Revolutionizing AI Performance Through Encryption
At a recent summit focused on artificial intelligence, Cornami and DESILO joined forces to unveil a cutting-edge technology that brings encrypted AI into the mainstream. This new approach utilizes fully homomorphic encryption (FHE), enabling large language models (LLMs) to process sensitive information while the data remains in encrypted form. The result is a significant advancement in both speed and accuracy for AI operations.
Security Meets Speed: Resolving the AI Enigma
For many years, organizations in various sectors, including healthcare and finance, have grappled with a pressing issue: the challenge of safeguarding data without sacrificing the performance required for effective AI deployment. This partnership marks a critical turning point, offering solutions that allow encrypted AI inference to be executed almost as quickly as plain text operations.
As Dr. Craig Gentry, noted as one of the foremost minds behind FHE, shared insights on this breakthrough: "For decades, FHE was thought unfeasible for real-world use due to its speed. However, with our efforts to speed up this encrypted computation, we are demonstrating that it's possible to achieve both privacy and performance in AI systems."
Innovative Techniques Propelling the Industry Forward
One of the key advancements facilitating this leap is the issue of plaintext ciphertext matrix multiplication (PCMM). This innovative technique significantly impacts over 90% of computations within LLMs, enabling the secure execution of these operations without incurring substantial overhead costs. By combining this with enhanced scalability through Cornami's proprietary Compute Fabric, they have delivered performance enhancements that far exceed traditional methods of encrypted computation.
Dr. Gentry further elaborated, stating, "PCMM is pivotal in making privacy-preserving AI feasible. It ensures that core operations, like matrix multiplication, can be executed efficiently within an encrypted environment. This transition effectively narrows the performance gap between encrypted inference and its unencrypted counterpart, while bolstering compliance and security measures across the board."
Applications Across Industries
The implications of these advancements are vast, particularly in sectors like healthcare. Clinical trials generate a substantial amount of data that must be both secure and readily available for analysis. This innovation allows researchers to examine encrypted patient records and imaging data without ever exposing confidential information, thus streamlining research while maintaining privacy.
Moreover, similar methodologies can be applied in finance, government, and cloud services, where stringent compliance with security protocols is essential. Seungmyung Lee, CEO of DESILO, remarked, "Our guiding principle has always centered around never decrypting sensitive data. Our collaboration with Cornami brings this principle to reality, setting the stage for enterprises to truly unlock high-value data without compromising security. This partnership aligns perfectly with the future release of our HARVEST platform, focusing on global healthcare collaborations."
About Cornami and DESILO
Cornami
Cornami stands out as a leader in performance for scalable computing technologies. Their innovative framework is engineered to effectively leverage advanced encryption systems like fully homomorphic encryption, meeting the dual demands of enhanced security and operational scalability.
DESILO
Operating from Korea, DESILO is a rapidly expanding privacy-enhancing technology startup making strides globally, with initiatives underway in a multitude of regions including North America and Europe. Their core offerings revolve around creating secure collaboration solutions embedded with FHE and federated learning, ensuring that sensitive information remains confidential at all times. The upcoming HARVEST platform will further enhance capabilities for secure multi-party data interactions within healthcare and beyond.
Frequently Asked Questions
What is the significance of fully homomorphic encryption in AI?
Fully homomorphic encryption (FHE) allows computations to be performed on encrypted data, enabling the analysis of sensitive information without exposing it, thus enhancing privacy and security in AI applications.
How does PCMM contribute to AI performance?
Plaintext ciphertext matrix multiplication (PCMM) streamlines essential AI computations, enabling operations to be performed efficiently and securely while maintaining high performance standards.
What sectors can benefit from this technology?
Healthcare, finance, government, and cloud services are key sectors that can significantly benefit from these advancements, as they require rigorous data protection along with dependable performance.
What is the HARVEST platform?
The HARVEST platform, set to launch shortly, will facilitate secure multi-party collaborations, particularly in the healthcare sector, ensuring that encrypted data remains protected while still allowing for valuable insights.
Who leads the technology initiatives at Cornami?
Dr. Craig Gentry, a prominent figure in the field of fully homomorphic encryption, leads the innovation efforts at Cornami, driving advancements that bridge the gap between security and performance.
