An online AI or machine learning course can be checked against the CS2023 curriculum report, the joint ACM, IEEE Computer Society, and AAAI guidelines for undergraduate computing programs, which were endorsed between January and February 2024. The report lists what an AI course should teach, which math it assumes, and how many hours each topic deserves. Whether a certificate carries any weight is a separate question, answered by who issued it and whether that institution is recognized where the learner lives.
Key takeaways
- CS2023 has a dedicated AI knowledge area with 12 hours of core material shared by all computer science graduates and 18 further hours in the area's own core.
- For AI, the report lists algebra, precalculus, discrete mathematics, linear algebra, and probability and statistics as required, with calculus desirable.
- Its suggested introductory AI course allots 12 of its listed hours to machine learning, and a dedicated machine learning course allots 32.
- Canada's CICIC directory covers only currently recognized, authorized, registered, or licensed institutions, which gives a way to check who stands behind a program.
- Research on Harvard and MIT online-course data found that low MOOC completion rates had not improved over six years.
What CS2023 expects an AI course to contain
The CS2023 report divides the AI knowledge area into a dozen units, running from foundations and search through machine learning and probabilistic reasoning to planning, agents, language processing, robotics, vision, and societal impact. Its table of core hours gives machine learning four hours in the shared core and six in the area core, the same area-core weight as search.
The report also tells instructors to keep an eye on recent advances. It says that since the first draft, techniques such as generative networks and large language models have become widely used and were added to the cores, and that faculty should consider whether new advances belong in the curriculum. A course page that never mentions evaluation, data preparation, or ethics is missing material the report treats as basic.
The machine learning unit, item by item
Within the machine learning unit, the report's shared core names supervised learning (classification and regression), reinforcement learning, and unsupervised learning with clustering. It adds the no free lunch theorem, sources of error, simple models such as linear regression or decision trees, overfitting and the bias-variance tradeoff, data preprocessing including missing values and categorical encoding, and the separation of training, validation, and test sets. It also lists basic neural networks, an introduction to generative networks such as large language models, and ethics, with bias in datasets and evaluation. One phrase stands out: the report asks for emphasis on real data, not textbook examples. A syllabus that never uses a messy dataset deserves a question.
Math a course should state up front
For AI, CS2023 requires algebra, precalculus, discrete mathematics, linear algebra (matrix operations and basis sets), and probability and statistics, including conditional probability, independence, and Bayes' theorem. It lists calculus-based probability, single-variable and partial derivatives, and optimization as desirable. A course that promises machine learning with no mention of any of this is either skipping the mechanics or leaving the learner to find the gaps alone.
Two course shapes the report suggests
| Course shape | Hours the report assigns to each unit | Prerequisites it names |
|---|---|---|
| Introductory AI | Introduction 4, search 9, knowledge representation 4, machine learning 12, probability 5, societal impact 4 (integrated throughout) | Programming fundamentals, data structures, algorithms, discrete mathematics, probability |
| Machine learning | Machine learning 32, knowledge representation 4, natural language processing 4 (selected topics), societal impact 4 (integrated throughout) | Listed in the report; not reproduced here |
The hours add up to 38 for the first shape and 44 for the second. Short online offerings usually fall below that, which is fine for a taster but should not be confused with a degree-level course.
Who stands behind the certificate
CS2023 is written for degree programs, so a certificate of completion is a different thing from academic credit. In Canada, CICIC, a unit of the Council of Ministers of Education, Canada, runs a directory of educational institutions. The directory states that it contains only currently recognized, authorized, registered, or licensed institutions, and results can be filtered by institution type and by province or territory. The page does not say whether a listed institution may grant degrees.
In the United States, the Department of Education does not itself accredit schools. It recognizes accrediting agencies, and its public database of accredited postsecondary institutions and programs lists the institutions those agencies and state approval agencies report. That database could not be opened for this article, so readers should look the institution up there directly. In both countries, a course platform's own badge is not the same as a recognized credential.
Why finishing is the hard part
A 2019 analysis in Science by Justin Reich and José Ruipérez-Valiente used online-course platform data from Harvard and MIT. It reported that the vast majority of learners never return after their first year, that growth in participation was concentrated almost entirely in the world's most affluent countries, and that low completion rates had not improved over six years. The authors did not frame this as a flaw in any one course; they were describing the open online model as a whole. For a prospective learner, the practical reading is to look for features that the course page can state, such as deadlines, graded assignments, and a human contact, and to treat a high enrollment count as a statement about interest, not results.
A checklist for reading a course page
- Does the syllabus name specific topics, such as the units above, or only broad promises?
- Are math and programming prerequisites listed, and do they match the content?
- Do assignments use real datasets and include evaluation on held-out data?
- Is there a unit on bias and societal effects?
- Who issues the credential, and does that body appear in an official directory for the learner's country?
- What is the refund and cancellation policy, and is the total cost shown before payment?
This article offers general information and does not rank, endorse, or recommend any provider. Prices, course contents, and credential recognition change, so the provider's current page and the official directories are the places to confirm details.
FAQ
Is there one best online AI course?
No source cited here ranks courses, and the right choice depends on prior math, programming experience, and the goal. The CS2023 topic list and the checklist above give a way to compare syllabi on the same terms.
How much math does machine learning need?
CS2023 requires linear algebra and probability and statistics for AI and lists calculus as desirable. The report treats the listed topics as the foundation, so a syllabus that skips them should explain how it covers the same ideas.
Does a certificate of completion count as a degree credit?
Not automatically. Credit depends on the issuing institution and on whether a receiving school or employer accepts it. Recognition status can be checked in official directories such as CICIC in Canada.
Why do so few learners finish open online courses?
The 2019 Science analysis documented low completion that had not improved over six years, without assigning a single cause. Courses with fixed deadlines and graded work are structured differently from fully self-paced ones.
Related reading
- what CS50 covers, week by week
- Stuart Russell and his AI textbook
- Marvin Minsky and perceptrons
- Timefold, an open-source planning solver
More in AI Basics.
Last verified: October 2026
Sources
- Computer Science Curricula 2023 (CS2023), ACM, IEEE Computer Society, and AAAI, 2024
- Directory of Educational Institutions in Canada, CICIC, Council of Ministers of Education, Canada
- The MOOC pivot, Reich and Ruipérez-Valiente, Science 363(6423), 2019
- Accreditation notices and the Database of Accredited Postsecondary Institutions and Programs, U.S. Department of Education, 2025 to 2026