WHAT’S NEXT?

“To move beyond carbon-based economies, beyond being trapped indoors because of a tiny virus, in order to have agency over our biology, our environment, industrial processes, and agricultural processes, we need better kinds of computation that can answer impossible questions,” as Peter Barrett puts it. “Today we’re fumbling in the dark, guessing at stuff and butterfly collecting rather than designing and engineering things.”

In 1950 the world’s population stood at roughly 2.5 billion. Now it’s 8 billion, and will reach 10 billion by 2058. The strain on Earth’s natural resources is already evident, of course, with soaring global food shortages, persistent drug shortages, and acute climate change events growing in frequency and magnitude – and in lockstep with carbon-based approaches to supporting humanity. Researchers and policymakers need significantly more powerful computational methods to solve these problems, but first computer scientists must simultaneously find new ways to squeeze better performance out of traditional silicon-based frameworks (now that Moore’s Law is a thing of the past), while also speeding the advent of quantum computing. All the while, this research must also account for the growing environmental impact of humanity’s appetite for data and computing, given the vast amounts of energy required to deliver it. The good news: quantum computing is close to becoming a reality, while strong incentives continue to drive innovation in silicon-based frameworks.

Quantum computing, advances in artificial intelligence and machine learning, and photon-based advances in traditional computing architecture.

Quantum moves toward the starting line; older technologies get incremental boosts.

“On quantum, we won’t have a million qubit machine running, but you will have an algorithm that’ll allow us to understand, say, the method of action of our most important drugs,” Barrett said. We’ll know the impossible questions across a bunch of different domains we’ll be able to answer, and how transformative those quantum machines will be. We have some examples of that, but I think we’ll see very concrete examples in therapeutics, battery chemistries, how we design superconductors – specific examples of what you pay for and what you do with it.  See how Mercedes plans to work with PsiQuantum to develop the next generation of EV batteries. So you’ll see an increasing focus in the Fortune 50 or Global 2000 on doing the preparatory work for when the quantum machine shows up in a handful of years. One of the biggest surprises over the last couple of years is that companies like Goldman Sachs now has 30 of the smartest quantum engineers on the planet. A lot of these companies now have serious people who are domain experts looking to find a quantum advantage.”

“That relentless and expensive march, like ‘Let’s go to 5 nanometers to 3 nanometers,’ will continue,” Matt Hershenson said. “But we’ll also see innovation that’s orthogonal and complementary to that, and silicon photonics and chiplets are two examples.” While “chiplets” allow computers to more efficiently allocate and execute subsets of computational tasks, optical interconnectivity moves data more quickly and demands less energy than traditional, copper-based data transport. Both exemplify the coming seismic shift from electron-based to photon-based computing, sometimes collected under the “silicon photonics” label.

“Some of these things will be in limited use next year, and much more broadly available the year or so after,” Hershenson said. “That’s important because the scale of computing is often limited by those two things – power consumption and how fast data can move around. So data centers will start to become more efficient and higher performing.”

“In five years we’ll have a working quantum machine,” Barrett said. “At that point you’ll see the most immediate changes happen in areas that won’t require a lot of infrastructure change: there’ll be discoveries in financial services like in portfolio management, for instance, or the way we schedule flights. Now quantum will certainly bring huge changes in pharmaceuticals, and help us discover materials for decarbonization. We’ll be able to plan on solar cells being twice as efficient, or having 50-megawatt windmills. It’ll just take longer to build the infrastructure to help make those kinds of discoveries possible.”

Hershenson: “Early movers will increasingly enjoy the benefits of silicon photonics, with mainstream enterprises watching closely from the sidelines and preparing to enter the fray as the technologies mature and industries shift their operations and infrastructure to suit the new technologies, and the possibilities offered. The overall impact of these advances will be significant, however. With major cloud-computing giants like Amazon, Google and others shifting to these technologies, they’ll be better positioned to serve the world’s ever-growing appetite for computing power more efficiently (read: with a lower carbon footprint) and more quickly. “When data movement and execution improve by tenfold, that will certainly have benefits in a lot of applications,” Hershenson said. “It’s just hard to be specific right now about what problems will be solvable based on that.”

Quantum machines will be yielding significant advances in computational theory and methodology, which will themselves produce advances across a range of disciplines and industries. “Ten years from now we will have superconducting materials. We will have ways of making photo fuels,” he said, referring to fuels created by harnessing the power of photosynthesis. “We will understand the mechanisms by which our important drugs work, and we’ll have designed drugs that can treat maladies that we currently don’t have good answers for. We will have far superior battery technology. We will have the ability to make renewables with very large scale and reliable storage, so that wind and solar can completely replace gas. We will have ways of using natural gas in a way that doesn’t emit carbon. We will have real scalable mechanisms for carbon sequestration. In other words, we’ll have the mechanisms and the materials we need for energy transformation.”

Market forces will of course direct some of the innovation, Barrett said. “If Elon wants to make 20 million cars a year, it’s $120 billion of metal, and we don’t know where most it is coming from. So we need to be able to mine more efficiently, and make steel without carbon. All of those processes demand quantum design of catalysts and materials. And we may have entirely new classes of materials which obviate the need for much of that metal. We’ll have things like potassium batteries instead of lithium batteries, and potassium is 1000 times more ubiquitous.”

“We’re bad at imagining things that don’t exist. But we are on thresholds of materials and chemistries and capabilities that we just can’t get our heads around.”

Innovations in exotic materials that allow for computation in using so-called “majority logic” and using ferroelectricity, allowing computation that sidesteps the manufacturing and voltage limitations of silicon. “Majority logic relies on computational gates that work differently from those made out of transistors,” Barrett said. “Transistor-based computation relies on traditional computing logic, where an operation requires a certain number of gates. With majority logic, you have different kinds of gates: gates that turn on if three of five inputs are true, for instance, or if one of seven inputs are true. The physical representation of those gates is much smaller, so you need as little as one-tenth as much geometry on the device in order to synthesize the same amount of logic, and if you fabricate those gates with various exotic materials, you can use 100-times less power for that computation. We’re looking at companies in that space to invest in, and haven’t yet found one.”

“Methane pyrolysis is the single best approach to decarbonization, and it’s being almost completely overlooked,” Barrett said. “While existing decarbonization efforts for coal, natural gas and oil seek to capture carbon after burning it, methane pyrolysis captures carbon from natural gas before burning. Energy suppliers can fairly easily retrofit existing equipment to enable this process, making it both efficient and scalable, and given that the world’s economy will rely on natural gas for the foreseeable future, methane pyrolysis is a powerful way to give gas companies a path to decarbonization.”