Dublin City University

都柏林城市大学 (Dublin City University，简称DCU) 创建于1980年，位于首都爱尔兰的都柏林。是为适应复杂多样的工业的不断变化的社会需求而建立的高等教育学府，学校设有五大学院：商学院，教育学院，工程与计算机学院，科学与健康学院，人文与社会科学院。都柏林城市大学拥有爱尔兰“企业大学”称号，大学以创新性和市场为导向，主张将学术创新与企业实践相结合。都柏林城市大学致力于提供优秀的教学和科研教育，位列2023年QS世界大学排名世界第471名。在2022年QS学科大学排名中，都柏林城市大学多个学科名列前茅，其中传媒学世界排名101-150，语言学世界排名101-150。 在2021年QS世界年轻大学排名中，位列61-70名。都柏林城市大学是“欧洲创新大学联盟”(ECIU)成员，也是首批加入“欧洲大学计划”的成员。DCU于2021年被《星期日泰晤士报》评为“爱尔兰年度大学”。2022年， 都柏林城市大学成立欧盟伊拉斯谟联合培养硕士项目——欧洲法律、数据和人工智能硕士。都柏林城市大学未来课程（Futures Programmes）获得2022年爱尔兰全国教育奖“最佳学生体验奖”奖项。 

The term artificial intelligence (AI) has recently become synonymous with deep learning using neural networks, as is applied for machine translation (MT), speech recognition, and the use of large language models (LLMs) for text generation. These are examples of subsymbolic AI based on training data that is converted to numbers and then processed via “a stack of equations – a thicket of often hard-to-interpret operations on numbers” (Mitchell 2020, 24). As a result, systems are often referred to as an impenetrable ‘black box’. Bender et al. (2021, 8) identified environmental and financial costs behind LLMs, and further commented that the apparent “fluency and coherence” of their output raises risks because people interpret text as “meaningful and corresponding to the communicative intent of some individual or group of individuals who have accountability”. However, output has been found to encode hegemonic societal views from training data, particularly from polarised content scraped from the web, producing output that is at times problematic, biased, and offensive. In order to minimise this effect, some developers use reinforcement learning using human feedback, for example OpenAI employ staff (remotely and on low pay) to make ChatGPT output less toxic (Perrigo 2023).

The proposed interdisciplinary PhD project will investigate the ethics of multilingual communication using LLMs and will be supervised by Dr Joss Moorkens (School of Applied Language and Intercultural Studies/SALIS) and Prof Bert Gordijn (Institute of Ethics). There has so far been relatively little research produced on this topic. For example, Bender et al. (2021) considered potential dangers and ethical issues related to LLMs and Mitchell and Krakauer (2022) considered the debate about whether LLMs can be said to understand content. However, these did not focus on multilingual communication and were written before ChatGPT launched on November 30th, 2022, reaching one million users in five days. The intervening months have seen some articles on the translation and translation evaluation capabilities of LLMs, along with one short article by Moorkens (2023), but now would appear to be the perfect time to systematically assess the ethical aspects of the development and use of LLMs for translation and multilingual communication in order to accommodate responsible research and innovation.

On completion of this project, the graduate should have published work that adds to the field. Experience in this topic should be highly beneficial, leading to opportunities for further research at postdoctoral level in industry or academia. There is an expectation that the student would contribute to teaching in SALIS, which should also increase employability on graduation. The project is co-funded by the ADAPT Centre and Dublin City University.

Informal enquiries may be directed to Joss Moorkens at joss.moorkens@adaptcentre.ie

信息来源：语言治理

点击“阅读原文”可跳转官网

推  荐

欢迎加入

今日小编：心得君

审     核：心得君

转载&合作请联系

"心得君"

微信：xindejun_yyxxd

点击“阅读原文”可跳转官网