Personal Information Agent

Full Title or Meme

Support for a personal assistant to help users control access to their personal information.

This is called a delegate in most of the content of this wiki.

Context

• The operating assumption on this page is that users want to delegate part of the decisions about releasing the personal data to an agent that operates solely on their behalf.
• Another assumption is that a large majority of users would not toleration interruptions every time some element of their personal information was accessed.
• Relationships on the internet are mostly Asynchronous where users would like to have some privacy, but demand that organization a fully transparent about themselves and the motives in accumulating user personal information. This page is concerned solely with natural persons and their desire to not have all of their personal information on display to all enterprises where they interchange information.
• Researchers at Carnegie Mellon University's CyLab Security and Privacy Institute assessed the degree of autonomy that people would feel comfortable giving to personalized privacy assistants (PPAs). The team surveyed users on three increasingly autonomous versions of PPAs; most participants reacted positively to the first version, which would simply let users know that devices were around them, while a few said it would make them anxious. A second version that knows users' personal privacy preferences, and makes recommendations from that information, also found wide favor, while the third PPA, which would exclude users from decision-making entirely, provoked mixed reactions. CyLab's Jessica Colnago said, "We found that people are definitely interested in having some sort of assistance like that provided by a PPA, but what that assistance looks like varies across the board. In different scenarios with different people, they want different ways of interacting with the system.”
• The Health Insurance Portability and Accountability Act of 1996 (HIPAA) Privacy Rule’s right of access provisions require that individuals or their personal representatives have timely access to their health information (within 30 days, with the possibility of one 30-day extension) and for a reasonable, cost-based fee.

Problems

• Decentralized IDs were created to give user's control of their identifiers, but at the cost of posting those identifiers on a publicly resolvable public ledger. The result is that any usage of the the DID can be correlated with any other use of the same DID. Once sufficient correlations have been accumulated against that DID, the person is uniquely identifiable.
• Most identifier authentication services on the web now offer a Service Endpoint that can provide information about the identifier. If the user is hosting their identifier on there smartphone, no service endpoint on that phone could be always available for query.
• Nearly all of the privacy solution require users to keep separate track of identifiers used in public and private settings. This is not likely to be acceptable to the majority of users if we don't also give them a User Agent that can track all of their relationships and the identifier that they use for that relationship.

Solutions

Any solution that met the context described above would need to have an agent that was responsible to, and trusted by, the person that owned the personal data.

Web based Agent

Also tracked in this wiki as a Personal Cloud Store (qv).

Personal Artificial Intelligence agent

Can a smaller language model with significantly less parametric memory emulate such emergent ability of these larger language models? The Berkeley Artificial Intelligence Research "goal is to develop Small Language Models (SLM) that are capable of complex reasoning that could be deployed securely and privately at the edge. Here we will discuss the research directions that we are pursuing to that end. First, we discuss how we can enable small open-source models to perform accurate function calling, which is a key component of agentic systems."

Demo of TinyAgent-1B along with Whisper-v3 running locally deployed locally on a Macbook M3 Pro. The framework is open sourced and available at https://github.com/SqueezeAILab/TinyAgent

BingBot

It sounds like you're interested in cloud agents related to GitHub. Here are a few notable ones:

1. **Amazon CloudWatch Agent**: This agent helps you collect and export host-level metrics and logs from instances running on Linux or Windows servers. It's useful for monitoring and managing your infrastructure¹(https://github.com/aws/amazon-cloudwatch-agent).

2. **Arduino Cloud Agent**: This agent allows you to use Arduino Create applications to upload code to any USB-connected Arduino board directly from your browser. It runs in the background and works seamlessly with Arduino devices²(https://bing.com/search?q=git+hub+cloud+agent).

3. **SSH Agent Forwarding**: While not a cloud agent per se, SSH agent forwarding is a useful feature for securely connecting to GitHub repositories. It keeps your SSH key loaded in memory, so you don't need to enter your passphrase repeatedly³(https://docs.github.com/en/enterprise-cloud@latest/authentication/connecting-to-github-with-ssh/using-ssh-agent-forwarding).

4. **AgentCloud**: This is an open-source platform that enables companies to build and deploy private LLM chat apps, allowing teams to securely interact with their data⁴(https://github.com/rnadigital/agentcloud).

Source: Conversation with Copilot, 10/8/2024

(1) Amazon CloudWatch Agent - GitHub. https://github.com/aws/amazon-cloudwatch-agent.
(2) arduino/arduino-create-agent: Arduino Cloud Agent - GitHub. https://bing.com/search?q=git+hub+cloud+agent.
(3) Using SSH agent forwarding - GitHub Enterprise Cloud Docs. https://docs.github.com/en/enterprise-cloud@latest/authentication/connecting-to-github-with-ssh/using-ssh-agent-forwarding.
(4) GitHub - rnadigital/agentcloud: Agent Cloud is like having your own GPT .... https://github.com/rnadigital/agentcloud.