of it as for one off things.

Yup. Thats why we usually just call it machine learning not AI. It doesnt work (well) without carefully curated data by vast amounts of cheap labor. Its takes a lot of mindless work to get it running. Mindless for us is still too complex for most machines and even ones that will become quite capable start out almost like infants with a blank slate and set of algorithms. Once its running the need for human input is reduced greatly but still needed. Self learning technologies will continue to improve but for now its not as complex as most people make it out to be. Although after the singularity all bets are off.....

Look what happened when Microsoft released Tay- a chatbot that used Twitter for training data. It hilariously became a genocidal racist in no time at all. We need to be very very very careful what we feed these programs as they learn and have more powers and autonomy in the future.

And yes- the group I worked with created the initial data to train Alexa, Siri, Erica........

 

[Blue Sky Country]

----------
Even if you don't actually put any of this into practice, it's still an interesting short read on how things work in case you need it.
----------

Just like guns, broadheads, field trauma kits, and fire extinguishers.........Better to have 'em and not need 'em, than need 'em and not have 'em. Thanks for the post, OP.

 

[Spblademaker]

 

[MarinePMI]

Hence why many of us have taken to calling "AI", "RI" (not machine learning, since it's not really "learning" anything). What many call "AI" is really just "Reflective Intelligence" since even the training programs really only do what you tell them to do. Even the miracle of inference is tightly controlled and programmed.

"Bu...bu..but Watson beat the human Go champion!" Hmmmm...a data warehouse, petabytes of data, thousands of CPU threads, gigawatts of power, a metric shit ton of water cooling....beat a human, with a brain, that uses 18 watts of power.

Think about that for a second...

Smoke and mirrors...

 

[Bender]

Ooooor. You don’t live in a nanny state?