Artificial Intelligence is undergoing a massive acceleration driven by rapid growth in available data and rapid evolution of algorithms.
Intel’s acquisition this Tuesday of our portfolio company Nervana Systems validates that our companies’ platforms, driving this acceleration, are disrupting the world’s largest industries.
Our thesis is that a) increasingly powerful and inexpensive hardware that is machine learning/deep learning-friendly (lots of multipliers and fast memory; e.g., Nervana Systems), b) a flourishing of learning approaches running on that hardware, and c) large, novel data sets, now inexpensive to acquire and refresh, to train and drive those novel learning algorithms, is fueling a transformation of major global industries right in front of everyone’s eyes.
Mission-critical decisions can now be made in the face of huge amounts of even chaotic data, and life-or-death actions can be implemented in the real-world, at large scale, with dramatically less cap-ex and op-ex than ever before, thanks to the speed, clarity, and efficacy of commercially practical AI.
We’ve been investing in this thesis for the better part of a decade along with a small number of like-minded folks‡, and we expect that it will fundamentally disrupt every industry vertical. Let’s take a look at how this combination of the rapid growth in available data and a corresponding rapid evolution of algorithms is playing out in the wild.
Disrupting the world’s largest industries
Increasingly powerful hardware, evolving machine learning approaches, and large new data sets are fueling a transformation of major global industries.
Drug discovery is moving from a paradigm of armies of expensive white coats in labs to algorithms in the cloud — leading to a drug discovery process that takes places over months rather than years. CapellaBio and Atomwise engage in drug discovery directly. Transcriptic and Vium are make experimental biology and pharma processes elastic and on-demand via fully automated warehouses for wet lab and in vivo experiments, respectively; think of them as AWS for wet labs and animal models, respectively. 3Scan is reinventing pathology as a high throughput, quantifiable, digital science. Omniome is revolutionizing diagnosis using a more complete set of genomic, proteomic, and clinical patient data.
Zymergen has integrated proprietary machine learning software solving a problem computationally harder than AlphaGo with robotic lab automation, and a huge storehouse of genomic data, to evolve in silico, test, and deliver microbes for its Fortune 100 global customers that radically change the economics of bio-produced products for the chemicals, health, agriculture and food industries. Atomwise, which uses deep learning at massive scale along with very large chemistry and genomics data sets, enables in silico discovery of non-toxic pesticides and safer agricultural products, as well as pharmaceuticals, for multiple Fortune 100 companies today.
Innovation in precision agriculture is occurring from a macro scale based on aerial imagery to a micro scale based on a deeper understanding of bacteria. Descartes Labs is detecting subtle changes in growth patterns from petabytes of orbital data to accurately predict harvests, impacting agriculture worth hundreds of billions of dollars. Pivot Bio has applied proprietary, multi-decade-depth genomics data along with novel machine learning to program networks of soil bacteria to generate fertilizer for plants from the air and soil, potentially eliminating nearly 10% of global energy use, as well as huge costs and toxic run-offs for farmers world-wide.
As Bill Gates often proclaims, breakthroughs in Materials Science will be one of the biggest accelerators of technology development in other fields. Citrine is a DCVC company that helps discover cheaper, less energy intensive, more environmentally friendly ways to produce existing materials, as well as in silico discovery of new materials that solve a unique product or production requirement, through novel deep learning approaches applied to extremely large and complex data sets.
Financial Services and Risk Management
From banking to insurance, data and algorithms are disrupting how we estimate and manage risk, price products, and make financial predictions. LendUp uses machine learning to score risk faster and more accurately, with less data, to unlock credit for the world’s unbanked. Raptor streamlines compliance processes and detects suspected money laundering and terror-financing patterns in huge transaction flows. One of our stealth companies unearths complex dependencies in multi-party financial instruments to prevent another 2008 event. Another stealth startup underwrites cyber insurance policies and provides board-level financial risk assessments for cybersecurity risk. Cape Analytics uses aerial imagery with breakthrough machine vision techniques to facilitate real estate insurance underwriting from above rather than sending someone out to visit buildings manually. Metabiota is the world’s leading company for tracking, qualifying, mitigating and pricing and making efficient markets to manage epidemic risk, applying machine learning to fast-moving, proprietary data sets from its network of epidemic sentinel stations around the world.
Digital and Physical Security
With the increasingly visible cataclysmic manifestations of digital and physical risks, detecting and neutralizing threats is top of mind for the public and private sector. Primer (stealth) uses a novel ensemble of deep learning techniques to ingest petabytes an hour of unstructured data in multiple languages and automatically generating analytical reports to identify patterns of terror or crime. SentinelOne uses compute-efficient machine learning (less than 1MB memory profile, runs in real-time on an ARM) to Identify the behavior of “zero-day” cyber-attacks in real time and stops them cold. Kentik ingests terabits an hour of network traffic and automatically detects DDoS attempts and network security risks. PerimeterX finds and stops bot networks in banking and trading, an area where other systems have failed to date. Area1 detects the subtle patterns of phishing attacks and stops them before they start.
As computation eats capex and opex, the world increasingly yearns for more powerful and specialized compute resources to unlock new possibilities and make existing programs run faster on larger amounts of data. Nervana is a great example of specialized compute — massively accelerating deep learning. Rigetti aims to deliver multiple orders of magnitude improvement in machine learning and deep learning in general, and in particular simulation, genomics/proteomics, materials science, and similar computation via its breakthrough quantum compute platform.
With costs growing at unsustainable rate and a system rife with inefficiencies and poor decisions, healthcare is ripe for technology disruption. From supporting physicians in making more accurate diagnoses to closed-loop optimization by analyzing outcomes and learning to predict the most effective treatments, applications abound for data-driven improved outcomes. CloudMedx synthesizes effective and inexpensive treatment plans from huge amounts of ambiguous health care data. Karius helps avoid the costs of mistaken diagnoses and poorly conceived treatments.
Manufacturing and Supply Chain
Manufacturing and supply chain have been under the optimization microscope since the industrial revolution, and the quest continues. Tradeshift streamlines how corporations manage their suppliers and leverages network effects of supply chain data to provide never before seen supply chain analytics and risk management. Kindred uses AI-driven robotics so that one human worker can do the work of four. SigOpt delivers custom manufacturing and supply chain optimization recommendations, on demand. Rescale enables the on-demand combination of existing industrial supercompute for simulation and manufacturing planning with advanced machine learning for optimization. Most of today’s private aerospace and rocket companies use Rescale to simulate how existing and novel materials will perform in real world systems at full scale.
†Standing on the Shore: as the late Dr. Carl Sagan said, “The surface of the Earth is the shore of the cosmic ocean. On this shore, we’ve learned most of what we know. Recently, we’ve waded a little way out, maybe ankle-deep, and the water seems inviting.” Similarly, DCVC believes that humanity is just as of yet standing on the shore, having waded a little way out, ankle-deep, into the huge fertile ocean of how AI will transform our lives, our industries, our economies, our health, our happiness, our safety and prosperity.
‡The team at DCVC wants to thank Steve Jurvetson, Managing Partner of DFJ, for vouching to the Nervana team that DCVC was a value-add investor, and for his key role in making Nervana happen for all of us. DFJ and our other co-investors in Nervana, Lux Capital, are both co-investors with DCVC in a number of the companies cited here, and we are grateful for their friendship, support, and shared vision.