英国AI本科 院校梯队 ALevel选科 TMUA 面试
英国本科阶段单独开设人工智能专业的学校不算多，大部分项目长在计算机科学系里，第三年才分方向。中国学生看排名选校，往往忽略了一件事：课程表里数学占多少，比学校名字更影响四年后的出路。人工智能在本科阶段是一门吃数学的学科，线性代数、概率、优化这三块基础不牢，后面的机器学习课只能靠背结论过关。这篇按申请链条拆开讲：选校看什么，A-Level 怎么搭，笔试与文书怎么准备，面试会被问到哪一步。文中的录取要求以 2027 入学为参照，各校每年会调，报名前请以学校官网与 UCAS 当年页面为准。
Only a handful of UK universities run artificial intelligence as a stand-alone undergraduate degree; most teach it inside computer science and let students specialise in the final year. Families often choose by league table position alone, which misses the part that matters most: how much mathematics the timetable actually contains. Linear algebra, probability and optimisation carry the whole degree, and a weak foundation there turns every machine learning module into memorisation. The sections below follow the application chain from course choice to interview, using 2027 entry as the reference point; requirements shift each year, so check the university website and the UCAS page for the year you apply.
课程骨架 前两年数学 后两年方向
英国的人工智能本科大多不是独立系，而是计算机科学系里的一个方向，课程表前两年几乎一样。大一通常是离散数学、编程入门、算法与数据结构，大二加概率统计、线性代数、计算理论与操作系统。真正的机器学习、计算机视觉、自然语言处理排在大三和大四，前面两年是在为这些课铺路。爱丁堡的 BSc Artificial Intelligence 是少数从入学就以此命名的项目，课程仍然以数学与编程打底。很多学生以为 AI 专业一开学就教深度学习，翻开第一学期的时间表会发现大半是证明题。这种安排不是保守，是因为没有概率与线性代数，调参只能靠试。选校时把四年课程表拉出来对一遍，比看排名有用。
British AI degrees usually sit inside a computer science department, and the first two years look almost identical across universities. Discrete mathematics, programming, algorithms and data structures come first, followed by probability, linear algebra, theory of computation and operating systems. Machine learning, computer vision and natural language processing only appear in the third and fourth years, once the mathematical groundwork is in place. Edinburgh is one of the few places where the degree is named artificial intelligence from day one, yet its timetable is still built on mathematics and code. Applicants who expect deep learning in week one are usually surprised by a term of proofs.
院校类型 宽口径与专业名之分
按项目类型分，英国的人工智能本科大致落在三类学校里。剑桥与牛津的计算机科学是宽口径课程，AI 内容分散在选修模块与毕业项目里。帝国理工的 Computing 系在三年级分出机器学习与人工智能方向，UCL 的计算机科学同样允许高年级选 AI 模块。爱丁堡、曼彻斯特、南安普顿、伦敦国王学院这几所把 AI 写进了专业名，课程更聚焦。选哪一类取决于你想要宽还是想要专，前者转向空间大，后者动手能力来得早。国际生学费与城市成本差别不小，伦敦几所的年开销通常比其他城市高一截。
UK artificial intelligence degrees fall into three rough groups. Cambridge and Oxford teach computer science as a broad degree, with AI appearing in optional modules and the final-year project; Imperial splits its computing degree into a machine learning and artificial intelligence stream in year three, and UCL allows similar specialisation. Edinburgh, Manchester, Southampton and King's College London put AI in the degree title, which makes the syllabus tighter but leaves less room to switch. Cost matters as well: London living costs sit well above most other cities, and international fees vary widely between institutions.
A-Level 选科 数学与进阶数学的分量
数学是几乎所有项目的硬性要求，进阶数学在很多学校的官网上被写成 strongly recommended 或 preferred。剑桥、帝国理工、UCL 的计算机类项目在筛选时会把进阶数学当作数学能力的证据，缺这一门不代表没机会，但要在别处补上同等强度的证明。计算机科学 A-Level 有帮助，不是必需品，物理对理解硬件、信号与嵌入式系统有实际用处。常见的错误是把四门选课全押在理科上，却忽略了书面表达，个人陈述与面试都要把技术想法讲清楚。选课表定下来之后，Year 12 的暑假就该开始补线代与概率的基础，因为这两块在 A-Level 里讲得浅。
Mathematics is required almost everywhere, and further mathematics appears on many university websites as strongly recommended or preferred. Cambridge, Imperial and UCL treat further maths as evidence of mathematical maturity; without it an applicant needs an equivalent signal elsewhere, such as a strong admissions test or an olympiad result. Computer science at A-Level helps but is rarely mandatory, while physics pays off when the course touches hardware, signals and embedded systems. A common mistake is filling all four options with sciences and leaving no room to practise written argument, which the personal statement and interview both demand.
入学笔试 MAT TMUA 与 STEP 的分工
笔试这一环变化最快，同一所学校隔两年换一张卷子的事发生过不止一次。牛津的计算机科学与数学方向使用 MAT，剑桥的计算机科学近年转向 TMUA，帝国理工的计算机类项目也把入学测试纳入筛选。STEP 主要服务于剑桥的数学方向，计算机申请者一般不需要，但拿到有条件录取后可能被要求考。报名节点通常在申请季的秋季，考试在十月或十一月，错过就没有补考。具体到某个项目用哪张卷、什么时候报名，只能查学校官网当年的 admissions tests 页面与考试机构公告。准备方式不复杂，把近五年的真题按题型刷一遍，错题整理成一份自己的清单。
Admissions tests move faster than any other part of the process, and a university can switch papers within two cycles. Oxford uses the MAT for computer science and mathematics, Cambridge has moved its computer science admissions towards the TMUA, and Imperial now includes a test in its computing selection. STEP belongs mainly to Cambridge mathematics and is rarely needed for computing, though a conditional offer can still require it. Registration closes in the autumn before the January deadline, with sittings in October or November and no second chance, so the paper and deadline for your course have to be confirmed on the university's admissions tests page.
语言与成绩换算 官网两个页面
语言成绩的门槛看起来明确，实际上每所学校都有自己的换算表。多数英国院校把雅思总分要求放在 6.5 到 7.5 之间，单项在 6.0 到 7.0 之间，计算机类项目通常取偏高的那一档。部分学校接受托福、PTE 与 IB 英语，也有的对在英语授课学校读满两年的学生免语言。成绩换算同样有坑，A-Level 的 A* 与 IB 的 7 分不是简单对应，学校会把科目组合一起看。查要求时看官网的 entry requirements 与 English language requirements 两个页面，别只看第三方汇总站。这两页每年秋季更新，申请当年的版本才算数。
Language requirements look precise, but every institution runs its own conversion table. Most UK universities ask for IELTS between 6.5 and 7.5 overall with 6.0 to 7.0 in each component, and computing courses usually sit at the higher end. Some accept TOEFL, PTE or IB English, and a few waive the test for students who spent two years in an English-medium school. Grade conversion is another trap: an A* at A-Level and a 7 in the IB are not equivalent, since subject mix is read alongside the grades. Use the entry requirements and English language pages on the university site rather than an aggregator, and check the version published for your application year.
个人陈述 三问结构与具体场景
UCAS 从 2026 入学申请季起把个人陈述改成三个固定问题，字数上限仍是 4000 字符。三个问题分别问为什么选这个专业、你的学历与学习经历如何支撑它、以及课程之外你做过什么准备。这种结构对中国学生反而友好，不用再猜招生官想看什么开头。常见的错误是把项目经历写成产品说明书，堆技术名词却不解释你为什么做、遇到什么、怎么改。招生官每年读几千份文书，能记住的是具体场景，比如为了让一个识别模型在低光环境下工作，你改了哪一步预处理。三个问题的答案要能互相咬合，别写成三段互不相干的自述。
From the 2026 entry cycle UCAS replaced the free-form personal statement with three fixed questions, still capped at 4,000 characters. They ask why you want the subject, how your qualifications and study so far support it, and what you have done outside the syllabus. The structure removes the guessing game about openings, but it also exposes thin content faster. A common mistake is turning a project into a product brochure, listing tools without saying why you built it, what broke and what you changed; admissions tutors read thousands of these and remember the specific problem, not the stack.
面试 现场推理与可教性
牛津与剑桥的计算机面试以解题为主，题目常常超出课本范围，考的是现场推理。面试官会给一个陌生问题，看你能不能把已知条件整理成可处理的步骤，卡住时会不会说出来。帝国理工与部分其他学校也有面试或线上评估，形式更接近对话。很多学生以为面试是考知识量，实际上考官在测你的思维是否可教。答错不要紧，沉默才致命，把思路讲出来让考官有机会给提示，这一步很多人做不到。练法是把近年的面试题当成口头练习，出声讲给同学听，别只在纸上推。
Interviews at Oxford and Cambridge are problem-solving sessions, and the questions usually sit outside the syllabus. The interviewer wants to see how you organise unfamiliar information into steps, and whether you say something when you get stuck. Imperial and several other universities also interview or run an online assessment, though the tone is closer to a conversation. Many students assume the interview measures how much they know; in practice it measures whether their thinking can be guided, so a wrong answer explained out loud beats silence.
项目与作品 小问题做扎实
AI 申请里最容易被识破的是套模板的项目，教程里的泰坦尼克号生存预测写进文书，招生官一眼就能认出来。能站住的项目通常很小，问题来自你自己的生活，比如帮学校社团做一个报名系统，或者用公开数据集分析本地公交的准点率。重点不在模型多复杂，而在你有没有定义清楚问题、有没有验证、结果不对时怎么处理。代码放到 GitHub 上，附一段说明写清数据来源与限制，比奖状有用。做两个小项目并写出复盘，胜过堆五个半成品。
The projects that get recognised as templates are the ones copied from tutorials, and a Titanic survival model on a personal statement reads as filler. Work that holds up tends to be small and local: a sign-up system for a school society, or an analysis of bus punctuality using an open dataset. Complexity is not the point; a clear problem definition, a validation step and an honest account of what failed carry more weight. Put the code on GitHub with a short note on data sources and limitations, which tells an admissions tutor more than a certificate.
三个高频误判 排名 课程 语言
很多学生以为英国本科的 AI 专业会教最新的大模型技术，实际课程更新要经过审批，讲的多是五到十年前就稳定的方法。另一个高频误判是把排名当成唯一标准，同一门课在不同学校的数学强度可以差一整年。还有人对语言成绩掉以轻心，总分够了单项差 0.5 被要求重考的情况每年都有。选校时把课程表、笔试要求、面试形式三样放在一起看，比单看一个榜单靠谱。这三样都能在官网查到，花两个晚上就能整理成一张对照表。
Many students expect a UK AI degree to cover the newest large model techniques, but syllabuses pass through approval rounds and usually teach methods that have been stable for five to ten years. A second misjudgement is treating a league table as the decision, when the same subject can differ by a full year of mathematics between two universities. Language scores get underestimated as well: an overall pass with one component half a band short leads to a retake every year. Compare the timetable, the admissions test and the interview format side by side, all of which the university publishes.
时间线 Year 12 暑假到面试季
时间线可以倒推。Year 12 的暑假确定方向并开始准备笔试，九月到十月完成报名，十月或十一月考试。UCAS 的常规截止在一月中旬，牛津与剑桥以及多数医学类项目在十月中旬。Year 13 的十一月到二月是面试季，部分学校在这段时间发有条件录取。语言考试建议在 Year 12 结束前考出可用分数，避免和面试、大考挤在一起。拿到 offer 后别松手，条件里的成绩要求才是真正的门槛，A-Level 最后一年掉档的例子每年都有。
Work backwards from the deadline. The summer of Year 12 is when the subject choice and test preparation should start, with registration in September or October and the test itself in October or November. The UCAS deadline for most courses falls in mid-January, while Oxford, Cambridge and most medicine courses close in mid-October. Interviews run from November to February, and some offers arrive during that window. Language tests are best finished before Year 12 ends, so they do not collide with interviews and final exams.
学而思学习机X5Pro经典培优【256G内存翻倍 享两年只换不修】小初高全学科 焕新升级12.6英寸6+256GB¥3999.0📱 去京东购买京东精选
A-Level / GCSE / IB / KS2-KS3 全科辅导，AI 与计算机方向申请规划
16621398022 同微信
更多咨询请联系16621398022（同微信），或访问 aleveler.com
或截图后用「扫一扫 → 相册」选择本图