Current AI Architectures have a variety of sizes and are rapidly evolving. But under the hood, the vast majority of their computing cycles when processing images are dedicated to convolution in one form or another.

Multimodal architectures will accelerate this trend. The next generation will increasingly concentrate on the largest currently unprocessed data … images.

But digital convolution is hard, which is why image pixels are largely ignored.

What if there were a better way?

A way that…

• • • was thousands of times faster on a thousandth of the power?
    • supported even the latest, most advanced AI architectures
    • integrated transparently with current AI tools
    • seamlessly combined the flexibility of a fully-programmable digital system with the inherent advantages of processing with Light
    • supported both Training and Inference?
    • performed super-speed array operations for advanced algorithms, not only image processing

 

Meet Look Dynamics’ PHOTONIC NEURAL NET

THE FUTURE OF MULTIMODAL AI

Unlike current Artificial Intelligence implementations using digital spatial convolutions, Look Dynamics’ Photonic Neural Net (PNN)  harnesses the ultimate parallelism of photons using optical Fourier transforms to enable processing of any digital data normally processed by CNNs. It offers much higher speed and power efficiency than even the fastest GPUs or custom Neural Network ASICs.

The PNN supports all existing CNN architectures and training methods. Except for the fact that they are calculated in a photonic Fourier space and are inherently more accurate, the convolutions are the same as those computed by traditional digital methods. Dedicated on-chip circuitry supports pooling,
ReLU, thresholds, deconvolution flags and all other linear and non-linear operations to fully implement any CNN architecture. Nothing to change and nothing to learn.

But there is more.

In addition to convolutions, the PNN is built on a silicon analog array device that implements super-speed array operations, combining nearly instantaneous image-processing with next-generation permutational algorithms to implement Large Graphing Models, which will leave Large Language Models in the dust.

The key differences between Look’s technology and current digital approaches are speed, power, and size. Reflecting an image off of a modulator is the fastest possible way to calculate a convolution, giving the PNN full resolution parallelism at the speed of light. Combined with its analog array architecture, where every data element is retained on-chip in the ideal location for the next stage, it is nearly 100% efficient.

A THOUSAND TIMES FASTER
A THOUSANDTH OF THE POWER

The PNN input image of 4K x 4K can be up to sixty-four layers deep, supporting everything from monochrome (1 layer) to RGB (3 layer) to hyperspectral (up to 64 layer) to non-image data blocks. The PNN’s execution speed is dependent on the Neural Net architecture for which it is configured. For example, a simple VGG-16 configuration executes in three microseconds while a Mask R-CNN + ResNet-152, with many more terms, takes about five microseconds.

Other architectures will yield similar results. In comparing performance, remember that these times are for full 4K resolution images. Total latency is typically around five microseconds. All this with a single PNN module.

Power consumption is extraordinarily low since all of the “heavy lifting” convolutions and sums are completely analog full-frame photonic calculations. Typically the module consumes less than five watts regardless of the configuration architecture.

WHAT MULTIMODAL AI
WAS MEANT TO BE

The Look Dynamics Photonic Neural Net can handle even the heaviest Data Center loads.

For AI computations, four Look Dynamics PNN Modules in a 1U rack tray could achieve a peak performance of 28 exaFLOPS. Realistic external I/O limitations will probably reduce this by an order of magnitude to “merely” 2 to 3 exaFLOPS consuming 40 watts (mostly I/O power).

A single rack of PNN modules with their associated digital computers to feed data and return the results are easily the equivalent of even the fastest Super Computers when calculating AI algorithms..

MASSIVE AI WITHOUT COMPROMISE
THE FUTURE OF MULTIMODAL AI