Would Plato Encourage G-AI and Photonics Today?

Written by Michael Lebby, CEO of Lightwave Logic Inc.

Plato is widely recognized as a leading figure in philosophy. People often say that Plato’s most significant contribution was his concept of ideas, which he developed to address the problems of universals and the study of human perception and perfection. 

It would be interesting to have insight into what Plato would think about our world today with the massive advancement of neural networks, machine learning, and artificial intelligence. Perhaps Plato would be pushing the field into new and innovative forms and warning us of the merits and demerits of Generative Artificial Intelligence (G-AI) on society.

This article questions how G-AI will affect hyperscale data centers. More specifically, I will explore the impact of integrated silicon photonics chips – otherwise known as photonics integrated circuits (PICs). The article also considers electro-optic polymer, a new technology platform that changes how we look at silicon photonics. Electro-optic polymers are organic polymer materials that enable high-speed optics and reduce power consumption for the internet in devices called modulators. 

G-AI is an electronic-based computing solution. Generally speaking, it is an approach to increase computational processing, i.e., allowing semiconductor ICs to process data faster and more efficiently. As an industry, we use photonics (and optics in general) to send information processed by MPUs and GPUs using fiber optic cables. These fiber optic cables form the interconnect architecture for the internet and optical network. Electronics does the computational processing and will likely continue to do so for the next decade or two, while photonics helps convey enormous amounts of generated information optically through fiber optic cables. Popular photonic components in fiber optic communication interconnects include laser diodes, modulators, and photodetectors. These components are now becoming integrated into PIC[1] chips – typically one PIC chip to send or transmit data and another PIC chip to receive data. PIC chips are found in pluggable optical transceiver modules used in hyperscaler-based switches and router equipment within data centers.

Related:Greeking Out: Aristotle's Guide to AIOps Mastery

Growing Impact of G-AI in Recent Years

While machine learning and neural networks have been a major center and focus for computational research over the past three to four decades, it has only been in the last few years that most of us have become aware of G-AI. We have learned to accept that G-AI will drive lots of traffic on the internet, mostly because of us: the users experimenting and figuring out innovative ways to use new applications.

Related:The Changing C-Suite: Chief AI Officer In, Chief Diversity Officer Out

Data Centers and the Internet

From a hyperscale data center perspective, the growing demands of increased traffic and data driven by G-AI are changing how the internet operates, leading hyperscaler companies to invest in new hardware equipment. In terms of photonics, we can now see that data center operators have ignored the assumed next incremental photonics bandwidth standard of 400 Gigabit per second that was talked about in 2021 and 2022 and are focusing on 800Gbps, 1.6Tbps (or 1600Gbps), and 3.2Tbps today. Just two years ago, most market analysts covering data centers (and more specifically pluggable optical transceiver modules) forecasted strong growth in 400Gbps modules as the main vehicle for client-side hyperscaler traffic. At that time, data centers were looking at 4-channel 100G (for an aggregate of 400Gbps) and 8-channel 50G (also for an aggregate of 400Gbps) as solutions to support 400Gbps traffic. This trend has changed substantially: Market forecasts for 400G are flat, and the current focus is on 800G modules for hyperscalers. At the 2024 Optical Fiber Communications Conference and Exhibition, it was evident that almost all photonic suppliers must now demonstrate 200G lanes for potential 4-channel 800Gbps modules. Further, hyperscalers were calling out for technologies that would take them quickly and efficiently to 400Gbps lanes and even 800Gbps lanes (where a 4-channel 400Gbps per lane module creates an aggregate 1.6Tbps, and a 4-channel 800Gbps per lane achieves 3.2Tbps modules). 

Related:AI Assistants: Picking the Right Copilot

Creating a 4-channel 200G lane module is a challenge for the industry. A significant part of the design factor involves optical modulators. An optical modulator generally switches and modulates light. Millions of these devices exist on the internet or optical network architecture today. The current semiconductor incumbent solutions are struggling to deal with higher data rates, higher traffic volumes, low power requirements, and small sizes that are necessary to cope with the rising popularity of G-AI. This challenge has caused the industry to look for faster optical modulators operating with low voltage levels[2] to reduce power consumption. Hyperscalers are looking for optical modulators that not only can operate at 200G lanes but do this at sub-1V levels.  

These metrics – higher speed, lower power, and smaller size – create substantial challenges for the electronics and photonics industries. The electronics industry is exploring addressing this via linear pluggable optics (LPO) and linear receive optics (LRO) to reduce power consumption. Meanwhile, the photonics industry addresses these challenges through faster, lower power, smaller, and easier-to-fabricate optical modulator technology for PICs. 

An exciting solution is the use of electro-optic polymers for polymer-based optical modulators. Polymers offer a heterogeneous solution for a PIC, as the material is organic (non-semiconductor) and can be spun or dropped onto an existing silicon wafer in a standard silicon fabrication plant. This is an easier implementation of a new technology into an existing process in a standard silicon fabrication plant. Already, polymer modulators have enhanced and ‘turbo-boosted’ silicon photonic PIC platforms with 1V drive 200G lanes. Future optical modulator designs will address 400G and 800G lanes. These results and the roadmap to 400G and 800G with polymer modulators exceed incumbent semiconductor technologies used on the internet today. With polymer optical modulator 3dB bandwidths exceeding 100GHz, which have been measured to over 250GHz, electro-optic polymers are well positioned to enable 800G, 1.6T, 3.2T, and speeds beyond over the next decade. In addition, polymer modulators with drive voltages in the sub 2V and sub 1V range provide power savings that hyperscalers always look for. Lastly, with electro-optic modulator physical device structures such as the silicon slot, footprint sizes are tiny, which easily accommodates many different form factors for pluggable optical transceiver modules, both from a module design philosophy as well as a co-packaged, on-board optics design philosophy. 

Electro-Optic Polymer Modulator PICs

Electro-optic polymers made from an advanced and mature organic material enable the manufacturing of optical modulators to replace existing semiconductor and lithium niobate modulators used in today’s internet. 

Lightwave Logic Inc. sources and creates organic materials to produce electro-optic polymers called Perkinamine®. The company starts with its own proprietary designed organic chromophores, which are a key ingredient of polymers, and these are deposited onto a silicon chip to add an optical modulator function. In fabricating these modulators, a high voltage is applied briefly to align the polymers, enabling ultra-fast modulation at ultra-low power. 

Polymer modulator devices are formed on a silicon-based chip roughly a few millimeters on each side where the Perkinamine® chromophore electro-optic polymer is deposited. These chips represent the engine of a fiber optic transceiver, which is a key component for data center switches and routers. One way to visualize a high-performance optical engine is to think about automotive vehicles. For example, polymer modulators represent upgrading a four-cylinder engine to a V8 engine while keeping the whole network infrastructure the same. By turbo-charging the fiber optic modules, the other parts of the data center infrastructure remain in place. These fiber optic modules go into large metallic racks that are typically 6’ to 8’ high, about 2’ wide, and 2’ deep. Each rack forms the data center system architecture and can include routers, switches, and memory units.

How Polymer Modulators Enable G-AI

The potential performance of new photonics technologies will help data center operators meet the demands of G-AI. Further, a new technology platform that can turbo-boost existing modules without changing the network architecture and infrastructure is a low-cost OpEx, CapEx approach to speed up the internet, control power consumption, and keep the footprint and size the same. By providing higher-performance optical components such as electro-optic polymers, the internet is in a better and stronger position to absorb the growing wave of data, information, and traffic. As G-AI rises in popularity, the demand for high performance in optical components will increase. The potential of polymer technology, which currently operates at 200G lanes with 1-volt operation and has a roadmap to 400Gs and 800Gs at low voltage levels, will allow hyperscale data center operators to stay competitive with their equipment upgrades. One question today is: Can these operators afford to ignore the signs that G-AI is bringing with increased machine learning and higher levels of computational processing? I very much doubt it.

Plato’s Perspective and Summary

In addressing the original question of this article: Would Plato encourage G-AI and photonics today? Absolutely. What makes us so sure? G-AI, machine learning, and neuro-networks explore different forms of our lifestyle, and it is clear we, as a society, are on a new frontier. I believe that Plato would have been ecstatic, thrilled, and even involved in the trend enabled by modern-day electronics, software, and photonics.

As G-AI changes our lives, it is healthy to look back to where we have come from. From Plato’s perspective, our future could not be brighter. We have grand challenges ahead, the motivation to use our toolkits to meet those challenges, and a driving force to change our daily lives.

[1] PIC is an abbreviation for Photonic Integrated Circuit, i.e., a chip that contains multiple devices and functionality.

[2] Low voltage is below 2V and ideally below 1V to allow for direct drive by generic CMOS IC chips as opposed to dedicated driver chips.