Per a recent article I wrote on Medium, Yes we all need a great guard for our data like Lassie.
So I chose to describe the oracle function as "Collie", to keep an eye on "AI" , so humanity stays in control.
As an aside, witness my #AI_Joe post on R2's Linkedin and ask yourself is this what we want guiding the direction of all humanity?
My flowchart notions below, are a first crack at how to keep the eVoter in control of AI, are now conceptually expanded. That is adding a 'Collie' oracle managing a pack of BorderBots which essentially are the harness "ends" controlling various AI instances, ensuring AI services stay with in their bounds of operation, serving the voter and not the financer/VC behind the AI. Note the absence of traditional government hierarchy, dictating things centrally.
Above is only part of the picture when it comes to reigning in AI to serve the people
Now, we might get the current paper system to work, if we could only get UPS Drivers to stop throwing out US 2020 Armed Forces Mail in Votes, into any ditch they find convenient. (not.)
Connecting the Collie oracle to the Voter Controlled SGR Blockchain
Realizing eVoting and eGoverance via the Blockchain is currently WIP. (Think '63 Chevy Biscayne just entering the "chop shop" for a make over.)
I earlier covered in my posts the Swiss #AGORA project and their early progress for eVoting on the Blockchain. Agora's early progress was and I believe still is, very promising. Frankly though, eVoting on the Blockchain in its current form is only 1/4 of the way to enabling a Voter Self-Governed Republic on the Blockchain.
Collie Commanded BorderBots: "At the Voter's Service" in Real Time, with the oracle Collie "Boss" in control of AI.
Having a 'Boss' Collie oracle direct a bunch of BorderBots to manage the 'blockchain' eVoting build, test, deployment and eVoting Epoch (voting round) process makes a lot of sense when, the 'Boss' oracle itself, is governed by the eVoter, including having such bots operate @ scale in different single use function capacities "Chinese walled" if you like, from other different BorderBot functions essentially tending to the "evoter" flock and their submissions, with both help as well as results protection and " indiscernible (bad or altered scan) vote quarantine", the latter ready for manual inspection of what becomes immutable data only accessible via the blockchain, so it's either approved by a signed inspector who has access to the public/private key to ascertain the acceptance of the vote ballot in question , or not, keeping everything "above the table" (no voter count fraud).
BorderBot "Defacto" Control over AI Instances- An AI "Empathy Injection".
eVoters being the sentient, cognizant beings they are, have one trump card over AI.
Collectively, the silent majority, if given the chance, will choose to insert empathy into the voting equation and, it's likely, that outcome will indeed have some flavour of empathy built into the result, expressed "by the people", "for the people".
The BorderBot receives it's marching orders from the Collie oracle in the form of encrypted instructions travelling over the Control plane. On receipt the BorderBot dutifully reports acceptance over the same Control plane, then reports BorderBot actions over the monitoring plane back to Collie. The BorderBot has a few mandatory items it must do with each action which is check the network clock and post the clock value to it's log locally and publish that action back to Collie over the control plane. This simple action of "punching the clock" frequently for every BorderBot move ensures Collie get's a 100% transparent audit trail of events from the BorderBot stored as an encrypted log file within the Collie oracle's own persistent data structure, which is in turn posted over secure monitoring plane and communication channels for posting private key access to the SGR Blockchain BorderBot Smart Contract which allows those with access to that "SCS" to gain access via Collie to the timestamp log pertaining to that particular BorderBot and what it has been doing with it's assigned AI instance.
TK Observation - "Is that really Joe on the left (with more forehead moles), or some "Bounder" body double wearing those Google Lenses reading AI delivered Tele-prompt responses in the latest US 2020 Election debate? "
BorderBots: Keeping AI within eVoter bounds- No AI "Bounders" Allowed!
Central to eVoter protection described above, "Whatever the eVoting populous decides on the SGR Blockchain determines both Collie and BorderBot behaviour/control of AI" should drive development of all eVoting imnsho. A little bit more below on how that might work.
The Collie oracle get's it's own marching order on how to deploy and manage the BorderBots from the SGR Blockchain SC attached to the latest Epoch of voting results via it's own encrypted control plane as a receiver. Within Collie's SC messages, always encrypted, are instructions for single or groups of BorderBots. Collie oracles are specialized. Collies store instances of BorderBots and deploy/provision those bots over APIs communicating to the AI instances to which they are assigned over a known encrypted channel and API. They log everything the BorderBots are doing and post the location of those logs with access credentials to wherever those encrypted logs are stored. (Such logs represent much data and require cheap fast storage as a service).
The AI instance knows only one master, Collie. The Collie assigned BorderBot is itself, simply a local masked "Collie" address running in the same VMware in its own container, much like any agent, a container which cannot be compromised by the AI instance, in that it can only receive and will only function if it receives periodic heartbeats"HB" from the BorderBot.
If the AI Instance does not receive the local BorderBot "HB" (a masked proxy for Collie), the AI instance gracefully goes into sleep state (for some period) and, after sleep state, shuts down according to the SGR Contract managed on its behalf by the Collie oracle and implemented by the Collie assigned BorderBot controlling that instance locally.
Should the Collie oracle assigned BorderBot cease to operate/intact with the it's assigned AI instance, in effect, the lack of HB acts as the AI instance "Kill Switch", with one intermediate step, that is, the AI instance transmits to the Collie, saying the AI instance no longer receives the HB from "Collie" ( with the AI Instance receives from the local masked address BorderBot acting on behalf of the Collie). This AI instance transmits to Collie is also an outbound masked address which the AI Instance understands to be Collie (not knowing it occurs locally or is relayed by some containerized code separate from the BorderBot to the Collie's actual address ). At that point, depending on the SGR SC assigned logic executed by the Collie oracle, Collie will either interact with the AI instance to remedy via the same source/destination address pair used by the AI Instance transmitting this new state to Collie (ie- in most use cases, send down another serialized replacement BorderBot to same container, over writing the old BorderBot and remotely activating the new replacement BorderBot) or, simply ignore the AI instance Beacon (3 times, 15, 10 and 5 sec intervals) at which point the AI instance, certified safe, gracefully shuts down. This means the new serialized replacement BorderBot, has to boot up in the previous BorderBot container and respond to the AI instance with a new HB within 30 seconds in this example.
Otherwise BorderBots as proposed do their jobs as the Collie oracle dictates, sensing and reporting anomalies to the Collie oracle such as the 'bounder' above (illegal use of AI props by Candidates) to protect the integrity of the related debate ( ie BorderBots scanning the live IPTV streamed video for various visual "prop" anomalies IDed to matching illegal criteria such as: Candidates can't wear wires, nor; can Candidates wear Google Lenses to get information feeds nor; can Candidates have microphone speaker wire up your sleeve, to ultimately protect the eVote (voting round) outcome.
AI Instances- Certified Safe- How does that work?
As to how that AI 'safety' is determined, such AI safety will come down to a voter created set of AI safety software build & test standards, lined up with voter created laws governing the deployment of AI instance.
How much voter delegated authority given to Collie oracles needs to be governed by "Rule of Law", 100% of the time, perfectly, to truly benefit 'the people'.
Sentient, Empathetic Control over AI: Created For the People, By the People- AI Safety Bylaws- asap!
Getting control over AI, before "it" gets control over us, the people, is THE paramount order of the day.
We as voters, have an urgent responsibility to table this fundamental issue of the ages to our local, state/prov. and federal governments, asap.
It's likely the official government response is lethargic at any of the levels of government. Therefore, the best course of action is to, as a concerned voting public, act en mass and, take two steps locally to correct the lack of AI Safety in your locale.
Thankfully, most municipal bylaws tend to have priority over state/provincial laws and federal laws, especially where no laws for a new issue exist.
In the above, common "lack of AI Safety" case ,the lack of any related bylaw is most likely pervasive where no law exists to govern AI for the safety and benefit if the people generally everywhere.
AI Safety Bylaws: How to get them in place in your locale.
A new "AI Safety" bylaw can be promoted, backed via voter petition (say 1000 signatures) is the first step and, then as a second step, the petition can be tabled locally with the right bylaw wording to your local council to get such a bylaw passed, asap. Doing so ensures the AI Safety bylaws drawn up (get a municipal bylaw savvy lawyer involved) will protect all public and private entities as will as private persons.
As side note, this likely means AI driven CCTV surveillance of the people has a big "cease and desist order" surprise coming, should such an AI Safety bylaw get tabled and passed by local council . (It's likely violating your rights and freedoms, right now as you read this blog.)
To all my readers, as the World 'Wyrds' in response to the most recent AI assault on our senses (US election debate) , please be safe and please do speak your mind about AI.
Post your thoughts in the comments below, thanks :)
TK over and out..