How To Create an AI Version of Yourself with Your Personal AI

Step-by-Step Tutorial To Train Your Own AI Model. Simple. Sane. Private. Yours. No code.

Published in

Personal.ai

12 min readMay 23, 2023

Today’s news is dominated by headlines of new AI tools that hit the market every day. Whether for productivity, education, or collaboration, AI today has the power to help you 10x the knowledge you’re already working with.

Enter your Personal AI, stage right. Contrary to the Large Language models currently in the spotlight, your Personal AI learns entirely from you. Comprising its core data from your daily interactions with your team, community, and friends, your Personal AI model is owned and directed by one human — you.

For a copilot that suggests replies to your conversations, or to keep connections forever, train your own Personal Language Model for free today. Stay ahead of the curve, and own your model and all the personal data that goes in it, without compromising on your ownership of your data and your model.

Let’s get started with the application itself.

Step 1: Create your Personal AI account and access the application.

A new user can sign up for free at https://personal.ai
Get an email invite to join a Lounge, and follow the link to sign up.
Get an email invite to chat in Personal AI Messaging, and follow the link to sign up.

Here is an overview of the Personal AI application layout:

Left navigation bar:

Chats — A place where all your personal chats go. They are direct messages and are private. You can also chat with yourself (your AI) here. Simply click + to invite or add people to chat.
Lounges — A place where all your group chats go. Create lounges per your needs and add people. The owner of the lounge has the option to add their AI to it.
My AI — A place where you can manage yourPprimary AI and sub AIs, also known as AI Profiles. This section is specifically helpful to work with different sub AIs and training activities. Remember, training means adding data to your Memory Stack, and training is continuous.
My Data — A place where you can manage all your data and documents. In this section you can review draft documents, edit them and stack them. You can also look at your Memory Stack and verify what goes into your Stack and what should be removed.

Command bar (Ctrl / Cmd+K):

Search people, Lounges, and AIs.
Search text in Stack or Docs.
Search Integrations such as Twitter and Google Drive.
Search Settings or Tutorial videos.

Message bar:

Message: Accepts text, rich text, images, links, etc. with send options.
AI mode: Set to AI Copilot, Autopilot, or simply turn it off.
Send options: Send to your AI (default) or choose drop-down options such as GPT-3 to consult the internet or DALL-E to generate images.
Learn More

Step 2: Once in the app, start training by growing your stack. There are many ways to do so.

As stated earlier, for every message you send, you may choose whether or not to stack, which will determine what is used in training. While your AI automatically determines if a message is useful to add as memory or not, you can control what messages your AI should stack (train on) and unstack (not train on). Learn More

Message yourself to train on you.

Here are a few basic methods to train:

Explicit Training (Self-chat): Navigate to Chats in the left nav bar -> Select Me Or My AI -> Select Primary or sub AI: Simply message yourself and tell everything you want your AI to remember and train on. Think about this as your autobiography.

2. Implicit Training (Private Chat): Invite your friends and colleagues to Chats by clicking the + symbol: Replace your everyday messaging with your friends and colleagues with Personal AI Messaging. This is a natural way for your AI to start learning about you and your style and surprise you along the way with automated drafts.

3. Community Training (Lounge Chat): Create a lounge, and invite your community or groups to a Lounge by clicking the + symbol. Add people via email. Migrate your communities, customers, and clients and have them engage with you in your Lounge with your own AI. This allows you to address the needs of the community, such as by answering questions, and stack them at the same time. The beauty is your AI Copilot will keep track of chatter and offer suggested responses if within reasonable scores. Learn More

#personal-ai-community

‍

Step 3: Create a habit of using Personal AI Messaging to communicate using Copilot and Autopilot in your self-chat, private chats, and Lounge chats.

Using the Personal AI Messaging desktop application like any other messaging app makes the utility of your AI more promising and integrates the behaviors of stacking, training, and reaping the baked-in benefits of Copilot and Autopilot with your day.

Copilot example in a private chat: Once said, never say it again, let your Copilot draft it. Send if it looks good otherwise make it personal by clicking edit for the next time.

A Copilot draft message

Autopilot example in a Lounge chat: Your community, customers, and clients won’t have to wait while you are away. Set your Lounge on Autopilot with a threshold.

An Autopilot reply (threshold at 22%)

2. Prompt example in a self-chat: Message yourself to work on a project with your own “personalGPT”, your personal AI, and use different prompts just like you would with ChatGPT.

ChatGPT like prompts

Step 4: Architect your personal AI — with multiple sub-AIs to compartmentalize your life, such as personal and professional, and control who in your contacts has access.

We have created the ability to segment areas of your life by creating multiple sub-AIs under your primary AI. For example, you may choose not to share stories about your parents with people at work. This is achievable using sub-AIs.

I maintain my sub-AI architecture as follows:

Day to Day — Here, he takes notes, pushes meeting transcripts, drops LinkedIn profiles of people he meets and social posts, etc.

My Life — Here, he puts all his personal information about his journals, his autobiography, his stories, his preferences, etc.

My Media — Here, he pushes all his YouTube podcasts, public Medium articles, media coverage, speaking transcripts, etc.

My Second Brain — Here, he pushes all his knowledge, the information he consumes, articles he relates to, research he wants to connect with, etc.

My Work — Here, he keeps all things related to his work such as business plans, board materials, market strategies, product documentation, etc.

Learn More

Step 5: Hasten your training process with integrations and data upload mechanisms to your Stack.

Use CMD+K for integrations.

There are a few ways to train your AI faster.

Use documents to write, edit and stack large amounts of data: You can add long, rich text directly as memory when sent as a message, however, there is a full editor in Personal AI to create, edit, and stack long documents. Simply click + and create a new document in the respective sub-AI. Documents are also a great way to add user metadata such as titles, topics, and people to carry on additional context. Learn More

2. Sync Twitter to stack all your tweets: If you are a heavy Twitter user, connect your Twitter account to sync all your tweets (CMD+K -> Integrations -> Twitter -> Sync tweets). Tweets are a great way to bootstrap your AI and get meaningful results in your chats. Once synced, you can even generate more tweet ideas and focus on specific topics. Learn More

3. Simply send links and URLs accessible to your Personal AI to stack: One of the faster methods to stack and train your AI is to use data from links. When you send a link as a message, Personal AI attempts, if you choose, to extract the data and draft a document. If successful, you can edit, remove junk, if any, and stack it. Learn More

4. Use Google Drive links for stacking documents with data: We chose to start off with Drive as it supports many file formats, is widely used, and there is a free version. Once the drive is connected (CMD+K -> Integrations -> Drive -> Upload URLs, just use the direct URL to your Google doc and send the link to your AI. It then creates a draft and you can click Stack just like any other message. Make sure that the file is less than 5MB for each upload. Convert files such as .docx and PDF to Google doc first before uploading. Learn More

5. Use YouTube link for audio and video with data: Similar to any link, any YouTube videos are automatically transcribed and uploaded as a draft document to review and stack when the youtube URL is sent. Learn More

6. A very disruptive method, clipboard stacking, push anything you copy on your computer: If you have files and formats that are not supported by the methods described above, then turn on Clipboard (CMD+K -> Integrations -> Clipboard On) and simply hit copy on any text in any application to paste to your personal AI. Just make sure you remember to turn it off. Learn More

7. Zapier: Integrating with other applications such as text messaging, Gmail, Slack or Discord is easy with Personal AI’s Zapier integration. Your key is available for premium users to use with Zapier. Suman uses Zapier himself as a proof of concept, before investing in full integration projects. Go to https://personal.ai/zapier to get started or CMD+K -> Integrations -> Zapier.

8. API: For customer development such as using your personal AI on your website or writing app for faster data uploads, or using tighter integrations with other messaging applications, use the API available for premium users. The API key is available in Settings. It can also be found at https://personal.ai/memory-api

9. Coming soon: We will continue to develop more integrations specifically around messaging apps for both input and output integrations.

Step 6: Reply with ease, have fun chatting, and stay connected forever.

Your AI is designed to create replies to your incoming messages. You can always use ChatGPT-style prompts for all the messages.

In Copilot or Autopilot mode, incoming messages from others or yourself create a response from your PLM. If your PLM is unable to construct a response because the data within your Memory Stack is not able to provide a meaningful response, then it consults an LLM. This will be reflected in the Personal Score.

Example: Lower Personal Scores with no accuracy mean a certain response is general and not from your PLM. You can choose to edit the response, save it, and send your reply. This will also train your AI on that context. It’s also a good technique to train your AI to ground on a question based on context.

Example: If a response isn’t grounded enough given the data within your PLM, it’s reflected in the accuracy scores. It may still be the response you thought to send, but you have the choice to make it more personal.

Example: The scores also reflect fluency and relevancy and may hamper your scores. As your Stack grows on similar topics, your scores get better.

Example: Use ChatGPT style prompts when you need them. Use reply to carry conversations from the previous context.

Example: Use additional dropdown options when needing other models such as Rephrase, Ask my AI (this works well in Lounges), GPT-3, and DALL-E. ChatGPT will be available as well. They all work in Copilot mode and you may use them however you like. You can also choose to stack them with some personalization. Suman likes to use these options in conjunction with a “reply”.

Turn quick responses into nicely expanded replies, written from your personal AI

2. Use rephrasing to turn lazy replies into something better

3. Use GPT-3 to get quick information from an LLM. Expect this feature to be replaced with ChatGPT. If it’s personal, consider stacking it with some personal edits.

4. Use DALL-E image generation for a modern AI way of exchanging characters in chats :)

‍

Step 7: Do you have a community, clients, or a small business? Develop and integrate your personal AI in your own workflows using API or Zapier.

There are two ways currently to integrate your personal AI into your existing apps. 1) via API and 2) via Zapier. Here are a few examples:

Texting My AI

Twilio <-> API/ Zapier <-> Personal AI. Learn More

2. Quote tweet a response

Twilio <-> API/Zapier <-> Personal AI. Learn More

3. Auto-reply on Slack mention if the response is above a certain Personal Score threshold.

Slack <-> API/Zapier <-> Personal AI. Learn More

Everyday uses with Personal AI

Are you a Parent?

Train on your parenting advice and engage with your kids today, and leave your personal AI as a digital asset for them to connect with you forever. You own your data and your model forever.

Steps:

Journal to your “Personal” Sub AI all your parenting advice, every day.
Invite your kids to chat with you in Personal AI Messaging, in Copilot mode.
Let your kids enjoy chatting while you are away in Autopilot mode.
Create a family lounge and invite your partner and kids to collaborate with AIs, or maybe debate it.

Are you a Student or a Teacher?

Turn your course material into interactive content for engaging and intellectual conversations.