This content was first published on IGR – Italian Gemological Review no. 17 in 2023. The information provided here is therefore current as of the original publication date.
The introduction of artificial intelligence is reshaping many areas of our lives and, more importantly, redefining the landscape of scientific research. Now it is the turn of gemology. In collaboration, the Gübelin Gem Lab and CSEM have pioneered a new platform that leverages artificial intelligence to ascertain gemstones’ heat treatments and their origin.
Traditionally, gemologists have relied on their own experience and knowledge to identify and classify gemstones based on separate and distinct criteria that generate distinct but unrelated data. This is defined as expert-based evaluation, which provides confident predictions though on a significantly smaller subset of stones. Today, artificial intelligence can accelerate and automate this process, as demonstrated by the Gübelin Gem Lab’s Gemtelligence project. This initiative employs convolutional and attention-based neural networks to process a set of data derived from various instruments. The algorithm organizing this process has demonstrated prediction accuracy by capturing correlations between different data modalities. This achievement is noteworthy, especially considering the use of relatively inexpensive analytical methods so to reduce dependency from highly costly and extensive use of laser-ablation inductively-coupled plasma mass spectrometry (ICP-MS)
The presentation of Gemtelligence to the scientific community, reported in the article titled “Gemtelligence: Accelerating Gemstone Classification with Deep Learning”, can mark a significant step forward.
Gem News published on IGR – Italian Gemological Review #17, Autumn 2023
