Topic: Grassland: A Scorched Earth AI "Eye" Recording ALL Human Activity (Read 288 times)

The current eon's model you can go and download now only detects people and cars but I've been looking at how well facial expressions are detected in some face detection models, as well as the gaze of the eyes (where they're looking) using 4k and 8k cameras. I doubt my neighbours are aware that I've built a machine that has learned the things about them that I've listed on the site (average height, individual walking gait identification pattern, estimated salary based on car model, family structure, daily schedules...) But in case that perturbs them before they get around to running nodes themselves, I'll want to know if when they look at my house their expression would fall into the category of displeased. This and the lip reading model won't be for quite a few eons yet for obvious technical reasons, one being we need to build a "foundation" for the detections first, but once we get there, it'll give Grassland users a good estimate for the state of people's emotions and dispositions in regards to us before those people are even fully aware of their opinions themselves.

Even so, remember, as I said on Reddit "... you're still not actually publishing the node's location to the wider network. The only thing anyone else on the network could be sure of is that there is a Grassland node that has access to a video feed that's coming from a camera and that camera can view a particular location. Your node could be in Argentina reading the digital video feed from an IP camera in St. Mark's Square in Venice."

In other words, just by looking at the map you can't really be sure where the camera is or how many there are watching that location. Even then, the node reading that camera's video feed could be half way around the world.

But again, I'd advise you to follow your country's laws in any case.

In many countries there is the legal concept of public and private settings. Do people use Instagram in your country? Do they take pictures with their phone in public? It's legal as long as it's legal in your country for people to appear in the background of a photo without their consent like people in the background of a photo in a park, on the street, shopping mall, concerts etc. places where people have no expectation of privacy. So if you end up in the background of someone's Instagram photo and they didn't commit a crime for doing according to your laws, well this isn't even as bad as that.

It's not as bad as that because nodes don't even keep nor publish what they capture, they're just identifying the type of objects in the frame using deep learning. Once that's done and verified, they discard the frames. So if it's illegal to identify that someone is a person using the visual spectrum (light), well... you're doing that every second of the day with your eyes and your brain, are you not?

As to the objective, we might not get there, but as long as there are people in the world who like money, the remaining "dark" areas of the map will continue to be "lightened up".

I've updated the site so you might find new information that you haven't seen yet https://www.grassland.network/. You might need to refresh your browser page in case you're viewing a cached version.

coming back to your idea

Despite that the objective sounds a bit too massive to attain, at the moment it does seem realistic to me, from the tech point of view.

Question about privacy is more from the legal side - this is an intrusion into the personal life, without a consent. Even if the system is distributed, people who maintain it are at risk. Surely it is likely that govts might have some similar systems, but they don't go public with them.

If when you're talking about "privacy" you mean identifying who is operating the node, no that's not revealed. See my answer to someone asking the same question on Reddit (Question #2) -> https://www.reddit.com/r/deeplearning/comments/9ze4cy/tensorflow_cnn_mapbox_3d_real_life_simcity/eabsg4m/

Yes, I do. First, most nodes won't need to keep the entire data set. Just the hashes of each block to have consensus that the database wasn't changed and history rewritten. If they want to view events much further in the past and/or in places they don't store the data set for, there'll be at least one website with the entire global data set running OpenStreetMap's global 3D map along with same hashes for verification.

If when you mention "privacy", you mean how the network deals with the data gathered from objects the nodes are detecting, read on (But if you aren't, see my next comment below). As regards the data the network gathers from objects the nodes are viewing, you'll have to throw away this veneration of privacy to grasp what this is really about; like esteeming the Divine Right of Kings, it's not going to work any more. We're fighting over the right to keep riding this precious dead horse while corporations and governments are riding over us in their AI ferrari's because they know that the ones who'll will win are those who privately control the most data to train their AI's to do, well... everything that a human can do and more. They don't care about your privacy. It's long gone and turned into glue. They just want you to think you still have it so the conjob, the sleight of hand can continue till one day we wake up and realize all our land is in the hands of the Conquistadors.

Grassland is a competition based on Game Theory^[1] --What strategy should you take considering the strategy of others? The strategy of players in Grassland's game is laid out for you. It's to gather as much data about human activities by pulling in data from cameras wherever they can place them or find them online by setting its calibration on the virtual 3D map, then feeding those frames into the current eon's AI model, hashing the output, signing it and broadcasting it into the network. The more hardware they can throw at the problem, the more proofs-of-work they're broadcasting, the more tokens they generate, gathering data about you and thousands of NPC's and essentially trading whatever currency their compute costs are denominated in into Grassland's coin, a fungible, cryptographic token backed by proof (of work) of the internal activations of an AI.

The data, the emergent property of all this greedy, self-interested and beautiful computation, is merely the exhaust from this engine. And it's freely disseminated to everyone not just because it stands as unassailable proof an AI performed a difficult computation on human behaviour and earned those coins but because it's important that everyone, especially NPC's, knows the truth about the world for science, democracy, health, justice and all rational decision making. An embargo on stupidity and ignorance in that node's corner of the world at least. Advantage must then come from helpful innovation not parasites rent seeking on people's data.




[1]. https://en.wikipedia.org/wiki/Game_theory

Do you think it's objectively possible to implement it in production, Mainly in terms of personal / data privacy ?

Also, I built this Serverless Lambda framework running Tensorflow -> https://github.com/grasslandnetwork/node_lite_object_detection to give node operators the same deep learning inference that would normally require buying $3,000 worth of hardware but instead it'd just be around $50 per month. And since it's on AWS Lambda, it gives them "infinite" horizontal scaling.

Thanks! Security and accountability is one aspect, but it's also about Big Data. And the data that is captured is reused to improve the very AI that we used to capture that data. So it's a feedback loop of constant improvement.



It's useful at every size of the network. Even if it were just one node whether it be ...

• for security
• plain curiosity about people's behaviour
• data science (what's this neighourhood's average height, walking gait, salary based on car, how many (visibly) pregnant women etc.)
• because it's interesting seeing people's lives from a bird's eye view and being able to rewind time
• multiple vantage points for a self-driving car
• an impartial, indifferent, semi-omniscient, non-human oracle for a smart contract
• or the fact you believe the crypto-currency will be worth something some day
• etc. etc.

Much cheaper to just wish for a god or two... 

Ive often wished there was a satellite for each person on the planet, watching and protecting them

Great Idea you have here!

https://www.grassland.network/
https://github.com/grasslandnetwork/ (version 0.1)

Grassland is an anonymous, peer-to-peer network of open source, computer vision software that can take any video feed and gives everyone on the network a permissionless, (politically) stateless, and indelible public record of the activities of any person, by allowing one to view any region of the globe in a 3D "god view" mode similar to the games SimCity® or Civilization®, but with the ability to rewind time and view their entire history.

Nash Equilibrium
It's structured so as to be inescapable by any one party (preventing them forming an asymmetry) as long as there remain other parties acting in their own self-interest (Nash Equilibrium^[1]).


2D To 3D
No sensors, SLAM or point clouds needed. Each instance of the software acts as a contributing node in the network. By giving a node access to any fixed perspective, 2-D camera feed, it converts it into a 3-D, searchable, simulated re-creation of events in the part of the world it's viewing using OpenStreetMap 3D. It's a dispassionate, non-human, semi-omniscient, public frame of reference that also incentivises nodes with a digital, financial reward.


Proof-of-Work
Camera frames aren't stored in the database, just the information gathered from each frame, which includes among other things object identification, object geospatial information, a timestamp, a SHA hash of the frame and a hash chain of the activations of certain hidden layers^[2] of the CNN during object detection. The frames themselves are only needed for a little while after for random confirmations and then they can be discarded. It's astronomically improbable to have hashed those hidden activations into the correct digest unless those hidden layers were actually computed, the "proof-of-work".

Grassland's digital coin is a fungible cryptographic token backed by this proof of AI computation.


Steadily Increasing Detail And Difficulty
Current computational requirements are low as it's only necessary to detect and track people and cars now but that will increase over time.

Every few months, nodes must download the next version of the network's deep learning model. Which will recognize and track more objects and activities in order to discern more and more about those objects. So over time the network becomes an increasingly accurate and harder to fabricate representation of the real world.


Why The Heck Did You Build This?
A few months ago, I just wanted to turn my neighbourhood into a game of Civilization. But I needed a way to incentivize other people to join without having to trust their cameras and that they weren't faking data... Having designed a proof-of-work for computer vision, I realized we could do so much more... Grassland takes such a unilateral and fait accompli approach to data as an answer to the power asymmetry enjoyed by corporations and governments with respect to all the data they have on the public. This only serves to encourage social manipulation, lies, propaganda, fake news and a trickle-down reality. Data is too valuable for powerful organizations not to hoard and use to their advantage. Well-meaning consumer data laws only protect us from ethical organizations and organizations who just lack the resources to skirt them, but not from the hackers who've already taken that data. Neither is there any conceivable benefit from the public having limited or warped information about our ever changing and complex political, social and (rapidly being destroyed) natural environment. So the only option left is to "burn the forest", "scorch the earth" and level the playing field, as it were, in a manner that's inescapable. While also providing an economic incentive to do so and seeing if it catches on.

For more information -> https://www.grassland.network/  && https://github.com/grasslandnetwork/



[1]. https://en.wikipedia.org/wiki/Nash_equilibrium
[2]. https://www.spiedigitallibrary.org/ContentImages/Journals/NEUROW/5/1/011008/FigureImages/NPH_5_1_011008_f002.png