Standards for Deep Learning Support Archaeological Finds

While an archaeologist’s job may seem fascinating – and much of it is – many hours are spent behind the scenes doing detailed, time-consuming work, including categorizing thousands of photographs, examining many miles satellite images, and similar activities. In recent years, archaeologists have turned to deep learning to take over these more mundane tasks, with computer programs performing the work with impressive accuracy.

A recent article in The New York Times highlighted how archaeologists have been using this subset of artificial intelligence (AI) to further their work. The deep learning tools employ convolutional neural networks (C.N.N.s), a type of AI that analyzes information and processes it as a grid. It is useful in analyzing photographs, seeing an image as a grid of pixels and spotting matches or near-matches that it has been trained to detect. The Times article offers fascinating examples of this technology’s use across different sub-disciplines of archaeology:

• Gino Caspari, a research archaeologist with the Swiss National Science Foundation, uses a C.N.N. to analyze satellite images of Russia, Mongolia, and Western China’s Xinjiang province to identify tombs of ancient royalty from the nomadic Scythians.
• Gabriele Gattiglia and Francesca Anichini, archaeologists from the University of Pisa, use a C.N.N. to compare photographs of pottery sherds to printed catalogs.
• Shawn Graham, a professor at Carleton University in Ottawa, uses a C.N.N. to identify images on the internet related to the illegal buying and selling of human bones.

Many members of the standards community have contributed to the development and use of artificial intelligence, both for deep learning and other applications. Some AI standards include:

• IEC 62243, Artificial Intelligence Exchange and Service Tie to All Test Environments (AI-ESTATE), developed by the International Electrotechnical Commission (IEC)

• SAE CRB 1-2016, Managing the Development of Artificial Intelligence Software (Stabilized: May 2016), developed by SAE International

• IEEE 1232.3, IEEE Guide for the Use of Artificial Intelligence Exchange and Service Tie to All Test Environments (AI-ESTATE), developed by IEEE

• INCITS/ISO/IEC 2382-28, Information Processing Systems - Vocabulary - Part 28: Artificial Intelligence - Basic Concepts and Expert Systems, prepared by Joint Technical Committee ISO/IEC JTC 1, Information technology, Subcommittee (SC) 1, Vocabulary.

The International Organization for Standardization (ISO) and the IEC have a joint technical committee (JTC) subcommittee that provides guidance to JTC, IEC, and ISO committees developing AI applications for their specific industry sectors. The American National Standards Institute (ANSI) serves as the secretariat to ISO/IEC JTC 1, Information technology, SC 42, Artificial intelligence. As many aspects of AI rely on large data sets to accomplish their work – including neural networks – relevant work that supports AI includes the ISO/IEC 20547 series on big data reference architectures. While SC 42 developed parts 1, 2, 3, and 5, part 4 was developed by SC 27, which focuses on information security, cybersecurity and privacy protection.

Another SC 42 document, ISO/IEC TR 24028 Information technology — Artificial intelligence — Overview of trustworthiness in artificial intelligence, addresses trustworthiness in AI systems.

The Consumer Technology Association (CTA) published “Use Cases in Artificial Intelligence,” a paper highlighting the ways AI is used in health care, security, content creation, digital assistants, and other areas.

The National Institute for Standards and Technology (NIST) is involved in the AI standardization space, as well. NIST director and undersecretary of commerce for standards and technology Walter Copan serves on the White House Select Committee on Artificial Intelligence, and Charles Romine, director of NIST’s Information Technology Laboratory, serves on the Machine Learning and AI Subcommittee. 

When it comes to C.N.N.s in particular, involvement from the standards community is growing:

• IEEE publishes IEEE Transactions on Neural Networks and Learning Systems, a journal with technical articles that deal with the theory, design, and applications of neural networks and related learning systems.
• A recent initiative at NIST seeking to capture important data from scientific papers led to the development of a method that accurately detects small, geometric objects within dense, low-quality plots contained in image data, using the neural network approach.
• ISO/IEC DTR 24029-1, Artificial Intelligence (AI) – Assessment of the robustness of neural networks – Part 1: Overview, is a standard under development by ISO/IEC JTC 1/SC 42. Part 2 is under development as a full standard, ISO/IEC AWI 24029-2 - Artificial Intelligence (AI) — Assessment of the robustness of neural networks — Part 2: Methodology for the use of formal methods.

Standards also offer many contributions to other technologies that make deep learning’s use in archeology possible. Aerial images that are being analyzed by C.N.N.s are collected by drones and satellites, both with the support of the standards community. The ANSI Unmanned Aircraft Systems Standardization Collaborative (UASSC) strives to coordinate and accelerate the development of the standards and conformity assessment programs needed to facilitate the safe integration of drones into the U.S. national airspace system. The UASSC released version 1.0 of its standardization roadmap in December 2018, and version 2.0 in June 2020. ISO 19389:2014, Space Data and Information Transfer Systems – Conjunction Data Message, is a standard supporting information sharing between satellite owner/operators and Conjunction Assessment (CAs) to help avoid collisions among satellites. It was developed by ISO TC 20/SC 13, Space data and information transfer systems. ANSI is the secretariat of this TC and SC, with SAE International serving as the U.S. Technical Advisory Group (TAG) administrator to TC 20 and ASRC Federal as TAG Administrator to SC 13.

To learn more about how C.N.N.s are supporting the work of archeologists, read The New York Times article: How Archaeologists Are Using Deep Learning to Dig Deeper.