Rigetti Founder Launches New Quantum Firm to Boost AI

Key Concepts

Quantum accelerated AI, AI training and inference, large language models, energy requirements, stochastic gradient descent, low-rank adaptation, qubit modalities (superconducting, neutral atoms, photonics), fault-tolerant architecture, multi-modality quantum architecture, quantum computing hardware, classical infrastructure (GPUs), cost, speed, scale.

Objective of the New Company (Secretary)

Secretary is building quantum accelerated AI servers to exponentially speed up training and inference for large language models. The company aims to address the growing energy and cost challenges associated with training these models by leveraging quantum computing.

The Problem Secretary is Addressing

• Growing Energy Requirements: Training large AI models demands significant energy, posing sustainability concerns.
• Increasing Costs: The financial burden of training these models is escalating.
• Limitations of Existing Quantum Computing Companies: Current quantum computing companies have roadmaps and technology choices that were established before AI became the dominant compute paradigm. These companies are also siloed with respect to their technology choices.

Secretary's Proposed Solution

• Quantum Accelerated AI Servers: Utilizing quantum computing to achieve exponential speedups in training and inference.
• Multi-Modality Quantum Architecture: Combining different qubit modalities (e.g., superconducting, neutral atoms, photonics) within a single fault-tolerant architecture. This approach aims to leverage the strengths of each modality.
• Laser Focus on AI Acceleration: Designing a quantum architecture specifically for accelerating AI training and inference, lowering costs, and reducing energy consumption.

Why a New Company is Needed

Existing quantum computing companies, including the speaker's previous company (Rigetti), have established roadmaps and technology choices that predate the dominance of AI. These companies are also siloed with respect to their technology choices. Secretary aims to address these limitations by:

• Building an architecture from the ground up specifically for AI acceleration.
• Combining multiple qubit modalities to achieve optimal performance.
• Focusing solely on AI as the primary application.

Multi-Modality Quantum Architecture Explained

The core idea is to integrate different types of qubits (superconducting, neutral atoms, photonics) into a single, fault-tolerant quantum computer. This allows the system to leverage the unique advantages of each qubit type, potentially leading to better performance and scalability.

The Role of Quantum Mechanics

The speaker argues that nature itself operates on quantum mechanics, especially in information-rich environments. Quantum computing, therefore, is a more efficient way of processing information at a fundamental level.

Integration with Classical Infrastructure

Quantum computers will not replace classical computers (GPUs). Instead, they will work in conjunction with existing classical infrastructure to accelerate specific tasks, such as AI training and inference.

Market Opportunity

The market for quantum accelerated AI is estimated to be a $1 trillion market.

Funding and Future Plans

Secretary is currently backed by Y Combinator and will need to raise additional capital to scale its operations. The company is in its early stages but is rapidly growing its team.

Rigetti's Response

The speaker avoids directly commenting on Rigetti's response but implies that Rigetti's architectural choices and roadmap, established before the rise of AI, may make it difficult for them to pivot towards AI acceleration. He emphasizes the importance of having the right DNA and vision from the outset.

Key Quote

"Ultimately, nature itself runs on quantum mechanics in the most information rich environments at the lowest level of nature. We use quantum mechanics to process that information in quantum reflects that mirrors that reality and ultimately is a more efficient way of computing long term."

Synthesis/Conclusion

Secretary is a new company focused on building quantum accelerated AI servers to address the growing energy and cost challenges of training large AI models. Its key differentiator is its multi-modality quantum architecture, which combines different qubit types to achieve optimal performance. The company believes that quantum computing is essential for the future of AI and that its approach will be more competitive than existing quantum computing companies due to its laser focus on AI and its flexible architecture.