DDE AT IGC 37 – BUSAN, KOREA, 25-31 AUGUST 2024

2024-09-26
437

The 37th International Geological Congress, hosted by the Republic of Korea in Busan from 25 to 31 August 2024 was a landmark event for DDE. It was the first IGC, a quadrennial world gathering of geologists, where DDE had the opportunity to showcase its achievements, products and services since DDE began its implementation in 2019.

DDE presence at IGC 37 was significant. The key events that DDE convened at IGC 37 included the following:

·       14 technical sessions under the IGC 37, theme 37 (T37)– IUGS Deep-time Digital Earth (DDE). The sessions covered key geological sciences themes, such as Evolution of Life and Biodiversity through Deep-time, Knowledge and Data Co-boosted Discovery in the Co-evolution of Minerals, Microbes and the Environment and Geoscience Data Standards and Knowledge Graph, as well as Progress, Challenges and Plans for 2024-2030 of DDE, Early Career Scientists and Large Language Models (LLM) in Geological Sciences. Readers may consult Scientific Program - IGC 2024 (igc2024korea.org) to know the conveners and co-conveners of each of the T37 sessions and contact them to learn of the key outcomes of sessions of interest to them.

·       A DDE booth where achievements of DDE since it began implementation in 2019 were presented for visitors to view, experience, learn and share data, knowledge and information. The demonstration of the use of the DDE Platform ad GeoGPT – an exploratory research project being carried out by DDE and the Zhejiang Lab, in Zhejiang Province of the Alibaba Group – attracted a regular flow of curious and enthusiastic potential users. A panel on Large Language Models in Geological Sciences was hosted on 28 August; please see XXXX for a separate report on this panel and T37 session 14 which addressed the same theme. The booth was also the venue for some DDE Members who also hosted their own booths at IGC37, for example, the Geological Survey of India. Others from important organizations, like BRGM France that was adjacent to the DDE booth experiences the use of the DDE Platform and GeoGPT and had discussions that may lead to possible, future collaboration.

·       The Chairs of the DDE Governing Council (Prof. Harvey Thorleifson), Executive Committee (Prof. Chengshan Wang) and the Science Committee (Prof. Hans Thybo) welcomed more than 60 guests representing DDE Members and partners, representatives of DDE WTGs as well as key personalities representing IUGS and a select number of Geological Science organizations from host nation Korea and others. The President of IUGS, Prof. John Ludden attended the reception and addressed the guests providing brief remarks on the work of DDE and its importance to IUGS’ global mission. Guests interacted and conversed with one another in an informal setting sharing ideas and information of mutual interest. DDE hopes that the friendships fostered during the reception will lead to new and innovative partnerships for international geological sciences collaboration in the future.

DDE also presented a 4-year summary report to the IUGS Council session on 28 August which was well received and approved by the Council.

DDE presence at IGC37 was a landmark occasion and the outcomes have been significant. They will be reflected in an expansion of DDE’s international partnerships and collaboration over the next 4 years leading to IGC 38 to be hosted by Canada in Calgary in 2028. 

 

 

Did you know that DDE is working with top-tier scientists to create the world's first GeoGPT? 

 

A slide from Prof Jieping Ye’s talk at the start of the booth session

 

Artificial Intelligence is becoming an interesting avenue of research in geology. Recently an International Union of Geological Sciences (IUGS) Deep-time Digital Earth (DDE) team, and partners Zhejiang Lab, have begun to develop a large language model for geology known as GeoGPT. DDE and GeoGPT were highlighted at the IGC 37 in Busan.

Two well attended events were held; one at the DDE booth in the conference exhibition area and one as a formal session within conference.

The booth event held on 29 August went extremely well with several people taking up places early, and the proceedings attracting passers-by. Prof Mike Stephenson was the moderator and Prof Jim Ogg, Prof Jieping Ye and Richard Chuchla, and Natarajan Ishwaran were panel members. The event began with a talk from Prof Ye and then a short demonstration of the main part of GeoGPT and then a demonstration of one of the ‘professional level’ use cases of GeoGPT known as ‘Taxonomy Assistant’.

The panel then discussed some of the following questions and fielded questions from the audience.

·       How are LLMs currently being applied in geoscience research?

·       How might LLMs change the way we approach geoscience research, geological fieldwork and data collection? 

·       What potential do LLMs have in accelerating the peer review process for geoscience publications?

·       What are the limitations of current LLMs in geoscience applications, and how might these be overcome in the future?

·       How can we ensure the reliability and accuracy of LLM outputs when dealing with critical geoscience applications?

·       What are the challenges in developing LLMs in the geoscience field?

·       What ethical considerations should geoscientists be aware of when utilizing LLMs in their work?

As a result of the event, many people registered with GeoGPT to become professional testers of the system.

The second event was held in room M312 on August 30th. The event was well attended and began with an introduction from Prof Jieping Ye, and then a discussion of some of the promising use cases for GeoGPT in the geoscience realm: these were in palynological taxonomy, igneous rock databases, and uses of GeoGPT in Nigerian geosciences.

Mike Stephenson introduced ‘Taxonomy Assistant’ an LLM-assisted taxonomic key to help palaeontologists in identifying fossils. Mike and the GeoGPT team wanted to know if large language model methods could augment a traditional taxonomic key. Could a palynologist sitting by their microscope get some help from an LLM in determining a species? It turns out that palynology very much lends itself to developing a professional LLM that might be very useful in teaching taxonomy to apprentice palynologists, or to professionals in environments that require expertise in many areas of palynology. It could also have particular use in Global South countries where access to reference materials may be difficult.

LLM-aided taxonomy is text based, and will probably be delivered through a series of structured questions and answers that converge on a determination, rather than being based on image recognition. In Taxonomy Assistant, the LLM asks the questions. But rather than provide a single answer, the unique power of the LLM provides one preferred species as an answer (with a certainty level), but also a set of ‘nearby species’. The taxonomist can look at their descriptions and the judgement as to the ‘right answer’ is theirs.

Prof Tao Wang showed how GeoGPT can be used in the complex analysis of data associated with tectonics, granitoids and continental growth in Asia. Maps and models developed by the team could revolutionize our understanding of the creation of Asia as a continent

Shadrach Sherriff, an exploration geologist at DMW LOVOL in Nigeria showed the use of GeoGPT in GIS applications for mapping and spatial analysis to identify potential mineral zones that could contribute to economic assessments, surveys and drilling programs.