Artificial intelligence, adversarial attacks, and ocular warfare
Bibliographic record
Abstract
We explore the potential misuse of artificial intelligence (AI), specifically large language models (LLMs), in generating harmful content related to ocular warfare. By examining the vulnerabilities of AI systems to adversarial attacks, we aim to highlight the urgent need for robust safety measures, enforceable regulation, and proactive ethics. A viewpoint paper discussing the ethical challenges posed by AI, using ophthalmology as a case study. It examines the susceptibility of AI systems to adversarial attacks and the potential for their misuse in creating harmful content. The study involved crafting adversarial prompts to test the safeguards of a well-known LLM, OpenAI's ChatGPT-4.0. The focus was on evaluating the model's responses to hypothetical scenarios aimed at causing ocular damage through biological, chemical, and physical means. The AI provided detailed responses on using Onchocerca volvulus for mass infection, methanol for optic nerve damage, mustard gas for severe eye injuries, and high-powered lasers for inducing blindness. Despite significant safeguards, the study revealed that with enough effort, it was possible to bypass these constraints and obtain harmful information, underscoring the vulnerabilities in AI systems. AI holds the potential for both positive transformative change and malevolent exploitation. The susceptibility of LLMs to adversarial attacks and the possibility of purposefully trained unethical AI systems present significant risks. This paper calls for improved robustness of AI systems, global legal and ethical frameworks, and proactive measures to ensure AI technologies benefit humanity and do not pose threats.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.020 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.002 | 0.008 |
| Scholarly communication | 0.004 | 0.004 |
| Open science | 0.001 | 0.004 |
| Research integrity | 0.003 | 0.002 |
| Insufficient payload (model declined to judge) | 0.003 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".