As the US and China race to develop ever more powerful AI algorithms and quantum computers, European countries are growing their own quantum initiatives in the public and private sectors to keep pace.
Over the last few years, governments have taken on an increasingly critical role in supporting the growth of quantum technology, while quantum research centers advance the technology from theory to practical application. Governments have committed over $40 billion over the next 10 years globally according to Lakestar’s State of Quantum 2024 report. The analysis showed that countries must catalyze growth by fostering quantum ecosystems, supporting homegrown talent, and enabling domestic startups to access the growth capital they need.
Enrique Lizaso-Olmos, CEO of Multiverse Computing, discussed the state of European quantum and AI development with Rob Strechay as part of the Supercloud 6 event hosted by theCUBE by SiliconANGLE.
“Europe has the feeling that we are losing the tech race to the U.S. and China again and again,” Lizaso said. “We need to be working harder and bolder in a variety of fields to catch up, particularly in both quantum and AI. This is a good time for Multiverse Computing as we work at the intersection of these fields.”
Nations across Europe have accepted this challenge. The UK’s National Quantum computing Center will open this year, aiming to promote cross-sector quantum collaboration to boost the UK economy. The VTT Technical Research Center of Finland has announced a national strategy to build quantum hardware with 54 qubits by the end of 2024, focusing on micro and nanotechnology. In Switzerland, Quantum Basel focuses on Industry 4.0 and pharma, offering access to IBM Quantum’s platform and IonQ’s quantum computer. House of Quantum in Delft is a collaborative hub for quantum development, expanding access to cryogenic quantum computing. France is seeing returns on its $1.9 billion national quantum strategy and Germany is spending $3.3 billion on quantum tech and aiming to have a 100-qubit machine by 2026.
Multiverse recently won a project with the German Aerospace Center’s DLR Quantum Computing Initiative to use quantum simulation to improve the reflectivity capabilities of superconducting nanowire single photon detectors. Multiverse will work with Single Quantum, a Dutch company that specializes in photon detection hardware, on this effort.
European leaders have taken notice of Multiverse’s accomplishments. The European Innovation Council’s Scaling Club just named Multiverse as one of the highest potential deep-tech scaleups in the EU. Last month the company won the Future Unicorn Award from DigitalEurope in Brussels, delivered by the EU Commissioner Iliana Ivanova, for its potential to reach a $1 billion valuation due to its work on large language models with quantum-inspired technologies.
Multiverse also just announced an oversubscribed $27 million Series B funding round that includes plans to expand into more industries and to develop more quantum-inspired software.
The Quantum-AI Intersection
Multiverse’s approach involves creating “green algorithms” using quantum-inspired techniques to address the escalating energy consumption of AI, which is becoming increasingly relevant with the rising demand for AI applications.
In his conversation with Strechay, Lizaso explained that these versatile techniques can be applied across manufacturing, healthcare, and sports analytics, to solve specific problems like defect detection, patient outcome prediction and injury prediction. It was during a collaboration with Bosch that Multiverse researchers realized that tensor networks applied for machine learning and simulation in manufacturing could also be applied to LLMs.
We discovered that techniques from quantum computing, such as tensor networks, can be applied to classical computing problems, leading to significant improvements in efficiency and performance,” Lizaso said. “Our first applications saw over 85% reduction in size, and more than doubling the speed in both training and inference time. We are really confident that we can push these figures farther.”
By combining quantum techniques with artificial intelligence, Multiverse aims to provide solutions that are smaller, faster, and cheaper, with a particular focus on training large AI models more efficiently.
CompactifAI, Multiverse’s solution for the high compute cost of LLM’s, uses tensor networks to reduce the number of parameters in a model, reducing its overall size as well as shrinking memory and storage space requirements.
“This new tool needs to be implemented now by companies who want to make AI more sustainable,” Lizaso said. “Our customers can use the ability to compress large AI models to ease the challenges of deployment by saving on development costs and energy requirements across training, general operations and retraining.”
The Future Quantum-Enhanced LLMs
As Multiverse builds more quantum-inspired techniques for AI, the company plans to release a solution that addresses concerns about detecting and removing bad data in large language models (LLMs). Retraining models initially trained with data that is bad or IP-protected is expensive.
“This Lobotomizer makes LLM’s forget, enabling the precise extraction and elimination of undesirable information without retraining the entire model,” Lizaso said.
As quantum and quantum-inspired techniques work alongside classical techniques to further improve compression and reduce the cost of training AI models, LLM applications will become more accessible.
“Our work with quantum AI techniques is democratizing the field by making AI more affordable, faster, and efficient, leading to increased competition and innovation,” Lizaso said. “As our technology rapidly advances in step with the technology hubs and government programs across Europe, we believe the European quantum-AI ecosystem will rival those of the U.S. and China.”