Carol导师介绍

裁判经历

曾在多个重要辩论赛事中担任裁判长，包括清华校际联合辩论赛、超新星辩论锦标赛等，在辩论赛事领域拥有丰富经验。

辩论经历

多次参加英国议会制辩论赛获得最佳辩手，并入围国际辩论赛事。曾获得广东辩论挑战赛最佳辩手、亚太大学生辩论锦标赛最佳辩手，也是世界大学辩论锦标赛英文非母语组半决赛选手，亚洲联合辩论锦标赛英文非母语组四分之一决赛选手。

Generative AI is a branch of AI that focuses on generating new content such as Images, Text, Audio, Video based on patterns learned from existing data. Unlike other forms of AI, such as predictive AI, which forecasts future outcomes, or analytical AI, which provides insights from input data, generative AI is designed to produce original content that can be similar to human-created material. For example, the famous Generative Pre-trained Transformer （GPT）model can generate text almost indistinguishable from human written text. Typically, it works by taking a prompt as input and generating a corresponding output based on that prompt.

The market for generative AI is expected to skyrocket to $207 billion by 2030, a massive increase from $5.67 billion in 2020[1]. Educational institutions are leading this growth, with 58% of university instructors already using AI tools in their daily teaching according to the latest statistics[2].

AI Integration in Education: A Map of Tools and Applications [3]

Due to space constraints, some of the mentioned examples do not include detailed statistics or citations.

Some argue that artificial intelligence might take away jobs, particularly in roles involving repetitive and predictable tasks like customer services or jobs requiring high accuracy and reliability such as diagnosing diseases from medical images. However, AI also creates new industries and jobs like prompt engineers, and drive business expansions by lowering cost, potentially creating more jobs. Additionally, by automating mundane daily tasks, people can focus on more meaningful work.

Training AI requires large amounts of data, including biometrics(e.g., fingerprints, iris scans) and behavioral data (e.g., purchase / browse history). Usersʼ data are being collected without noting, as most people do not thoroughly read or understand privacy agreements. This extensive data aggregation can infringe on privacy, allowing AI models learn and predict personal behaviors such as voting preferences, which can then be used to manipulate online experiences like Youtube feeds. Improper data handling might also have security problems such as data leakage.

AI basically learns from existing data, extract their patterns and generate content similar to them, which raises serious copyright issues. For example, The New York Times has sued OpenAI and Microsoft for the unpermitted use of Times articles to train GPT large language model. Even if AI-generated output are not near-verbatim of original works, the synthetic output still derives from the original content, capturing implicit patterns. The extent to which AI usage infringes on original content remains significant issue, especially as AI-generated works are becoming more popular.

If AI goes wrong, itʼs hard to determine who is responsible. Itʼs impossible to hold programmers accountable, since they only create the model architecture(AI is literally a black box) and do not do case-by-case test. Additionally, general public often lacks the capability to understand rationality of the model or have access to its specifics, making it harder for them to call AI out. This issue is particularly concerning in critical areas like military applications, healthcare and autonomous vehicles, where AI decisions are made in such short notice and have such profound impacts, making it essential to establish clear accountability.

AI models might prone to bias because they learn from historical data, which might have racist / sexist content. This lead to biased decision-making and discriminative analysis. For example, artificial intelligence used in criminal justice system would tend to impose harsher penalties on black defendants based on historical data showing higher recidivism rates. Whatʼs more, AI often lacks adequate data on minority groups, leading to unfair treatment. For example, facial recognition algorithms in the U.S. are more accurate for white people and less so for other ethnic groups.