Unconventional AI
Making computation for AI radically more efficient
Early transistors were error-prone, and classical computing began to achieve real scale when engineers learned how to group transistors together and use error codes to smooth out the noise. Today the same thing is happening in quantum computing — and the question is how many physical qubits will be needed to power each logical qubit. Iceberg Quantum has developed an architecture using quantum low-density parity check (QLDPC) codes that reduces the number of physical qubits needed for real-world computations by a factor of 10 or more. Quantum hardware companies say Iceberg’s work will speed up their march toward utility-scale quantum computing by years.
Making computation for AI radically more efficient
Building human-centric frontier AI models
Building scalable quantum computers using optically trapped neutral atoms
Building a marketplace to buy and sell short-term contracts on compute/GPU hours