Interview, Conference Presentation
The Chip That Could Unlock AGI.
- Naveen Rao anticipates that within approximately five years, Unconventional AI will achieve a paradigm analogous to intelligence and deploy scalable manufacturing solutions, with the company initially operating as an open-ended research phase to establish existence proofs before engineering trade-offs.
- The organization plans to release open-source materials in the near future and intends to recruit experts in energy-based models, flow models, gradient descent, system architecture, and analog and digital circuit design.
- Rao expects the first chip prototype to be one of the largest analog chips ever constructed, though he acknowledges it may not function as initially theorized.
- Strategic partnerships are expected with TSMC for close collaboration and fast prototyping turnaround, while Google will focus on internal TPU improvements; the company's relationship with NVIDIA remains uncertain regarding potential collaboration or competition.
- The company aims to develop computer systems based on dynamic time and causality to address the limitations of the current digital computing paradigm and move humanity toward AGI, potentially resolving the global energy problem through scalable manufacturing.
- Significant infrastructure risks are identified, including the fear that the 1970s-era transmission grid may fail under increasing AI loads and the difficulty of adding the estimated 400 gigawatts of new power generation capacity required over the next decade.
- Success is projected to have a generational impact recorded in history books, contingent on employing high-agency individuals capable of experimentation and learning to build necessary products.