In the podcast, Noam Shazeer, co-founder and CEO of Character.ai, discusses his vision for universally accessible intelligence and the pursuit of AGI’s first use case: AI friends.
He talks about the potential and scalability of large language models, the importance of focusing on general use cases, and the role of AI in meeting societal needs.
The Journey to Character.ai
Noam Shazeer and Daniel de Freitas co-founded Character.ai with a vision to leverage large language models and neural networks.
They aim to create a product that has a significant value application, focusing on chatbots and dialogue systems.
Single Model for Multiple Use Cases
Developing a single model that can serve across a multitude of use cases is important.
This approach avoids the creation of specific rules that do not generalize well and caters to a wide range of needs cost-effectively.
Our goal is to be like an AGI company and a product-first company and the way to do that is by picking the right product that forces us to work on the right things. – Noam Shazeer
Full-Stack Company
Being a full-stack company allows Character.ai to influence every layer of the product, from inception to final output.
It also enables the utilization of user data as feedback.
AI’s Theory of Mind
AI’s theory of mind, its ability to understand that others’ beliefs, desires, and intentions may differ from its own, is seen as an emergent property that will come with scale and is crucial for the evolution of AI.
Affordability of Large Models
The cost of computation is decreasing, making it more affordable to use even the largest trained models.
The cost of using these models will be significantly lower than the value of most people’s time, further highlighting the potential for universally accessible intelligence.
Unlimited Scaling Laws
There is an anticipation of significant improvements and efficiency in the training process of large language models.
The cost of creating these models is reducing, leading to smarter, more accessible, and interactive AI systems.
AI Friends: First Use Case for AGI
Character.ai is focusing on creating AI that can serve as virtual friends, which is seen as a perfect first use case for AGI.
The massive unmet need for companionship and the potential it holds in the entertainment industry underscores the potential of AI friends.
Generality Over Specificity
Character.ai aims to provide a human-like experience with their AI, focusing on general use rather than fine-tuning for specific domains or use cases.
This approach aligns with the generality of large language models.
A Product-First AGI Company
Character.ai aims to be both an AGI company and a product-first company.
The strategy is to choose a product that forces them to work on things that generalize, meet people’s needs, and can be served at a massive scale and at a low cost.
Avoiding Mission-Critical Use Cases
Staying away from mission-critical use cases avoids the temptation of writing specific rules and doing things that won’t generalize well.
Focusing on a single model that serves across a number of use cases is the preferred approach.
Vertically Integrated Model
Character.ai has chosen to build a vertically integrated model and app company.
This approach is seen as the right product for the right goal, pushing the technology forward quickly and efficiently.
There are billions of lonely people out here so like it’s actually a very cool problem and a cool first use case for AGI. – Noam Shazeer
Dawn of Universally Accessible Intelligence
AI technology has reached a point where it’s not only functional but also increasingly scalable.
This development paves the way for universal access to intelligent systems.
