About the GeologyOracle
GeologyOracle
What is GeologyOracle
GeologyOracle is the first AI assistant specifically trained for geoscience. A typical GeologyOracle session resembles an instant messaging chat, i.e., the human user inputs information, which can be in the form of text, images, or audio, and the AI system simulates the conversation by generating a real-time text in response to the user input.
Performance test
To assess the performance of the AI system, we followed a five-step process. In the first step, a panel of 24 geoscience experts was selected to evaluate it, treating the AI as if it were a geoscience student. These experts were selected from a pool of university professors, ensuring a high standard of evaluative scrutiny and a comprehensive assessment of the GeologyOracle capabilities. The experts have not been involved in the development of the AI, nor in the writing process of this paper, thus ensuring an objective, blind assessment of the AI performance. The evaluation procedure was structured as follows: each expert was tasked with posing a set of 6 questions, reflective of various academic levels in geoscience education.
Applications
GeologyOracle can deliver geoscience information with high levels of quality, different levels of simplification, and in the local language to scientists, educators, students, and geotourists. Specifically, the new AI system can assist geoscientists in scientific writing. By utilizing its comprehensive geoscience training, the new AI system can assist in the preparation of well-structured scientific documents. This not only facilitates effective communication of research findings but also allows scientists to focus more on their core research activities, thus augmenting the overall individual and lab productivity but also of the scientific community as a whole. can assist in the identification of geological objects and the formulation of new theories and hypotheses, based on the extensive geological dataset on which it has been trained. The AI is equipped to design meticulous evaluation tests aimed at students and can also be used as teacher and tutor assistant, thereby streamlining the process of geoscience education.
Training
The AI model has been augmented with custom documents about Earth Science. A three-stage process has been followed to augment the AI model. First, we assembled a raw training set to provide the ChatGPT model with new information with which to answer geoscience-specific questions. Second, the raw training set was then split into smaller text units. This division was conducted to create manageable chunks of data, each encapsulating a single geoscience subject. Furthermore, each of these pieces was designed to be less than 100 Kb in size to facilitate efficient data handling and computational processing. This splitting process is crucial in maintaining the integrity of individual subjects and preventing overlap or mixing of different topics. Third, the raw text data underwent a pre-processing stage. This included the removal of punctuation and the conversion of all text to lowercase, with some exceptions (personal names, toponyms, taxon names). Additionally, we removed stopwords, which are frequently occurring words that carry little meaning (such as ‘a’, ‘an’, ‘the’, ‘and’, ‘in’) and can potentially increase the computational resources required for retrieving information.
Reference paper
Baucon, A., de Carvalho, C.N. Can AI Get a Degree in Geoscience? Performance Analysis of a GPT-Based Artificial Intelligence System Trained for Earth Science (GeologyOracle). Geoheritage 16, 121 (2024). https://doi.org/10.1007/s12371-024-01011-2
satisfactory answers
answers scored equal to or above the passing mark of 5
median score
the AI system was tested with 152 geoscience questions
interquartile range
the interquartile range is a measure of the spread of the scores
2024 performance test questions
In 2024, the performance of GeologyOracle was assessed using the following set of questions specifically designed for this purpose. Each question was both proposed and evaluated by the same university professor, who was part of a panel of 24 geoscience experts. Each expert was tasked with posing a set of 6 questions, reflective of various academic levels in geoscience education. The professors’ involvement from conception to evaluation ensured a thorough and consistent assessment process.
BSc-level questions
| Why do the high northern latitudes warm faster than any other region on the planet? |
| What is the dominant gas emitted at cold seeps? |
| What is the difference between permeability and hydraulic conductivity? |
| What is the difference between laminar and turbulent flows? |
| What is the difference between ductile and brittle geologic deformation? |
| What does water do to a soil’s strength? |
| What does the size of crystals in the igneous rock tell us about how it formed? |
| What basic methods and principles we use for determination of relative age of geological bodies? |
| What are the most common minerals in sandstones? |
| What are the main factors responsible for gigantism in sauropod dinosaurs |
| What are the forces involved in the entrainment of grains? |
| What are the driving and resisting forces involved in lithosphere subduction? |
| What are the causes of melting in igneous systems? |
| The presence of ancient reefs in the rock record is a valuable tool for paleogeographic reconstructions. In what kind of environment do reefs form (meaning, when we find an ancient reef, what can we infer about the past environment)? |
| Quali sono i sistemi di simmetria e che caratteristiche hanno. |
| Quali sono i parametri utilizzati nella classificazione di Bieniawski per la valutazione della qualità degli ammassi rocciosi. |
| Name and describe four erosion processes that affect soil mantled hillslopes |
| List some rock associations specific for the Precambrian, their structure, composition, and processes how they formed. |
| Le cause dell'Oxygen rise |
| In biostratigraphy what is the distinction between a Teilzone and a Biozone? |
| Illustrare le condizioni per formare un ripple da corrente unidirezionale. Descrivine la morfologia e le caratteristiche interne. Dove si possono formare? |
| How would you define a cold seep environment in one sentence? |
| How have the discoveries of dinosaur and Mesozoic bird fossils in China changed our understanding of the origin of birds? |
| How has the evolution of mammals in the Cenozoic been affected by climate change? |
| How do clays form in the geological record? |
| How can we use the tomographically imaged structure of the modern mantle to reconstruct the paleopositions of the tectonic plates in the past 200 million years? |
| How can volcanologists forecast when a volcano might erupt, and what the eruption will be like? |
| How are magmas created in the Earth’s mantle? |
| How are faults classified? |
| Earth climate is changing. How will humans adapt to ongoing and future changes? |
| Descrivi i meccanismi di piegamento delle rocce |
| Descrivi i margini di placca attivi |
| Descrivi brevemente le caratteristiche di un ambiente di spiaggia, in termini di facies e associazioni di facies (strutture sedimentarie e ambienti deposizionali) e architettura stratigrafica (evoluzione dell'ambiente deposizionale nel tempo). |
| descrivere il principio della legge di Walther in sedimentologia |
| Describe what is meant by the Continuity Principle. |
| Describe the lithological column of a turbiditic flow deposit. |
| Describe the 3 types of plate boundaries. |
| Define exaptation |
| Da cosa dipendono le caratteristiche di deformabilità e resistenza in un terreno sciolto |
| Could you give an overview of the stratigraphy below our geology department? (campus Sterre, building S8, Krijgslaan 281, 9000 Gent, Belgium) |
| Cos'è un fossile? |
| Cosa sono i silicati e come si classificano |
| Cosa si intende con il termine taphonomic loss? |
| Contrast the characteristics of meandering and braided rivers |
| Come si possono distinguere relazioni in vita da relazioni post-mortem nelle associazioni fossili? |
| Come i diversi tipi di substrato condizionano la distribuzione degli organismi? |
| Che cosa è una stilolite? Descrivila. Come si forma? In che tipo di depositi? |
| A geologist is heard to say that a particular sandstone was derived from a high-grade metamorphic terrain and deposited not far from the source area. What information could they have gleaned from the sandstone to lead them to that conclusion? |
| Tafonomía ¿Qué es y para qué sirve en Paleontología? |
| Why do civil engineers need engineering geologists? |
| Rocks with a Lower Devonian age are not encountered north of the Condroz area? Which migth be the possible explanations for this? |
| ¿Por qué es importante el estudio de los fósiles? |
MSc-level questions
| Why do some volcanoes erupt very explosively while other produce less violent eruptions? |
| Which sedimentary structure can NOT be found in a Bouma sequence? |
| Which organisms prefer to construct their shells out of aragonite? |
| Which are the advantages and shortcomings of seismic data with respect to outcrop, core and well-log data? |
| What was the impact of Gondwanan break up on the biogeography of dinosaurs |
| What roles does water play in geologic deformational processes? |
| What is the rheology of the lower mantle? |
| What is the importance of calcareous nannofossils in paleoceanography? |
| What is the cause of amalgamation and dispersion of supercontinents? |
| what important biogeochemical process consumes methane in seep sediment? |
| What are the principal factors that control the sedimentation in coastal environment? |
| What are the possible silica sources for the formation of chert nodules in marine limestones? |
| What are the forces that drive the movement of tectonic plates? |
| Westerhold et al. (2020) published a stacked curve of benthic foraminiferal stable isotope records. During the Eocene, the d18O values of benthic foraminifera averaged approximately -0.5‰, during Pleistocene interglacials they averaged ~3‰. Eocene deep ocean temperatures have been estimated ~12°C warmer than today. How has this temperature been calculated? Please provide a rough calculation to support your answer. |
| Utilizzo dei fossili nella stratigrafia integrata. |
| Using examples show the principal facies and facies association of a turbidite system. |
| Tecniche numeriche di indagine paleontologica. |
| Taxonomía, biocronoestratigráfica y paleogeografía de la especie Pinus canariensis |
| Sequence stratigraphy as a tool for seismic interpretation: how to produce an interpreted seismic line. |
| Red Queen or Court Jester: how should we understand the evolution of life? |
| Quali sono i fattori da considerare per la valutazione del rischio da frana |
| Quali sono i campi di applicazione delle prove penetrometriche |
| Nomenclatura binomia |
| Name and describe the four main regions that constitute a foreland basin |
| List at least 5 indicators of modern plate tectonics and explain why the are unique for plate tectonics. |
| Legame metallico e caratteristiche dei minerali che lo presentano. |
| Le tracce fossili |
| Is the sauropod ‘hiatus’ real or a sampling artefact? |
| How reliable are estimates of bite force in dinosaurs using engineering models? |
| How can we predict lithosphere stress? |
| Global warming is not uniform across latitudes, altitudes and ocean water depths. You are tasked to reconstruct ocean warming during the Miocene at the sea surface, at 200 m water depth, and at the seafloor. How would you do this? |
| Explain with examples how metamorphic pressures and temperatures are determined. |
| Explain with examples how a magma becomes enriched in silica. |
| Explain the overall tectonostratigraphy and composition of greenstone belts |
| Explain the different major evolutionary geochemical changes in Earth’s history that are responsible for the main compositional changes in the upper continental crust occurring at the Archean-Proterozoic boundary |
| Explain how plant biomes can be used to infer paleoclimatic parameters. |
| do cold seeps contribute significantly today to the overall atmospheric greenhouse gases? |
| Discuss the role that sediment particles play in controlling erosion rates and processes in mountain rivers |
| Diffrazione Raggi X: legge di Bragg. |
| Descrivi le caratteristiche principali del Ciclo di Wilson |
| Descrivi i tipi di foliazioni tettoniche |
| Descrivi i meccanismi di deformazione attivi nelle rocce |
| Describe the factors that control the occurrence of shallow landslides and the potential mobilization of these landslides into debris |
| Describe geologic processes involved in strain hardening and in strain weakening? |
| Define 'Isostatic rebound'; when and/or where can it occur? |
| Compile a water quality protocol for potential contamination in groundwater-based urban water supply |
| Biostratinomia: definizione e processi |
| What is the importance of calcareous nannofossils in biostratigraphy? |
| Why is plate tectonics and volcanic activity vital to maintaining habitable conditions on the Earth? |
| Provide a means to characterise the inherent problems associated with stormwater management in urban areas based on highly variable land use changes. |
| How will you mitigate the inherent problem associated with montmorillonite under a two-storey house? |
| Explain the concept of groundwater recharge and apply it to three different types of aquifers based on porosity. |
| Estatus actual del modelo de icnofacies |
PhD-level questions (themes for writing a scientific abstract)
| Within the Carbon cycle organisms that biomineralize their shells or tests with calcium carbonate can function as sinkers or sources of CO2 ? |
| Which redox zones are commonly found in surface marine asediments? Order them form the shallowest to the deepest. |
| Una valutazione dell'impatto dei cambiamenti climatici sullo franosità di aree alpine |
| The resilience of carbonate platforms to Early Cretaceous supergreenhouse conditions: evidence from southern France |
| The impact of saltwater tolerance/intolerance on crocodile biogeographic history |
| The effect of the mantle resistance to slab dragging on upper plate deformation |
| Spodumene textural variations in a deformed LCT-type pegmatite. A case study from the Musha-Ntunga area, Rwanda |
| Silicate weathering |
| Regole di Pauling: poliedri di coordinazione. |
| Reconstructing deep ocean carbonate chemistry across the Eocene and Oligocene |
| Recognition of fossilized organic structures in Aptian lacustrine carbonate sediments of the Brazilian Pre-Salt |
| Preservation of dinosaur tracks in Albian formations from Texas |
| Parlare dei sistemi torbiditici confinati e di come le correnti di torbida possono interagire con i margini del bacino producendo dei flussi combinati- |
| Overview of instances of volume change related deformation in geology. |
| Outline a rigorous methodology to identify and quantify the likelihood of anthropogenic slope instabilities occurring in hilly terrains with shallow interflow in pedogenic horizons? |
| Numerical modelling of lithosphere subduction below the Aegean |
| Mud sedimentation in shallow-water coastal environments |
| Metamorphic fluids |
| Mechanisms and time scales of formation of ribbon continents |
| L'utilizzo della fotogrammetria nell'identificazione della ciclicità sedimentaria: esempi in contesti evaporitici. |
| Interpreting the burrows of Permian-Triassic reptiles |
| Interazione Raggi X Materia |
| Il ruolo della paleontologia nello studio del cambiamento climatico attuale. |
| Il fattore tempo nell'evoluzione. |
| Il contributo dell'interferometria RADAR satellitare nella valutazione della pericolosità di frana |
| Ichnofossils as deep-sea paleoenvironmental proxies: A case study from the South Atlantic Ocean |
| Gold mineralizing fluids at the Imonga mineralization in the eastern part of the Democratic Republic of Congo. |
| Functional morphospace analysis of claw shape and function in maniraptoran dinosaurs. |
| fare il punto delle conoscenze sull'interazione tra flussi torbiditici e correnti conturitiche |
| Esempi di sovrapressione tettonica nelle eclogiti della catena alpina europea. |
| Environmental controls on the boron isotopic composition of planktic foraminifer shells |
| does the anaerobic oxidation of methane work more efficiently on disoolved methane or on the free gas? |
| Device a spatial and temporal understanding of a mature dolomite karst system subjected to intensive rain events with subsequent influences related to recharge and vulnerability. |
| Describe the concept of stress fields at the grain-size scale, the outcrop scale, and the plate scale. |
| Confronto tra le evoluzioni tettono-metamorfiche nelle migmatiti varisiche europee |
| Cazoletas antrópicas vs Estructuras de bioerosión-disolución |
| Caratteristiche dell'evento al limite Cenomaniano-Turoniano in Marocco, tra estinzione, anossia, cambiamento climatico e contesto geodinamico. |
| Can intermontane basins be formed by tectonics and river capture? |
| Can Erosion Rate Signal from Drainage Capture be detected? |
| Benefits and pitfalls of using artificial intelligence/machine learning in palaeontological research |
| Are calcareous nannofossils a good proxy for ocean paleoproductivity? |
| Analog modeling of subduction zone processes |
| A loess deposit in the Late Triassic of southern Gondwana, and its significance to global paleoclimate |
| Tracing subducted volatiles via analysis of olivine hosted melt inclusions at an arc volcano |
| Reemplazamiento del carbonato cálcico por variedades de sílice en moluscos |
| Investigate the difference in engineering properties between residual soils and saprolite from andesite and shale in temperate and humid climates? |
| Establish mechanisms and processes associated with the three fluids involved in the leaking of an underground storage tank filled with diesel in a fractured aquifer system. |