CES 2024: Announcing Personal AI MODEL-2

January 9, 2024

Personal AI enables everyone to create and own their unified memory and their personal AI for universal access. With MODEL-2, we are bringing in multi-persona, multi-modal, multi-channel experience to your own personal AI.

Recap of MODEL-1

In April 2023, Personal AI launched MODEL-1, a memory based personal AI model that is user-trained, user-owned and user-controlled.

 The image displays a presentation slide about a "Personal Language Model (PLM)" that is customized for individual users or brands. It illustrates the concept with server icons representing the personalized model and emphasizes the PLM's key features: widely accessible, cost-effective at 40 times less expense, does not require coding skills to train, highly accurate, authentic to the user's needs, and completely owned and controlled by the user.
What is a Personal Language Model (PLM)

MODEL-1’s core technology focused on an individual’s long-term memory and a personal language architecture driven by an ensemble of expert models, separately trained on each user's data. 

MODEL-1 provided the foundations for replicating the digital mind of a human – a continuously learning model with unlimited memory, enabling use cases such as hyper-accurate question answering and hyper-personalized content generation from personal memory. [Read More]

Announcing MODEL-2

On January 9th, 2024, we are proud to announce Personal AI MODEL-2, a comprehensive Human OS system that is multi-persona, multi-modal, and multi-channel. 

 The image is a flowchart on a dark background that describes the interaction between two models. "MODEL-1" processes documents, transcripts, facts, and beliefs into long-term memory and historical knowledge. This knowledge undergoes learning and updating through training and is used in inferencing to recall or relate information, resulting in factual and informational use cases. "MODEL-2" deals with short-term memory and continuous knowledge, taking input from conversations, messages, opinions, and preferences. This model also goes through learning and updating, with its inferencing resulting in creative and conversational use cases. Both models interact with each other, indicating a continuous flow of information and learning between them.
Evolution from MODEL-1 to MODEL-2

Long Short Term Memory: MODEL-2 integrates both long-term and short-term memory components (independent from a LSTM) with a conversational model that mirrors the interaction between long-term and working memory. This architecture goes beyond conversational understanding of single human-AI interactions, enabling it to understand multi-party dialogues involving several humans and AIs.

Multi-Persona Model: A hierarchical model that incorporates multiple sub-persona, each possessing a separate memory bank. These personas are individually trained and shared, ensuring data privacy and control within the scope of its memory. The primary persona integrates the unified memories of all its sub-persona. This model can be applied to an individual (personal and professional), a team (researcher and chief of staff), or an organization (CEO and Head of Marketing).

 The image is a visual diagram of a "Multi-Persona Model" by personal.ai, featuring a central AI with different personas branching out for various scopes such as personal use, team collaboration, storytelling, private brainstorming, and research assistance. Each persona is designed for specific user interactions like family, team members, journalists, and researchers.

Multi-Modal Personalization: MODEL-2 integrates the capabilities to create personal voice clones (beta) and personalized image generation (DALL-E 3 powered). The personal voice clones are created with ~2 minutes of speech and not only mimics the speaker’s voice but also reflects the speaker’s tone and personality traits. 

Multi-Channel Integrations: Personal AI can now be integrated into various communication channels with SMS (beta) and Messenger (coming soon), making Personal AI universally available and accessible. It's a substantial step towards scaling communications, both for automating team communication as well as facilitating external client interactions.

Personal AI Users and Use Cases

Personal AI is a full stack Human AI communication platform for individuals, businesses, and developers. [Watch Demo]

Tela, stock coach & consultant team and client community communication based on internal SOPs and Q&As. 

Jacob, immigration lawyer automated leads and 24/7 client replies with SMS and social media. 

Bob, law firm internal team expert knowledge and mentorship for incoming lawyers. 

Christine Ha, celebrity chef global fan interactions and personalized recipe creations and recommendations.

Image highlighting different users and their different use cases - including tela for team communication, jacob for lead auto-response, bob for law research, christine ha for fan communication
Personal AI Users and their Use Cases

What’s to Come 

MODEL-2 is just the beginning of the future of Human AI.

Over the next few weeks, we will be highlighting the different aspects of MODEL-2 capabilities, why we built them and how they are being used by real humans. We’ll be diving deep into training spaces, unified memory, channel integrations, voice cloning, and many more. 

At Personal AI, we are committed to building technology responsibly and intentionally with the pillars of user ownership, control, and sovereignty

Stay Connected
More Posts

You Might Also Like