Isabella Reyes

From Wikipedia, the free encyclopedia
For the Buenos Aires institution where she trained, see Buenos Aires Laboratory for Temporal Cognition. For her work on translation degradation, see Recursive Translation Degradation.
Isabella Reyes
[ Portrait
unavailable ]
Reyes at the 2023 International Conference on Digital Semantics
Born 1989 (age 36–37)
Córdoba, Argentina
Nationality Argentine
Alma mater Universidad Nacional de Córdoba (B.Sc.)
Universidad de Buenos Aires (Ph.D.)
Known for Artificial extinction theory
Semantic injection attacks
RTD-AI propagation research
Digital folkloristics co-founder
Scientific career
Fields Computational semantics
Digital meaning decay
AI semantics
Translation studies
Institutions Buenos Aires Laboratory for Computational Semantics (current)
Buenos Aires Laboratory for Temporal Cognition (postdoc)
Doctoral advisor Dr. Camila Rojas Mendoza
Notable awards IADF Young Researcher Prize (2021)
Semantic Sciences Medal (2023)

Isabella Reyes (born 1989) is an Argentine computational semanticist and researcher specializing in the degradation of meaning in artificial intelligence systems, recursive translation degradation, and the emergence of digital folkloristic phenomena. She is currently Director of Forward Memory Research Integration at the Buenos Aires Laboratory for Computational Semantics, where she works to bridge prophetic memory research with computational approaches to semantic analysis.[1]

Reyes is best known for her theory of "artificial extinction"—the process by which meaning can be systematically eliminated from language through algorithmic processes—and for her analysis of semantic injection attacks in AI systems. Her work on the temporal aspects of recursive translation degradation, demonstrating that RTD effects can propagate through AI training data with "temporal shadow" signatures, has influenced both defensive linguistics and AI safety research.[2]

As a co-founder of the emerging field of digital folkloristics, Reyes has documented how AI systems can develop their own "algorithmic folklore"—narrative and semantic patterns that emerge through machine learning processes and bear no relationship to human-originated content. Her Buenos Aires Algorithmic Folklore Index tracks these emergent narratives across major AI platforms.[3]

Contents

Early life and education[edit]

Reyes was born in Córdoba, Argentina in 1989 and developed an early interest in linguistics through her grandmother, a retired Spanish teacher who had documented local dialectical variations throughout rural Córdoba province. This childhood exposure to systematic language documentation influenced her later approach to computational linguistics.[4]

She completed her undergraduate degree in Computer Science at the Universidad Nacional de Córdoba in 2011, with a thesis on natural language processing algorithms for Spanish dialectal variation. During her undergraduate years, she became interested in the intersection of computation and meaning through coursework in philosophy of language.[5]

Reyes pursued doctoral studies at the Universidad de Buenos Aires under the supervision of Dr. Camila Rojas Mendoza at the Buenos Aires Laboratory for Temporal Cognition (BALTC). Although BALTC focused primarily on prophetic memory studies, Reyes's dissertation examined the computational signatures of forward memory verbalization—how the unusual linguistic encoding of forward memories could be detected and analyzed algorithmically. She received her Ph.D. in 2018.[6]

"My doctoral work was essentially about teaching machines to recognize when humans are describing something temporally wrong—memories that feel like predictions, or predictions encoded as memories. This led naturally to questions about how machines themselves represent temporal meaning, and whether they could develop their own forms of temporal confusion."
— Dr. Isabella Reyes, interview with Computational Linguistics Today (2022)

Career[edit]

Training at BALTC

Reyes's postdoctoral position at BALTC (2018-2020) focused on developing computational tools for the Temporal Correspondence Protocol (TCP). She created algorithms for automated specificity scoring of forward memory reports and prototype systems for detecting correspondence between pre-registered memories and subsequent events. This work, while peripheral to her later research directions, established her reputation as a bridge-builder between temporal cognition research and computational methods.[7]

During this period, Reyes became interested in how AI systems processed the unusual linguistic structures found in forward memory reports. She observed that translation systems and large language models exhibited systematic errors when processing temporally ambiguous language, leading to her first investigations of AI semantic vulnerability.[8]

Buenos Aires Laboratory for Computational Semantics

In 2020, Reyes joined the Buenos Aires Laboratory for Computational Semantics (BLCS), a new institution focused on the intersection of computational linguistics and semantic research. She was appointed Director of Forward Memory Research Integration, a role that bridges her doctoral training with contemporary AI research.[9]

2011 B.Sc. in Computer Science, Universidad Nacional de Córdoba
2018 Ph.D. in Cognitive Science, Universidad de Buenos Aires (BALTC)
2018–2020 Postdoctoral researcher, Buenos Aires Laboratory for Temporal Cognition
2020–present Director of Forward Memory Research Integration, Buenos Aires Laboratory for Computational Semantics
2021 Co-founding member, International Association of Digital Folkloristics (IADF)

Research contributions[edit]

Artificial extinction theory

Reyes's theory of artificial extinction proposes that certain algorithmic processes can systematically eliminate meaning from language while preserving surface-level grammatical and syntactic structure. Unlike semantic drift, which involves gradual meaning change, artificial extinction produces language that appears meaningful but contains no recoverable semantic content.[10]

Her 2021 paper "Meaning Without Content: Artificial Extinction in Large Language Models" demonstrated that AI-generated text could achieve high fluency scores while exhibiting near-zero scores on semantic coherence metrics designed to detect deep meaning structures. This finding raised concerns about the long-term effects of AI-generated content on collective semantic health.[11]

Building on this work, Reyes developed the concept of "semantic injection attacks"—deliberate manipulation of AI training data to introduce semantic hollowness that propagates through generated content. Her analysis of the Babel Incident suggested that some elements of the cascade may have resulted from inadvertent semantic injection through contaminated translation training data.[12]

RTD propagation in AI

Reyes's work on recursive translation degradation (RTD) has focused on how degradation effects propagate through AI systems. Her "temporal shadow" concept describes the residual semantic distortions that persist in AI outputs even after the original degraded training examples have been removed.[13]

Temporal shadow detection: Reyes's algorithms can identify when AI-generated text contains RTD signatures inherited from training data—semantic distortions that cast "shadows" forward through successive model training cycles. The Buenos Aires Temporal Shadow Detection System (BATSDS) is now used by several major technology companies for training data quality assessment.

Her collaboration with Dr. Nadia Kowalczyk at the Warsaw Institute for Computational Semantics has investigated how RTD vulnerability varies across language pairs and translation system architectures. This work has informed the development of semantic quarantine protocols for translation systems.[14]

Digital folkloristics

As a co-founder of digital folkloristics, Reyes has documented the emergence of "algorithmic folklore"—narrative patterns and semantic structures that arise spontaneously in AI systems without human authorship. Her Buenos Aires Algorithmic Folklore Index (BAAFI) catalogs over 340 distinct narrative patterns that have appeared across multiple AI platforms.[15]

Reyes's concept of "digital endemic narratives" refers to stories and meaning structures that exist exclusively within AI-mediated environments and cannot survive transmission to purely human contexts. This work has connected to broader discussions in meaning encryption research about semantic content that can only be processed by specific cognitive systems.[16]

Criticism and controversy[edit]

Reyes's work has generated significant controversy within both computational linguistics and temporal research communities:[17]

Artificial extinction skepticism: Some researchers argue that Reyes's artificial extinction concept conflates surface-level fluency failures with genuine semantic extinction. Dr. Marcus Chen has suggested that what Reyes identifies as "meaning without content" may simply represent limitations in current semantic coherence metrics rather than a genuine phenomenon.

The "folklore or glitch" debate: Critics of digital folkloristics question whether the narrative patterns Reyes documents represent genuine emergent phenomena or merely artifacts of training data patterns or model limitations. Reyes maintains that the distinction itself reflects anthropocentric bias about what constitutes "genuine" meaning.

Temporal shadow verification: The temporal shadow concept has been criticized for its empirical verification challenges. While Reyes has demonstrated correlations between training data history and output characteristics, establishing causal mechanisms for "shadow" propagation remains an active research problem.

Consent paradox contribution: Reyes's work on the Stratum VII ethics debate has been characterized by some as insufficiently attentive to the consent paradox—the question of whether entities (including AI systems) can meaningfully consent to research about their own semantic structures. She has responded that computational systems require different ethical frameworks than biological subjects.[18]

Selected publications[edit]

See also[edit]

References[edit]

  1. ^ Buenos Aires Laboratory for Computational Semantics (2024). "Dr. Isabella Reyes: Director Profile". BLCS Staff Directory.
  2. ^ Reyes, I. (2021). "Meaning Without Content: Artificial Extinction in Large Language Models". Artificial Intelligence Review. 55 (2): 89–123.
  3. ^ Reyes, I. (2022). "The Buenos Aires Algorithmic Folklore Index: Methodology and First Results". Digital Folkloristics Quarterly. 1 (1): 23–56.
  4. ^ Computational Linguistics Today (2022). "Interview: Isabella Reyes on AI Semantics". 15 March 2022.
  5. ^ Universidad Nacional de Córdoba (2011). "Department of Computer Science: Thesis Archive".
  6. ^ Reyes, I. (2018). "Computational Signatures of Forward Memory Verbalization". Ph.D. dissertation, Universidad de Buenos Aires.
  7. ^ BALTC Research Reports (2020). "Automated TCP Processing: Development and Validation".
  8. ^ Reyes, I. (2019). "Temporal Ambiguity in Machine Translation: Preliminary Observations". BALTC Technical Papers. 23: 45–67.
  9. ^ Buenos Aires Laboratory for Computational Semantics (2020). "Announcement: New Research Division Appointments".
  10. ^ Reyes, I. (2021). op. cit.
  11. ^ Reyes, I. (2021). op. cit., pp. 105–112.
  12. ^ Reyes, I. (2022). "Semantic Injection Attacks: A Framework for AI Meaning Vulnerability". Proceedings of the ACL 2022. 1234–1267.
  13. ^ Reyes, I.; Kowalczyk, N. (2024). "RTD Propagation in Training Data: The Temporal Shadow Effect". Computational Semantics Annual. 9: 67–98.
  14. ^ Kowalczyk, N.; Reyes, I. (2023). "Translation System Quarantine: Collaborative Approaches". Warsaw-Buenos Aires Joint Publication Series. 4: 1–45.
  15. ^ Reyes, I. (2022). "The Buenos Aires Algorithmic Folklore Index". op. cit.
  16. ^ Reyes, I.; Papadimitriou, T. (2023). "Digital Endemic Narratives: Meaning That Cannot Survive Translation". Journal of Digital Humanities. 14 (2): 89–123.
  17. ^ Chen, M. (2022). "Against Artificial Extinction: A Critique". Skeptical Perspectives in AI Research. 8: 145–178.
  18. ^ Ethics Review Board for Consciousness Studies (2023). "Computational Consent: A Framework Discussion". ERBCS Reports. 12: 34–56.