English

Scientists demonstrate a new optical neural network training method that can crush electronic microprocessors

200
2023-09-27 15:24:41
See translation

The current deep neural network system (such as ChatGPT) can quickly improve energy efficiency by 100 times in training, and "future improvements will greatly increase by several orders of magnitude. Scientists from MIT and other institutions have demonstrated a new optical neural network training method that can crush state-of-the-art electronic microprocessors.

Moreover, the computational density of the demonstrated system is about two orders of magnitude higher than that of Nvidia, Google, or Graphcore systems.

Basically, this means that the most advanced models can be trained with 100 times less energy and occupy less space at the same speed.

Artificial neural networks mimic the way biological brains process information. These artificial intelligence systems aim to learn, combine, and summarize information from big datasets, reshaping the field of information processing. Current applications include images, objects, speech recognition, games, medicine, and physical chemistry.

The current artificial intelligence model has reached hundreds of billions of artificial neurons, showing exponential growth and posing challenges to current hardware capabilities.

This paper demonstrates that optical neural network (ONN) methods with high clock speed, parallelism, and low loss data transmission can overcome current limitations.

Our technology opens up a path for large-scale optoelectronic processors to accelerate machine learning tasks from data centers to decentralized edge devices, "the paper wrote.

The ONN method is expected to alleviate the bottlenecks of traditional processors, such as the number of transistors, data mobility energy consumption, and semiconductor size. ONN uses light, which can carry a large amount of information simultaneously due to its wide bandwidth and low data transmission loss. In addition, many photonic circuits can be integrated to expand the system.

In order to move light for calculation, the team led by MIT utilized many laser beams, which were described as "using mass-produced micrometer scale vertical cavity surface emitting lasers for neuron coding".

The researchers explained, "Our scheme is similar to the 'axon synapse dendrite' structure in biological neurons
They believe that the demonstrated system can be expanded through mature wafer level manufacturing processes and photon integration.

Dirk Englund, Associate Professor of Electrical Engineering and Computer Science at the Massachusetts Institute of Technology and the head of this work, explained to SciTechDaily that the size of models such as ChatGPT is limited by the capabilities of today's supercomputers. Therefore, training larger models is not economically feasible.

He claimed, "Our new technology can make it possible to cross machine learning models, otherwise it would not be possible in the near future.

This paper titled "Deep Learning Using Coherent VCSEL Neural Networks" was published by a large team of scientists. This work has received support from the Army Research Office, NTT Research, and NTT Netcast Awards, as well as financial support from the Volkswagen Foundation. The three researchers of the team have applied for patents related to this technology.

Source: Laser Network

Related Recommendations
  • Chuangxin Laser Industry Dedicated Laser and Solutions Help Promote the Intelligent Development of Cladding Application Industry

    Laser cladding technology, also known as laser additive manufacturing technology, uses high-energy laser as the heat source and metal alloy powder as the cladding material. Through the synchronous action of laser and alloy powder on the metal surface, it quickly melts to form a molten pool, and rapidly solidifies to form a dense, uniform, and controllable thickness metallurgical bonding layer, the...

    2023-11-01
    See translation
  • Tsinghua University has made progress in the field of magnetic field and laser composite processing

    The National Key Laboratory of Interface Science and Technology for High end Equipment at Tsinghua University has made progress in the field of magnetic field and laser composite processing - magnetic field assisted laser shock strengthening of Ti6Al4V alloy. The relevant research was published as a cover article titled "Magnetic Field Assisted Laser Shock Peening of Ti6Al4V Alloy" in the journal ...

    2023-09-16
    See translation
  • Invest 13 million euros! Tongkuai opens its Southeast European headquarters in Hungary

    Recently, German company Tongkuai invested 13 million euros to open its headquarters in Southeast Europe in Hungary and jointly established a digital network demonstration factory in the Gothler Business Park. Its business focuses on machine tools for digital manufacturing and laser sales for batteries and other automotive components.Nicola Leibinger Kamm ü ller, CEO of Tongkuai, said, "It is...

    2023-09-16
    See translation
  • Internationalization Strategy Enters Stage 2.0 | HSG Hsglaser Thailand Manufacturing Base Holds Grand Opening

    At 9:00 am local time on June 26th, the opening ceremony of Hsglaser Thailand Manufacturing Base was grandly held in Bangkok Industrial Park, Thailand. This not only marks a significant expansion of Hsglaser's global strategic map, but also signifies that its international layout has officially entered a new 2.0 stage, and is another important milestone for Hsglaser to showcase its outstanding str...

    2024-06-27
    See translation
  • The technological iteration route of automotive millimeter wave radar chips

    The rapid development of intelligent cars and autonomous driving technology has made millimeter wave radar inconspicuous, and the widespread application of millimeter wave radar has driven the technological evolution of MMIC.From the expensive gallium arsenide (GaAs) process in the early days to the mainstream CMOS and SiGe processes today, and then to the future promising FD-SOI process, the cont...

    2024-12-07
    See translation