Make No Small Plans: Why Venture Capital Needs to Return to Big Bets on Deep Technology

By Peter Barrett

At a time when the combustible mixture of ignorance and arrogance has led to the deaths of hundreds of thousands of Americans, venture capitalists need to focus on technologies and ideas that will move society forward. Incrementalism won’t get us there, nor will the next behavior modification network or attention economy app. We should instead commit ourselves to big bets on transformative technologies that create durable value and real progress.

We need better computers. We are a technological society that entirely depends on computation, but there is an exponentially growing gap between our computational needs and Moore’s Laws’ ability to deliver. CPUs used to double their performance every 18 months. Now, it’s only every 20 years.

We’ve tried to fill the compute gap by piling on more cores, but constraints on available power and limits on the size of things we can fabricate in silicon have set an upper-bound on how far we can push the performance of traditional CPUs. We need to invest in new, non-traditional (post-Von-Neumann) architectures that have the potential to fill the growing compute gap.

We also need better domain specific architectures to meet the super-exponentially growing demands for deep learning. We need radical approaches to the problem if we are to continue to make progress with larger and larger models. By several accounts Open.AI’s GPT-3 model takes many millions of dollars of machine time to train. Without approaches that fundamentally rethink the problem we will hit a brick wall.

We also need better data connectivity between chunks of compute. The shift from electrons to photons at every scale will allow exponential improvements in power and throughput. Moving electrons even a tiny distance along a wire or through silicon takes a lot of energy and time, and generates heat. The vast majority of power used by modern computers comes from this data movement, not logic.

Recent improvements in silicon photonics allow large systems to be built where data moves as photons, avoiding the charge-discharge friction that electrons exhibit. These kinds of technologies allow ridiculously fast chip to chip interconnects that never cross copper and systems with internal bandwidths that are orders of magnitude better than anything one could create with electrons.

Silicon photonics is also an ideal technology for building large scale fault tolerant quantum computers. A practical quantum machine requires millions of physical qubits to provide the redundancy required for long lived computations. By leveraging the trillions of dollars that we as a species have invested into silicon technology we are now able to build an entirely new class of computer that uses photons to do calculations that would be simply impossible on classical machines.

We need better AI. For all the early promise of deep learning, and its unreasonable effectiveness in solving a certain class of problems, it does not always feel like a power for good. Social networks, dating apps, and other behavior modification services, in addition to the broader attention economies, have siphoned off some of our best and brightest engineers to work on things that seem trivial at best, insidious at worst.

The same deep learning technique that can generate a cartoon version of you on Instagram can be used to create novel therapeutic molecules. On which of those do we want to be spending our time and capital? Handling or preventing a pandemic seems like a better investment.

As Gary Marcus is fond of pointing out, our current A.I.s are “Brittle, Greedy, Opaque, and Shallow”. They are easy to fool, require extraordinarily large amounts of training data, provide no information about how they work internally and cannot match the cognitive performance of an earthworm. We need a new level of robust artificial intelligence that can synthesize information from a variety of sources and that has the ability to reason within different contexts, just as we would expect from animal intelligences.

We need better tools for making things. When 3D printing was first conceived, there were lots of not-so-interesting applications suggested. If you broke a cup, for instance, you could print another one. In reality, the power of 3D printing technologies is that you can create parts that are physically impossible to make by any other means. We are just scratching the surface as to what is possible. Already today, if you want to go to space, you need these kinds of technologies. The most powerful rocket engines simply cannot be made without 3D printing. At least one company is printing almost the entire spacecraft.

We need better digital cities and infrastructure. Code is the new concrete. Without digital platforms that allow management and marketplaces, our infrastructure will drown. From ride-share overwhelming airports to clots of one-hour delivery, to empty trucks and skies darkened by delivery drones, our lives are not improved by unchecked strip-mining of our public infrastructure. We need platforms that allow digital marketplaces to be created to manage and price infrastructure and allow commerce to thrive without ruining the quality of our lives.

We need a better understanding of the mechanisms that support life. Civilization essentially depends on two enzymes — nitrogenase and rubisco — to fix nitrogen and carbon dioxide. We would all be dead without them, yet we have no idea how they work at a molecular and atomic scale. It’s incredibly embarrassing.

If we are to manage and reverse the effects of climate change, or understand and engineer the dynamics of a virus like Covid-19, we need an entirely new kind of computer.

Large-scale quantum computers will allow us to understand and ultimately improve on these and other fundamental natural mechanisms. They have the potential to help us create new kinds of materials, medicines and intelligences. The computing revolution really hasn’t happened yet; our current computers will look like toys compared to these new quantum machines. They will create a nexus of opportunity to invest in new technologies and industries enabled by these exponentially more powerful computers.

We need better medicine. We spend $1.5 trillion annually to develop drugs, often guessing about the actual mechanism — we have no idea how a simple molecule like Tylenol works, for example. We need to be able to rationally design therapeutics, engineering them directly with new computational tools. We also need to automate every step of the process from design to production. Covid-19 is a wakeup call here, too.

The Life sciences has not had its industrial revolution yet, and we are stuck inside for a year because we don’t yet have the tools to quickly create the programmable medicine we need to tackle a 30 kilobyte virus. New collaborations at the boundaries of computation and biology will yield new kinds of therapeutics and diagnostics to tackle current and emerging maladies.

We don’t have to wait decades to fill these needs. There are companies actively working on these and other transformative technologies today. We have agency over how and when these kinds of technologies find their way into our daily lives. We can choose to hide in our caves as a political and viral pandemic swirls outside or we can make big bets on the technologies that move the world forward. I vote for the latter.

Peter Barrett is an investor and general partner of Playground Global; formerly CTO at WebTV and CloudCar; inventor of the first widely used video codec and Microsoft Distinguished Engineer.

Make No Small Plans: Why Venture Capital Needs to Return to Big Bets on Deep Technology was originally published in The Startup on Medium, where people are continuing the conversation by highlighting and responding to this story.

Related Articles

Intelligent models for smarter decision-making

The popularity of the design, build, and test approach to engineering is fast-waning as today’s engineers face unprecedented pressure to innovate, keep pace with the latest technologies, and design creative solutions to urgent problems.  Consider, for example, automated driving systems. Although autonomous vehicles promise to significantly improve mobility, engineers must test these frameworks for critical…

These five AI developments will shape 2021 and beyond

The year 2020 was profoundly challenging for citizens, companies, and governments around the world. As covid-19 spread, requiring far-reaching health and safety restrictions, artificial intelligence (AI) applications played a crucial role in saving lives and fostering economic resilience. Research and development (R&D) to enhance core AI capabilities, from autonomous driving and natural language processing to…


Your email address will not be published. Required fields are marked *

Receive the latest news

Subscribe To Our Weekly Newsletter

Get notified about chronicles from TreatMyBrand directly in your inbox