Name: LMP - Labs AI Term: 1 Cycle Cycle Amnt: 15,000 Total Amnt: 15,000 Author: JSKitty Receiver: PIVX Labs Address: DLabsktzGMnsK5K9uRTMCF6NoYNY6ET4Bb Created: 14-09-2023 Status: Active Vote Hash: 94bf4d8c8cc9ce3de1338da5d2a545b01a687fcd7a1f7e2e40dfd7c33d9ef7b4
Downvote! Labs AI funding will be delayed and re-submitted next cycle to preserve budget!
This proposal is aimed to cover the full costs of AI used at PIVX Labs to augment and enhance the PIVX Labs team and workflows.
This comes after a 'restructure' of Labs designed to push for efficiency. and remove team members that caused bottlenecks in our workflow, iron out areas that had unnecessary friction (dev-to-report flow such as The Superblock Report, developer overhead such as DevOps, code tests and documentation, as well as complete automation within areas of PIVX Labs that it makes sense: such as Social Media control, which is soon to be delegated in totality to an AI which consumes PIVX Labs 'report data' and spits out polished, automatic, update-style Tweets on a routine basis).
What does this help to solve?
- The need for writers: Labs no longer employs content writers, instead, we augment our existing team of Developers with an AI that compliments their workflow, rather than hinders it with overhead, the key word is Augmentation, not 'Replace', content is still accurate because it is still derived from our work, but it is converted in to a standard, non-techie format for all to consume with ease.
- An example of such writer augmentation, is the PIVX.Poker proposal, which was created and polished from very "rough" developer report text, using GPT-4 and a browser integration to use GPT-4 directly within the PIVX Forum.
- The need for common DevOps: due to recent improvements in the GPT space, all of PIVX Labs' servers are now partially managed by AI, such as the configuration of our MPW Nginx server, our safety systems, our nodes, and fast setup of new infrastructure, all of which take up a surprising amount of dev-hours, such tasks are now completely minimised with software such as Open Interpreter, which Labs servers are equipped with:
- Recently, all PIVX Labs infrastructure was switched to an encrypted SSH key system, AI completed this entire task itself via SSH on our Primary server.
- Open Interpreter created scripts to: open tabbed terminals, decrypt, and login to ALL Labs servers, in a single hotkey on JSKitty's main system, all while SSH keys remain encrypted on-disk, which saves a ton of overhead time, without harming security.
- Developer Overhead (or "Grunt Work"): PIVX Labs devs are all equipped (if so requested) with GitHub Copilot and PIVi, both of which are extremely effective at quickly handling grunt-work tasks, such examples below:
- PIVi wrote almost all of the 100s of JSDocs in the PIVCards backend, speeding up development of PIVCards to less than 3 weeks for the full platform, with a single dev.
- PIVi created Prodder almost completely solo, which removed the friction between Developer PRs and the Quality Control team.
- PIVi created the PIVCards Proxy System, which enables PIVCards to load-balance and hop between various continents and regions, using Labs Proxy servers located around the globe, hence PIVCards has 'region' options).
Where exactly does this go?
Somewhat surprisingly: mostly not to PIVi!
Our experiments in powering up the community with general-purpose AI, intended to empower the community to create easy and fast content to grow and promote PIVX; were essentially a failure.
More specifically, rather than using PIVi to empower Content Creation (with a few exceptions to some PIVians that did excellent Twitter content with PIVi), most simply used it for entertainment, or couldn't grasp the concept of "Prompt Engineering", and those whom did: abused PIVi to create 1000s of irrelevant art pieces of cats and dogs.
Instead, we are going to focus on using AI for more specialised and industrial purposes, and withdraw our efforts from 'Community AI'.
We will equip knowledgeable people with knowledgeable tools, rather than hoping the community help us.
Cutting to the exact question without the fluff:
- Open Interpreter (GPT-4): this is used for Labs infrastructure, DevOps, security, node and scripts management, and various other system overhead tasks that are now handled autonomously.
- GitHub Copilot: GitHub's own AI integrated with IDEs, augmenting simple developer tasks, documentation, JSDocs and more.
- ChatGPTBox (GPT-4): a browser integration that allows using LLMs directly within the browser: this is how Labs Writing is augmented, such as Proposals, Forum Posts and The Superblock Report.
- PIVi (GPT-4): PIVi will soon be equipped with smarter tools that greater automate the team workflow: one of such ideas are connecting PIVi with the Labs Twitter, to allow creating, proofing, and posting Tweets with updates of the PIVX Labs team as we work inside Discord, since PIVi is trained to understand GitHub PRs and Tweets, this could mean simply passing PIVi a PR, and it goes to create content with that PR, with minimal overhead on the team to create the content themselves, these tools will be unavailable to the community.
This proposal in specific aims to cover between 3-4 months of augmented or enhanced work, but depending on usage and integration, results may vary.