Can we Trust Chatbots for now? Accuracy, reproducibility, traceability; a Case Study on Leonardo da Vinci's Contribution to Astronomy - LAAS - Laboratoire d'Analyse et d'Architecture des Systèmes
Journal Articles Advances in Artificial Intelligence and Machine Learning Year : 2023

Can we Trust Chatbots for now? Accuracy, reproducibility, traceability; a Case Study on Leonardo da Vinci's Contribution to Astronomy

Didier El Baz
Équipe Services et Architectures pour Réseaux Avancés

Abstract

Large Language Models (LLM) are studied. Applications to chatbots and education are considered. A case study on Leonardo's contribution to astronomy is presented. Major problems with accuracy, reproducibility and traceability of answers are reported for ChatGPT, GPT-4, BLOOM and Google Bard. Possible reasons for problems are discussed and some solutions are proposed.

Domains

Cognitive science Computers and Society [cs.CY]
Fichier principal
Vignette du fichier
AdvancesinAIRevised21.2.pdf (282.06 Ko) Télécharger le fichier
Origin Files produced by the author(s)

Didier EL BAZ  : Connect in order to contact the contributor

https://laas.hal.science/hal-04078350

Submitted on : Monday, October 23, 2023-11:34:57 PM

Last modification on : Tuesday, January 16, 2024-4:27:23 PM

Download

Dates and versions

hal-04078350 , version 1 (23-04-2023)
hal-04078350 , version 2 (23-10-2023)

Licence

Copyright

Identifiers

Cite

Didier El Baz. Can we Trust Chatbots for now? Accuracy, reproducibility, traceability; a Case Study on Leonardo da Vinci's Contribution to Astronomy. Advances in Artificial Intelligence and Machine Learning, 2023, 3 (2), pp.1094-1109. ⟨hal-04078350v2⟩

Export

BibTeX XML-TEI Dublin Core DC Terms EndNote DataCite

Collections

UNIV-TLSE2 CNRS INSA-TOULOUSE LAAS LAAS-SARA UT1-CAPITOLE LAAS-RESEAUX-ET-COMMUNICATIONS GENCI TDS-MACS INSA-GROUPE LAAS-RISC ANR CIMI-TOULOUSE TOULOUSE-INP UNIV-UT3 UT3-TOULOUSEINP
230 View
98 Download

Altmetric

Share

More