Cartesia claims its AI is efficient enough to run pretty much anywhere

Cartesia, a startup co-founded by Karan Goel and Albert Gu, is developing state space models (SSMs) as a more efficient alternative to transformer architectures in AI. SSMs compress data into summaries, allowing them to handle large amounts of data more efficiently than transformers, which could reduce AI operational costs. Cartesia's Sonic, an SSM-based voice cloning tool, demonstrates the potential of SSMs in handling long-context data like audio. Despite ethical concerns regarding data usage and voice cloning, Cartesia is gaining traction with customers and has secured $22 million in funding to further develop its technology.

Cartesia aims to become a leader in multimodal AI models that can operate on various devices, offering real-time intelligence across different data types. The company is expanding its offerings with products like Sonic On-Device for mobile applications and Edge, a software library for optimizing SSMs. With a vision to challenge traditional transformer-based models, Cartesia is positioning itself to drive the next wave of AI innovation by providing faster, cost-effective, and scalable AI solutions.

Key takeaways:

```html

Cartesia, co-founded by Karan Goel and Albert Gu, is developing state space models (SSMs) as a more efficient alternative to transformer architectures for AI.
SSMs compress prior data into summaries, allowing them to handle large data volumes more efficiently than transformers, which could reduce AI operational costs.
Cartesia's Sonic model, a voice cloning tool, has raised ethical concerns due to its training data and potential misuse, but the company is working on moderation and verification systems.
Cartesia recently secured $22 million in funding, aiming to expand its SSM technology for real-time, multimodal AI applications across various industries.

```

Cartesia claims its AI is efficient enough to run pretty much anywhere | TechCrunch

Key takeaways:

Comments (0)

Newsletter