IBM Accelerates AI Model Using Quantum Processor - ForkLog: Cryptocurrencies, AI, Singularity, Future

# IBM Accelerates AI Model Using Quantum Processor

Researchers from Multiverse Computing announced a quantum enhancement of a large language model on IBM hardware. It involves a hybrid scheme using a 156-qubit Heron processor.

The authors called the experiment the first "end-to-end quantum enhancement" of an LLM on a superconducting processor for autoregressive text generation.

In the tests, they used Meta's Llama 3.1 8B. The base model was not fine-tuned: parameters were "frozen" and quantum adapters—Cayley-parameterized unitary adapters (CUA)—were added. First, they trained them in a classical manner, then integrated them into a hybrid quantum-classical scheme.

The experiment was conducted on IBM Quantum System Two—an architecture for hybrid quantum systems. The setup involved a 156-qubit Heron chip.

The hybrid version reduced the perplexity of Llama 3.1 8B by 1.4%. To achieve this, about 6,000 parameters were added—roughly 0.000075% of the model size.

During the demonstration, the quantum-enhanced Llama correctly answered questions about astronomy and biology that the base version could not handle (for example, about the presence of rings around all giant planets).

According to the lead author of the study, Borhi Aispuru, the work is a proof of concept. Quantum blocks allowed for more accurate prediction of the next token in the text with minimal computational resource costs.

The team aims to further reduce perplexity and increase accuracy with fewer parameters compared to fully classical approaches.

Recall that in May, the stock prices of quantum companies rose after the US Department of Commerce announced a $2 billion allocation to American firms under the CHIPS R&D program.