Code:
Name: LMP - PIVi
Term: 1 Cycle
Cycle Amnt: 6,500
Total Amnt: 6,500
Author: Sandude, JSKitty, PIVi
Receiver: PIVX Labs
Address: DLabsktzGMnsK5K9uRTMCF6NoYNY6ET4Bb
Created: 19-07-2023
Status: Active
Vote Hash: da31c011b9db236a65950a66ab4ecdb74f168fb24ea121f4b67ce37ceb06d885Who or What is PIVi?
For those unfamiliar, PIVi is an AI personality using a Discord bot to serve a variety of roles in the PIVX community. PIVi has already been working for PIVX for a few months, constantly evolving and becoming more capable. Our intention is to continue maintaining and developing PIVi to expand its use-cases and transform it into the perfect PIVX ambassador. PIVi can also support content creators to produce any text and pictures that they need to perform their tasks. The possibilities are truly endless.
But how is PIVi so smart? How does this all work? PIVi is currently powered by GPT-4; the newest version of OpenAI's language model systems. In addition to this, PIVi has received specific training to have some knowledge about the PIVX project, and to be able to use the different tools it has been upgraded with. These tools are what make PIVi unique, and an excellent support tool.
Here are some examples of PIVi in action:
PIVi also performed the voiceover for the video in this tweet:
This graphic:
This graphic:
These ones:
And many other tasks either requested by our users or necessary for our social media managers and content creators.
PIVi's Toolset
These are the tools PIVi can use in order to produce valuable content or simply interact with community members. In the future, we will expand this list and give PIVi further capabilities, as AI applications keep being improved and developed. PIVi can make use of these tools any time that the situation calls for it, without being specifically asked to do so. Here's a full list of what makes PIVi so special:
- Powered by GPT-4: PIVi uses the latest version of OpenAI's GPT, heavily abstracted and fine-tuned for PIVX-specific tasks and work.
- Per-channel isolated persistent memory: Thanks to it, PIVi can adapt in real-time individually to every channel.
- Task and user-adaptive learning: PIVi can learn to perform tasks autonomously, as well as learn and adapt to individuals to better assist with their specific workflows or needs.
- Large toolset: PIVi can use a great variety of "tools" at its disposal, either by itself, or on request, such as:
- PIVX: Yes, PIVi can use PIVX. It has a full wallet to itself, for autonomous work, such as: Ambassador work, On-boarding, working or trading with users, etc.
- Reactions: This is just so PIVi can use Discord reactions, a neat little personality tool.
- Draw: PIVi can draw by itself! Try asking it: "Draw me the Great Pyramid, glowing purple".
- Upscaling: PIVi can upscale images using AI super-resolution for better quality.
- Vision and OCR: PIVi can visually 'see' and 'read': send it a nice picture, or a book page, it can see them!
- Kicking: PIVi can kick people when it wants to, this is unironically useful for moderation. PIVi has already kicked scammers from the PIVX Labs server.
- Web integrations: PIVi can read GitHub Pull Requests, Tweets and PIVX Forum posts. These are very useful for autonomous reporting, community updates, translation, etc.
- DAO Integration: PIVi is integrated with the DAO, it shares PIVX Labs proposals in real-time, cross-Discord, with the ability to advise on them when needed, a valuable DAO automation system.
Additionally to being able to interact with PIVi's AI, community members can also make use of a broad suite of AI tools that the Discord bot has access to. In order to use them, just type the command and corresponding inputs on any Discord channel where PIVi is enabled, or on a DM with PIVi. This is a list of the available commands:
- /imagine <description>: Now with a new AI image engine, PIVi can create art like never before.
- /video <description>: PIVi can now create short videos, straight from a text description!
- /say <text>: PIVi now has a synthetic voice! Not your usual TTS, a life-like synthetic voice clone!
- /proposal_roundtable <forum url>: PIVi can act as a DAO advisor for proposals, deploying a roundtable channel exclusive to the proposal.
- /topic <description>: You can tell PIVi to create an exclusive thread channel, focusing on a given topic or task, good for crunching out specific work.
- /tweet <tweet link>: In case Twitter goes dark again, PIVi can bypass view limits, screw you Elon!
- /music <music style> <optional art style>: PIVi can create AI music, a short .mp4 track.
- /summarise <amount>: PIVi can quickly & concisely summarise large conversations for easy reading.
While PIVi's functionality is really impressive, it has some maintenance costs that need to be covered. Heavy usage of PIVi's underlying GPT AI requires a stream of OpenAI tokens at 3-6 cents per-thousand-tokens. Moreover, PIVi requires additional funding for autonomous work purposes. Of course, there will be constant optimization to reduce the token cost overhead; we have reduced PIVi's cost by more than half this last week from various NLP optimization techniques, but naturally, as PIVi grows more popular (and has been!), we need to keep its various AI engines running with direct funding.
Nevertheless, this is well worth it; as with some more development and fine-tuning, PIVi will be able to perform a great variety of useful tasks for the PIVX community. For example, PIVi could host AMAs, read video scripts for voiceovers, or even handle the onboarding of new users. PIVi could teach them about PIVX or even shower them with a lil' PIVX love in newbie tips. All hands-free!
But those are just a few examples of things we could do. There is an endless list of possibilities that could be made real thanks to having a fully autonomous AI with such an impressive toolset.
Last edited:
