Description: OpenAI's GPT-4o was found to have its Chinese token training data compromised by spam and pornographic phrases due to inadequate data cleaning. Tianle Cai, a Ph.D. student at Princeton University, identified that most of the longest Chinese tokens were irrelevant and inappropriate, primarily originating from spam and pornography websites. The polluted tokens could lead to hallucinations, poor performance, and potential misuse, undermining the chatbot's reliability and safety measures.
Entidades
Ver todas las entidadesAlleged: OpenAI developed an AI system deployed by OpenAI y GPT-4o, which harmed OpenAI , Chinese-speaking users of ChatGPT , Researchers y OpenAI users.
Estadísticas de incidentes
ID
729
Cantidad de informes
1
Fecha del Incidente
2024-05-14
Editores
Daniel Atherton
Informes del Incidente
Cronología de Informes
technologyreview.com · 2024
- Ver el informe original en su fuente
- Ver el informe en el Archivo de Internet
Soon after OpenAI released GPT-4o on Monday, May 13, some Chinese speakers started to notice that something seemed off about this newest version of the chatbot: the tokens it uses to parse text were full of spam and porn phrases.
On May 14,…
Variantes
Una "Variante" es un incidente que comparte los mismos factores causales, produce daños similares e involucra los mismos sistemas inteligentes que un incidente de IA conocido. En lugar de indexar las variantes como incidentes completamente separados, enumeramos las variaciones de los incidentes bajo el primer incidente similar enviado a la base de datos. A diferencia de otros tipos de envío a la base de datos de incidentes, no se requiere que las variantes tengan informes como evidencia externa a la base de datos de incidentes. Obtenga más información del trabajo de investigación.
Incidentes Similares
Did our AI mess up? Flag the unrelated incidents
Incidentes Similares
Did our AI mess up? Flag the unrelated incidents