AI is coming to an end… devices

Cyber.Fund sponsors TinyML Foundation

The TinyML Foundation is a non-profit organization committed to accelerating the growth of a global community focused on energy-efficient machine learning technologies. 

With over 19,000 meetup members across 38 countries, significant educational initiatives, and partnerships with academia and industry, the foundation is at the forefront of advancing AI at the edge. 

Cyber.Fund’s partnership with the TinyML Foundation aligns with our vision of a cybereconomy, where AI and advanced technologies democratize entrepreneurship and integrate seamlessly with daily life. 

By supporting the foundation, we are investing in a future where AI-driven solutions enhance the real economy, creating a healthier and more sustainable environment for all.

The partnership rationale for Cyber.Fund consists of at least three main points:

1. A significant majority of the global economy, potentially over 70-80%, is driven by offline, physical-world interactions – often referred to as the "real economy". Therefore extracting value at scale using AI requires integration with the real world. Sensors and robotics bridge the digital and physical realms, enabling AI systems to perform tasks ranging from simple monitoring to complex autonomous operations. This integration is crucial for industries such as manufacturing, healthcare, agriculture, and logistics, where AI-driven automation can significantly enhance efficiency, accuracy, and scalability. 

2. Communication-efficient models training is extremely important to enable decentralized AI. We believe that a significant part of expertise in this field is accumulated within TinyML foundation. Early work in this direction produces extremely promising results with 1B+ parameters size models having been trained over internet-grade networks (as opposed to the datacenter-grade networks). Techniques like forward-forward training, pipeline parallelism, offloading, tiling, etc… reduce the communication overhead significantly, making decentralized AI feasible with low-bandwidth connections. 

3. For every $1 spent on a GPU, roughly $1 needs to be spent on energy costs to run the GPU in a data center. Performance per watt is the main efficiency metric for large scale deployments today. Currently training is a by-product of electricity-to-heat conversion and we believe TinyML community is working hard to change it by introducing low-power hardware and energy-efficient AI architectures such as spiking neural networks. 

At Cyber.Fund, we are driven by the vision that convergence of advanced technologies will empower individuals to reach unprecedented levels of productivity and innovation. It also includes innovations in the public sector, financial institutions and markets, etc...

We seek first-of-a-kind solutions that create their own blue ocean markets, gaining a first-mover advantage. Investing in pioneering technologies that redefine market boundaries aligns with our vision of fostering innovation that transcends existing market limitations.