AI in Mining

Our Perspectives on AI

Baidu Ventures is one of the world’s leading Venture Capital Funds focused on Artificial Intelligence and Machine Learning. We believe AI is creating a new generation of devices, products, service industries, and value chains through its radical recasting of precision in production, flexibility in supply chains, labor costs, and effective forecasting & planning. Just as airplanes gave-rise to airlines, and cars gave-rise to Uber, these new devices, solutions and machines will enable a new class of operator to leverage the power and efficiency gains. New ultra-low-cost, low power, high-dimensional sensors, cameras and data collection will increase by 10000x, and large, new suppliers will be built. Quantum computing, optical chips, new computing architectures, data transmission & communications, new operating systems and algorithmic breakthroughs will lead to a new generation of technology infrastructure providers. All the things humans couldn’t see: from the small (quantum level, cellular, microbial level) to the large (earth scale imaging, exabyte-scale data), our devices and computers are beginning to know more about the world around us than ever before. Technology will interact with humans and the physical world through new embodiments — robotics, screens, AR/VR, implanted devices, building control systems etc. will give rise to all-new paradigms. We’re excited to back entrepreneurs building and scaling these new, data-enabled businesses. This article is the first in a series on industry applications of AI where we are excited to invest our time, capital and connections.

AI in Mining

For the past couple years, any discussion of ‘mining’ and ‘startups’ tends to focus on the energy-intensive process by which bitcoin and other cryptocurrencies are generated. It was recently determined that it takes 3 times more energy to mine a bitcoin than an equivalent dollar amount of gold.

At Baidu Ventures, we are big believers in the power of blockchain and cryptocurrency, but we’re also putting our efforts towards accelerating a more traditional type of mining startup: those focused on metal and mineral extraction.

Traditional resource mining is one of the largest industries on Earth, representing nearly 2 trillion dollars of value. While mines employ over 1% of the world’s workforce, they are also responsible for 8% of the world’s fatal accidents, making mining one of the most dangerous jobs on the planet. Finally, resource mining is responsible for some of the world’s worst environmental and ecological disasters. At Baidu Ventures, we see a tremendous opportunity for startups leveraging new technologies and data-enabled business models to transform the mining industry. These startups can help make mining more efficient, safer and less environmentally-destructive.

The Mining Value Chain

Why now?

Mines have existed since the bronze age, so why are we only now becoming interested in this industry? To answer this question, it’s important to recognize a few significant changes in the industry that are enabling innovation.

First, the global mining industry is becoming increasingly connected. Advances in blasting and drilling systems have allowed miners to extract a variety of resources from the surface more cheaply and safely than from deep under the ground. New mines are being built as open pits or ‘strip mines’ rather than traditional underground mines. These open pit mines are easier to instrument and monitor than underground mines as they have better access to wireless connectivity. Surface mines are also less space-constrained, allowing mine operators to leverage larger, more-efficient industrial equipment such as dump trucks. Finally, commoditization of telecommunications equipment has made it more affordable than ever for a mine operator to deliver Wi-Fi and LTE connectivity to a mine site.

Open Pit Mines Enable Better Connectivity

Changes in the types of commodities being mined today are also driving new startup interest in the sector. Worldwide, demand for structural metals (steel/iron) has declined over the past few years, particularly in China. This decline is a symptom of global reduction in new building construction, as well as increased substitution/recycling of existing waste metals. New mines are mainly being constructed to extract high-value rare-earth metals like cobalt, nickel and lithium, which are expected to power the coming electric vehicle and battery revolution. Rather than expand their footprint, structural metal mines are choosing to implement new technologies to increase the yield and efficiency of existing facilities. With new growth and change in the sector, startups are able to enter the market early and capture a significant share. Furthermore, startups that can help existing mines extract resources more efficiently are poised to see significant adoption.

Finally, the global mining industry has taken a renewed interest in PR. A combination of ecological disasters and highly-publicized accidents have forced the industry to take significant steps to rebuild its image. Coupled with tighter worldwide environmental and labor regulations, this change in approach represents an opportunity for startups who can make mines safer and/or less environmentally destructive.

At BV we have identified several exciting areas for startups to leverage these shifts within the mining industry and improve the sector as a whole. We’re excited to see startups leveraging new sensors and computer vision to unlock new sources of data that improve mine planning and decision-making. Mines are also increasingly adopting automation technologies to take humans out of harm’s way. Finally, we are seeing companies develop comprehensive mine planning and operations platforms that leverage these new data sources and autonomous technologies.


Traditionally, the ultimate source of data to direct mine operations and decision-making has come from human fieldwork. Mines employ armies of geoscientists and geotechnical engineers to identify new exploration opportunities, assess existing mine safety, set the rate of mine advancement, and optimize operations. New types of sensors allow mine operators and exploration companies to tap new sources of data and make smarter decisions. For example, startups like Rockmass Tech and Geoslam leverage Lidar systems to make mines safer and more efficient. With these systems, a geotechnical engineer can better map and monitor the rock face. This technology streamlines and automates this data collection process and allows mine operators to process and feed data directly into existing mine planning software.

New data sources are also impacting mine exploration and environmental monitoring. Startups like Earth AI and Descartes Labs are using satellite data, along with high-altitude, multi-spectral imagery of Earth’s surface to identify new areas for exploration and track global productivity. These new data sources have been shown to increase the success rate of mine exploration by as much as 3X. By increasing success rate, startups can help mining companies to save money and reduce environmental impact of unnecessary exploration. In addition, researchers and startups are leveraging low-cost environmental sensors to help mines comply with safety and environmental regulations. GUIA, a startup spun out of the University of Arizona’s College of Mining and Geological Engineering has developed a suite of environmental sensors equipped with a robust mesh networking stack. These sensors can be deployed on miners themselves, and around the job site, to track dust/particulate emissions and identify the buildup of toxic or explosive gases. Because the sensors are mesh-networked, they can communicate with each other at low latency across a surface or underground mine without reliable internet connectivity.

Finally, advances in camera technology, computer vision and image analysis are allowing mines to increase productivity and reduce waste. Startups like Mine Vision Systems and Emesent are using cameras, lidar and drones to efficiently map and build 3D models of existing mines. Startups and researchers are deploying advanced computer vision to more precisely analyze rock cores and mine outputs. These systems can pick out subtle differences in color between different grades of ore, allowing mine operators to more efficiently segment and categorize the quality of their products, resulting in overall higher revenues. CV systems are also helping mine operators to spot defects earlier in the production process, thereby increasing overall yield and reducing waste.


The holy grail of mining tech involves the use of automated machinery to keep human operators safe. The goal of automation is not necessarily to reduce headcount, but rather to move humans away from the most dangerous activities. A company called SAFE AI has partnered with Bobcat to automate heavy equipment used in mining and other industries. The startup’s technology enables heavy equipment to function without a human operator. SAFE AI’s algorithms and sensors are designed to operate in the remote, dusty conditions typical of a surface mine or construction site. By automating these operations, mining companies can both prevent fatalities and reduce the amount of time equipment sits idle. By reducing idle time, mine operators can lower CO2 emissions and fuel consumption.

Mining incumbents such as Caterpillar and Rio Tinto are also investigating the potential for automation of drilling and extraction of resources. Rio Tinto’s 11 Automated Drilling Systems have now drilled more than 5,000 kilometers without a human operator on-site. This system allows a single operator, at a remote location, to monitor and control multiple drilling rigs. As automated drilling and blasting becomes more commonplace, mines will be able to increase their rate of advancement without putting more miners in danger.

Planning and Decision Making

Most large-scale mines employ mine-planning software developed by one of a few larger mine equipment or operation companies beginning in the 1970’s. The most popular software packages provide a comprehensive suite of tools allowing mine operators to complete tasks including: 3D visualization, modeling, database management, reserve calculation, mine design and mine-planning. While these software packages are designed to be modular, allowing mine operators to select the precise modules for their specific needs and data sources, the underlying models and algorithms have largely remained constant since inception. As a result, existing mine-planning software is not well-suited to incorporate next-generation sources of sensor and satellite data, or to optimize and manage the operations of a fleet of autonomous equipment. Modern modeling and CAD algorithms used in other industries, such as manufacturing or construction, simply do not translate well to the specialized needs or safety requirements of mining.

New innovations in mine planning software are necessary to incorporate new types of sensor data and to efficiently manage autonomous equipment. The next generation of mine planning software will incorporate actual specifications of the particular machinery deployed on-site, as well as live sensor/telemetry data from those machines, in order to better consider the advantages, behaviors and limitations of the actual equipment when making planning or operational decisions. Advanced mining software platforms will also replace existing triangulation models with tetrahedral models, which allow operators to model structures not just by their external surface, but also within their volume. This modeling technology will allow operators to build larger, more precise and comprehensive models of underground ore bodies.

Finally, the next generation of mining software will leverage the growth of high-speed connectivity on the job site. Unlike virtually all other industrial software, most traditional mine planning systems cannot be updated via the internet due to legacy connectivity challenges at remote mines. As more and more mines come online, software developers will be able to deliver more frequent updates, remote licenses and better communication tools to mine operators, bringing the User Experience (UX) of mine planning software up to par with that of other industries. Future mine planning software packages will also increasingly leverage autonomous fleet management tools and routing algorithms. This software will become the conductor of a complex array of manned and unmanned vehicles, optimizing route planning to minimize downtime and improve safety.

What we’re looking to fund

The potential for advanced sensors, software and automation to revolutionize the mining industry is significant. As the industry evolves, and becomes more conscious of improving its image, these tools will become more common and cost-effective for mine operators of all scales. When evaluating startups in mining, we feel it is important to leverage an innovative, defensible technology; thorough knowledge of the market; and a clear understanding of the industry’s changing technological landscape. We also look for a clear competitive edge over incumbents and other startups. In some cases, this edge could be proprietary access to a unique data source, or a unique way of analyzing/processing publicly-available data. In other cases, a viable competitive edge could be a unique type of sensor or algorithm that is difficult for competitors to replicate or reverse-engineer. By contrast, we tend to be less interested in technology startups whose competitive edge centers around proprietary customer relationships or a first-mover advantage. While these strategies can provide a near term benefit, they tend to be more difficult to defend in the long-term than a unique technology.

Ultimately, we see mining as somewhat of a canary (pun intended) for the AI industry itself. Technology solutions that can work in the harsh, remote environment of a mine will likely work in other heavy industries such as construction and agriculture. By supporting startups applying AI and technology to legacy industries with a clear use case, we can grow and support the overall market for AI infrastructure. As demand for AI infrastructure grows, costs will continue to drop and technology will improve, benefitting the ecosystem as a whole.