European Startup Unveils 'Animal Brain' AI Models, Targeting Tiny IoT Devices
AI
Machine Learning
Startups
Small AI Models
Multiverse Computing
LLMs
TechCrunch Disrupt
8
Miniature Revolution
Media Hype
7/10
Real Impact
8/10
What is the Viqus Verdict?
We evaluate each news story based on its real impact versus its media hype to offer a clear and objective perspective.
AI Analysis:
While the hype surrounding AI is currently very high, this news represents a genuinely disruptive innovation – a focused effort to drastically reduce the size of AI models. The combination of a strong technical foundation and a clear market opportunity suggests a real, long-term impact, though it’s not a fundamental shift in AI, but a vital enabler.
Article Summary
Multiverse Computing, a Spanish AI startup focused on model compression, has released its ‘Model Zoo’ – a family of incredibly small AI models named after animal brain sizes. The flagship models, ‘SuperFly’ and ‘ChickBrain’, are designed to address the growing demand for on-device AI. ‘SuperFly’, a compressed version of Hugging Face’s SmolLM2-135, utilizes 94 million parameters, likened to a fly’s brain, and is targeted at tasks like controlling smart appliances through voice commands. ‘ChickBrain,’ a compressed version of Meta’s Llama 3.1 8B, boasts 3.2 billion parameters and provides more advanced reasoning capabilities, capable of running on a MacBook without an internet connection. Both models outperform their original counterparts in several benchmarks, demonstrating that size doesn’t necessarily equate to reduced performance. The startup’s ‘CompactifAI’ technology, inspired by quantum physics, significantly reduces model size without sacrificing accuracy. Multiverse is already in discussions with major device manufacturers like Apple, Samsung, and Sony, and offers its compressed models via an API on AWS. This innovative approach could revolutionize the deployment of AI in everyday devices.Key Points
- Multiverse Computing has released two AI models, ‘SuperFly’ and ‘ChickBrain’, designed for ultra-small devices.
- The company’s ‘CompactifAI’ technology compresses AI models without sacrificing performance, relying on quantum-inspired algorithms.
- These models can run locally on devices like smartphones and IoT devices, enabling functionality even without an internet connection.