Typology of Risks of Generative Text-to-Image Models

Published: Aug. 8, 2023

This paper investigates the direct risks and harms associated with modern text-to-image generative models, such as DALL-E Midjourney, through a comprehensive literature review. While these models offer unprecedented capabilities for generating images, their development use introduce new types of risk that require careful consideration. Our review reveals significant knowledge gaps concerning understanding treatment despite some already being addressed. We taxonomy across six key stakeholder groups, inclusive unexplored issues, suggest future research directions. identify 22 distinct types, spanning issues from data bias to malicious use. The investigation presented here is intended enhance ongoing discourse on responsible model deployment. By highlighting previously overlooked gaps, it aims shape subsequent governance initiatives, guiding them toward responsible, secure, ethically conscious evolution models.

Language: Английский

---

The role of artificial intelligence in generating original scientific research

International Journal of Pharmaceutics, Journal Year: 2024, Volume and Issue: 652, P. 123741 - 123741

Published: Jan. 3, 2024

Artificial intelligence (AI) is a revolutionary technology that finding wide application across numerous sectors. Large language models (LLMs) are an emerging subset of AI and have been developed to communicate using human languages. At their core, LLMs trained with vast amounts information extracted from the internet, including text images. Their ability create human-like, expert in almost any subject means they increasingly being used as aid presentation, particularly scientific writing. However, we wondered whether could go further, generating original research preparing results for publication. We tasked GPT-4, LLM, write pharmaceutics manuscript, on topic itself novel. It was able conceive hypothesis, define experimental protocol, produce photo-realistic images printlets, generate believable analytical data range instruments convincing publication-ready manuscript evidence critical interpretation. The model achieved all this less than 1h. Moreover, generated were multi-modal nature, thermal analyses, vibrational spectroscopy dissolution testing, demonstrating multi-disciplinary expertise LLM. One area which failed, however, referencing literature. Since appeared though, suggest certainly play role but input, interpretation validation. discuss potential benefits current bottlenecks realising ambition here.

Language: Английский

---

Artificial intelligence in plant breeding

Trends in Genetics, Journal Year: 2024, Volume and Issue: 40(10), P. 891 - 908

Published: Aug. 7, 2024

Harnessing cutting-edge technologies to enhance crop productivity is a pivotal goal in modern plant breeding. Artificial intelligence (AI) renowned for its prowess big data analysis and pattern recognition, revolutionizing numerous scientific domains including We explore the wider potential of AI tools various facets breeding, collection, unlocking genetic diversity within genebanks, bridging genotype–phenotype gap facilitate This will enable development cultivars tailored projected future environments. Moreover, also hold promise refining traits by improving precision gene-editing systems predicting effects gene variants on phenotypes. Leveraging AI-enabled breeding can augment efficiency programs holds optimizing cropping at grassroots level. entails identifying optimal inter-cropping crop-rotation models agricultural sustainability field.

Language: Английский

---

Large language models in plant biology

Trends in Plant Science, Journal Year: 2024, Volume and Issue: 29(10), P. 1145 - 1155

Published: May 26, 2024

Language: Английский

---

The Development of ChatGPT and Its Implications for the Future of Customer Service in the Hospitality Industry

Advances in information security, privacy, and ethics book series, Journal Year: 2024, Volume and Issue: unknown, P. 100 - 126

Published: April 9, 2024

This chapter summarizes the function that artificial intelligence, especially ChatGPT, plays in providing help to customers hotel business. Implementing AI-driven solutions radically altered how hotels and other businesses hospitality industry communicated with their customers. Chatbots driven by AI models such as ChatGPT were used address regular inquiries services. These chatbots helped around clock, individualized recommendations, supported several languages communicate Integration previously installed management systems ongoing development response patrons' comments standard operating procedures hospitality. However, it was noted deliver unique experiences for guests, still necessary strike a balance between human interaction. The information presented this implies intelligence will play an increasingly important role customer service within industry; nonetheless, interaction always be indispensable.

Language: Английский

---

Confronting The Data Deluge: How Artificial Intelligence Can Be Used in the Study of Plant Stress

Computational and Structural Biotechnology Journal, Journal Year: 2024, Volume and Issue: 23, P. 3454 - 3466

Published: Sept. 17, 2024

Language: Английский

---

Can a plant biologist fix a thermostat?

New Phytologist, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 2, 2025

Summary The shift to reductionist biology at the dawn of genome era yielded a ‘parts list’ plant genes and nascent understanding complex biological processes. Today, with genomics in full swing, advances high‐definition enabled precise temporal spatial analyses systems down single‐cell level. These insights, coupled artificial intelligence‐driven silico design, are propelling development first synthetic plants. By integrating approaches, researchers not only reimagining plants as sources food, fiber, fuel but also ‘environmental thermostats’ capable mitigating impacts changing climate.

Language: Английский

---

Rise of the machines: artificial intelligence in plant science and publishing

Trends in Plant Science, Journal Year: 2024, Volume and Issue: 29(2), P. 101 - 103

Published: Feb. 1, 2024

This special issue of Trends in Plant Science presents a collection tools that enable plant researchers to tackle the challenges 21st century, with artificial intelligence (AI) taking centre stage. I have confess being sci-fi buff. For me, one most fascinating aspects genre is it offers glimpse future. The current hot topic AI best example. Although seems very 21st-century topic, 'rise machines' was first predicted movies more than 50 years ago. Many possibilities for application been foreseen, from simple tasks, such as speech recognition, machines over world entirely. Famous takeover examples are space ship's computer system HAL 1968 movie 2001: A Space Odyssey, and Skynet AI-system Terminator (1984–2019). In nutshell, benefits lie its power analyse vast amounts complex data, identify certain patterns relevant specific query, produce model, then adapt continue improve model new data input. enables automation routine tasks humans would not only find monotonous, but also be unable complete same accuracy or speed. Current versions go far providing support decision-making. has yet taken by equivalent, undoubtedly crept into many our life will transform it. That poses several questions: Where we now? going? Should afraid, should embrace this transformation? There few areas computer-based left already infiltrated AI. Some mundane, robotics manufacturing, facial recognition smart phones, surveillance cameras homes, chatbots dealing customer queries, content moderation social media, predicting market developments finance, tailoring educational material schools. Other applications might still appear futuristic fast gaining momentum, self-driving cars, talking humanoid robots, spacecraft navigation. We quite reached point where AI-led technology can cure diseases, common cold cancer genetic disorders (as showcased original 1966–1969 Star Trek TV series). However, medicine greatly benefitted introduction services including highly accurate diagnostics, personalised treatment, robotic surgery. advances robots now found robot science, part labs large-scale performed minimal human oversight. Another prominent example use high-throughput phenotyping facilities used research. addition, farming, oldest activities, seen an revolution recent years. leading way precision agriculture practices monitoring health weather conditions, managing soil growth innovative approaches, vertical [1.Kumar S. et al.Vertical farming: moving environmental.Trends Sci. 2020; 25: 724-727Abstract Full Text PDF Scopus (103) Google Scholar]. What does future hold science? Several articles highlight some uses AI: prediction [2.Lou Y. al.Genomics refined: AI-powered perspectives on structural analysis.Trends 2024; 29: 123-125Abstract (2) Scholar], processing [3.Sears R.G. al.AI cell metabolic engineering.Trends 126-129Abstract (1) machine learning [4.Matese A. al.Are unmanned aerial vehicle-based hyperspectral imaging advancing crop science?.Trends 196-209Google Given above AI, there potential concerns regarding proper fast-evolving, game-changing technology. An raising question replacing permanently 2023 US actors' union strike fear actors replaced AI-generated digital twins. dispute mainly who gains control 'twin' financially, highlights create likeness quality allows improper 'deepfakes'. October issue, Geitmann Bidhendi highlighted shortcomings context research, bias lack diversity [5.Geitmann A.J. blindness language models.Trends 2023; 28: 1095-1097Abstract It important aware these ethically responsibly. Harfouche al. [6.Harfouche A.L. ethics road responsible science societal welfare.Trends 104-107Abstract Scholar] provide useful summary ahead necessary guidelines policies prevent biases implement so benefit all researchers, ethical socially way. With improving lives, what about threat actually us? Agathokleous [7.Agathokleous E. al.One hundred questions facing derived using large model.Trends 210-218Google nicely showcase how ChatGPTi approach speed up long-term project relies input, shows creativity cannot matched So far, when unique thinking required, outperform machines. predictive help make better informed decisions quickly, replace factor. extends publishing. launch 2022 shone spotlight models (LLMs). specialised class generative LLMs were designed comprehend generate language. Through memory deep techniques, able meaningful topically replies queries. answer (e.g., service chatbots) translate languages. When trained fine-tuned performance task provided initial story elements, creative human-like text. freely available BLOOM, Claude 2, GPT-J, Llama, RoBERTa), tempting leverage writing scientific manuscripts. obvious publishing, saving time information retrieval, summarising content, and/or translating foreign languages, mining literature. Elsevier/Cell Press assist editorial process years, identifying suitable reviewers, applying AI-based anti-plagiarism service, designing user-friendly workflows (a full list Elsevier's principles onlineii). member Committee Publishing Ethics (COPE), committed uphold publishing and, such, limits AI; example, editors allowed upload manuscripts tool review, evaluation, decision-making process. authors standards therefore, updated author clarify below guidance refers process, draw insights research process:•Where AI-assisted technologies they do readability Applying done oversight control, carefully review edit result, because authoritative-sounding output incorrect, incomplete, biased. listed co-author cited author. Authorship implies responsibilities attributed to, by, humans, outlined policy authorsii.•Authors must disclose adding statement at end their manuscript. published work. no doubt here stay, while secretly hoping predictions come true. am voting 'yes' curing 'no' domination. any case, keeping strict important. Of featured broadest possible range takes role here. Nevertheless, fingertips scientists century. include novel methods, click chemistry [8.Chen M.-M. al.Applications opportunities science.Trends 167-178Google well hardware materials, wearable electrochemical sensors [9.Zhou al.Wearable small-molecule detection.Trends 219-231Google nanomaterials [10.Mukherjee al.Nanovehicles melatonin: journey agriculture.Trends 232-248Google Scholar,11.Gohari G. al.Next generation chemical priming: little nanocarrier friends.Trends 150-166Google Two combination another innovation [12.Sarkar al.Cyber-agricultural systems breeding sustainable production.Trends 130-149Abstract (3) Scholar,13.Harfouche al.Metaverse innovating learning.Trends 266-267Abstract means comprehensive related topics follow throughout year. invite you explore welcome feedback [email protected]. ihttps://chat-gbt.io/ iiwww.elsevier.com/en-gb/about/policies-and-standards/responsible-ai-principles

Language: Английский

---

Unmasking Large Language Models By Means of OpenAI GPT-4 and Google AI: A Deep Instruction-Based Analysis

Intelligent Systems with Applications, Journal Year: 2024, Volume and Issue: 23, P. 200431 - 200431

Published: Aug. 27, 2024

Language: Английский

---

Generative artificial intelligence usage by researchers at work: Effects of gender, career stage, type of workplace, and perceived barriers

Telematics and Informatics, Journal Year: 2024, Volume and Issue: 94, P. 102187 - 102187

Published: Sept. 1, 2024

The integration of generative artificial intelligence technology into research environments has become increasingly common in recent years, representing a significant shift the way researchers approach their work. This paper seeks to explore factors underlying frequency use AI amongst professional environments. As survey data may be influenced by bias towards scientists interested AI, potentially skewing results perspectives these researchers, this study uses regression model isolate impact specific such as gender, career stage, type workplace, and perceived barriers using on AI. It also controls for other relevant variables direct involvement or development, collaboration with companies, geographic location, scientific discipline. Our show that who face adoption experience an 11 % increase tool use, while those cite insufficient training resources 8 decrease. Female 7 decrease usage compared men, advanced 19 Researchers associated government advisory groups are 45 more likely tools frequently than roles. for-profit companies %, medical institutions hospitals 16 15 respectively. contributes deeper understanding mechanisms driving valuable implications both academia industry.

Language: Английский

---