MIT unveils system using light for computations, achieving 100-fold energy efficiency boost and 25-fold compute density improvement


  • MIT has unveiled a groundbreaking system that employs light for computations, achieving a remarkable 100-fold improvement in energy efficiency and a 25-fold increase in compute density.
  • This innovation has the potential to supercharge machine-learning models, paving the way for powerful applications beyond the capabilities of current technology.
  • The system utilizes micron-scale lasers, which could empower small devices like cell phones to perform complex computations.
  • Its scalability and compatibility with existing fabrication processes make it a commercial possibility in the near future.
  • MIT’s quantum leap builds upon years of collaborative research and addresses the limitations of traditional digital technologies in machine learning.
  • Optical neural networks offer unparalleled energy efficiency and the ability to transmit vast amounts of data in confined spaces.
  • Challenges like energy conversion and bulkiness are overcome through a compact architecture using vertical surface-emitting lasers (VCSELs).
  • This technology holds the potential to accelerate AI systems and revolutionize machine learning, including models like ChatGPT.

Main AI News:

MIT has been at the forefront of groundbreaking technological advancements, and once again, it has not failed to impress. In a recent issue of Nature Photonics, a remarkable development was unveiled, promising to revolutionize the world of machine learning. This innovation, pioneered by an MIT-led team, harnesses the power of light to drive computations, catapulting efficiency to previously unimaginable heights.

The traditional digital systems that underpin contemporary machine learning are approaching their limits, both in terms of computational capacity and energy consumption. Enter MIT’s quantum leap: a system that leverages the movement of light, rather than electrons, to perform computations. The significance of this achievement cannot be overstated, with the team reporting a staggering 100-fold improvement in energy efficiency and a 25-fold boost in compute density, firmly establishing it as a game-changer.

The heart of this cutting-edge system lies in its use of hundreds of micron-scale lasers, marking a monumental shift away from traditional electronic-based computing. The implications are profound. As the researchers themselves point out, this innovation ushers in “an avenue to large-scale optoelectronic processors to accelerate machine-learning tasks from data centers to decentralized edge devices.” In simpler terms, this technology has the potential to empower everyday devices, such as cell phones, with the capability to run programs that were once the exclusive domain of massive data centers.

Moreover, what sets this innovation apart is its scalability. The components of this system can be seamlessly integrated into existing fabrication processes, including those already utilized in cell phone face ID and data communication. Zaijun Chen, the first author of this groundbreaking work, envisions that this technology could become commercially viable within a few short years.

Dirk Englund, an associate professor at MIT, offers a tantalizing glimpse into the future: “We don’t know what capabilities the next-generation ChatGPT will have if it is 100 times more powerful, but that’s the regime of discovery that this kind of technology can allow.” As the leader of MIT’s Quantum Photonics Laboratory, Englund is at the helm of a revolution that promises to redefine the possibilities of machine learning.

This achievement is but the latest milestone in a remarkable journey. The work represents a culmination of years of relentless research and innovation by Englund and his team. From theoretical groundwork in 2019 to the practical demonstration we see today, the path to this moment has been paved with determination and expertise.

This journey has been possible through the collaborative efforts of talented individuals, including Ryan Hamerly, who played a pivotal role in both the theoretical and practical aspects of this endeavor. The list of coauthors reads like a who’s who of experts in the field, underscoring the significance of this achievement.

In a world where deep neural networks are the backbone of cutting-edge machine learning, the limitations of existing digital technologies are increasingly apparent. The voracious appetite for energy and the confinement to colossal data centers necessitate a fresh approach. MIT’s quantum leap, with its reliance on optical neural networks, offers a lifeline.

The potential here is boundless. By harnessing light instead of electrons, this technology promises not only unprecedented energy efficiency but also the ability to transmit vastly greater amounts of information across smaller spaces. While optical neural networks face challenges of their own, such as energy conversion and bulkiness, the current work deftly tackles these issues head-on.

The secret lies in a compact architecture built upon state-of-the-art vertical surface-emitting lasers (VCSELs). These lasers, developed by the Reitzenstein group at Technische Universitat Berlin, hold the key to a future where high-speed optical neural networks are not just a dream but a reality.

As Logan Wright, an assistant professor at Yale University, aptly puts it, “The work by Zaijun Chen et al. is inspiring, encouraging me and likely many other researchers in this area that systems based on modulated VCSEL arrays could be a viable route to large-scale, high-speed optical neural networks.” While challenges remain, the potential to accelerate AI systems, including popular textual ‘GPT’ models like ChatGPT, is nothing short of transformative.


MIT’s groundbreaking achievement in machine learning efficiency, using light-based computing, has the potential to disrupt the market significantly. It promises to unlock new possibilities for AI applications, making them more energy-efficient and accessible on smaller devices. As this technology becomes commercially viable, it could lead to a wave of innovation in various industries, from mobile devices to data centers, with the potential to reshape the landscape of AI and computing as a whole.