6 Ways Cloud AI Empowers Global Causes
For AI to function as intended, a host of IT resources, data, and infrastructure must come together seamlessly. Microsoft puts these tools into the hands of innovators and organizations that care about the future of our planet.
Artificial Intelligence applications need much the same resources as any other computer science — namely, data, processing power, memory, and analytics tools. AI for Earth is part of Microsoft's AI for Good program, which provides grants and data science support to organizations working at the forefront of challenges in sustainability, accessibility, humanitarian action, and cultural heritage. AI for Earth releases open-source tools, models, infrastructure, data, and APIs to support sustainability and environmental science.
Eco-conscious startups and research groups apply, compete for, and win grants from Microsoft that include funding and support, in addition to the technical resources mentioned above. Let's look at the ways Microsoft supports their efforts on a global scale:
1. Data
AI starts with data, and more data gets better predictions. The challenge begins with collecting and capturing sufficient amounts of data. The Internet of Things is a growing source of extensive and flexible data capture capability. Humans may generate data manually, but far more is now collected by sensors, cameras, microphones, satellites, mobile device scanners, and myriad other devices.
On Microsoft's side, the various AI for Earth projects collect a variety of data, much of which is analyzed using specialized datasets for comparisons. AI for Earth hosts key geospatial and conservation datasets on Azure so that partners – and anyone else applying technology to conservation – can use the resources of the Azure cloud to analyze global-scale environmental data. These datasets include radar, temperature, high-resolution satellite imagery, crowdsourced labelled images, global surface reflectance, and more.
2. Research
All of this collected data is augmented by human research. Relevant studies and models are also included in the AI for Earth resources, so they may be incorporated into future analyses. In addition to the research conducted by partner groups, Microsoft Research itself drives innovation in a variety of environmental sciences. These initiatives include land cover mapping (which applies machine learning to satellite and aerial data to facilitate land management), subseasonal weather forecasting, precision agriculture information, biodiversity and pathogen monitoring, and more.
3. Storage
Once captured, the data must be stored somewhere that can scale elastically with its massive volume and support the velocity at which it needs to move for processing and analytics. When Microsoft awards an AI for Earth grant, the package includes extensive use of Azure facilities, including highly scalable, dynamic storage and data management services. Having the power of a global cloud at one's disposal makes the work of environmental science that much more efficient and focused.
4. Infrastructure
As data travels from storage to processing to end users, it's moving along network infrastructure, comprising copper and fiber-optic cabling, wireless connections, and more. The global internet is the largest and most extensive network infrastructure ever conceived. To assure that data is not compromised, it must always be encrypted for security whether it is in transit, or at rest in storage.
A major goal of the AI for Earth initiative is to make all research available and consumable immediately for environmental scientists around the world. Microsoft's application programming interface (API) framework helps developers get from machine learning (ML) models to web-based APIs. In other words, they can build functions into their software that make their work accessible to the environmental science community. These APIs are hosted on Azure's scalable infrastructure. Geoscientists, for example, can leverage the Geo AI Data Science Virtual Machine, which comes pre-configured with specialized tools including Esri and ArcGIS Pro. Again, this is a feature that speeds time to results, as developers don't have to load tools manually and configure the machines needed to host them.
5. Open-source code
All the data management and processing instructions in AI ecosystems are delivered via programming code. To enable greater collaboration, much of this code is developed and contributed to the open source community, where it is available freely. Many developers work together to improve these code bases and their documentation.
Microsoft makes all AI for Earth work available to the open source community via GitHub, including not only code, but trained AI models, as well. The environmental use cases facilitated by Microsoft code include accelerated camera trap survey workflows, classification of plants and animals in citizen-science photos, land cover classification, aerial wildlife surveys, and more.
6. Application programming interfaces (API)
As we mentioned, APIs are the connectors that enable developers to pull results from AI analytics and integrate them with existing programs that users are already familiar with. Microsoft provides APIs that are used by participating scientists, so their conservation applications are now empowered with machine learning capability.
All the investments Microsoft is making in services, training, code, research and other resources are meant to galvanize organizations that work to improve life on Earth and enable greater collaboration between them. To explore the resources that Microsoft provides, check out AI for Earth technical resources.
To learn more about how Microsoft AI is changing the world for the better, please visit https://www.microsoft.com/en-us/ai/empowering-innovation.