Start with a self-test, then run a small reproducible task using real data. Answer feedback and review results stay in this browser, separate from reading progress.
These eight introductory questions sample a few topics; they are not an assessment of every concept on the site. Answer independently before reading feedback.
Task: predict one of three Iris species from four sepal/petal measurements. The material contains 150 real observations. Introductory level; allow 20–30 minutes including code reading. Basic lists, means and distances are prerequisites.
Download CSV · Download Python · Expected results JSON · Source and license
python practice-iris.py
This fixed procedure should classify 25/30 correctly (about 83.33%). Matrix rows are true classes, columns predictions, ordered setosa, versicolor, virginica: [[10,0,0],[0,5,5],[0,0,10]].
This is a teaching baseline on a small dataset, not a best-score claim or evidence of performance on other flowers or datasets. The lab uploads no data and trains no online model.
Source: Fisher, R. (1936), Iris — UCI; DOI 10.24432/C56C76; CC BY 4.0
Source: scikit-learn — Data leakage
The download uses UCI’s corrected bezdekIris.data, adding a header and normalized newlines while preserving row order. Retrieved 2026-11-01.
These are locally written reading aids, not full translations. Check the originals. Previews do not embed external sites or automatically send reading content.
Read the abstract for the problem, method and evaluated tasks, then inspect how retrieval connects to generation. A gain on evaluated tasks does not establish reliability for every question. Terms: retriever; generator; knowledge-intensive.
Source: Lewis et al., Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (2020)
Map the architecture diagram to encoder, decoder and attention, then inspect training and evaluation. Translation results have specific experimental conditions and cannot be generalized to every modern model. Terms: encoder; decoder; self-attention.
Focus on splitting first and fitting on training data only, then distinguish fit from transform in the examples. This lab computes means and scales on training data; test data is used for final evaluation.
Source: scikit-learn — Common pitfalls and recommended practices
Models can produce emotional language and detect some emotional cues. Those observations alone do not establish subjective feelings. How to identify consciousness and experience remains debated. Distinguish observable behavior, mechanisms and subjective experience.
The 1955 proposal for the Dartmouth summer research project used “artificial intelligence”; the project was planned for 1956. The name denotes a research direction, not a claim that machines have every human mental property. Distinguish the proposal date from the event year.
Source: McCarthy, Minsky, Rochester & Shannon — Dartmouth research proposal (1955)
This is an open question. Discuss technical capability, sustained resources and choices about deployment separately. Capability and adoption are different; costs, policy, risks and public choices affect the path. State a testable condition and evidence that could change your view rather than treating a position as a single-answer quiz.
Source: Stanford AI100 — historical background (2016; not a forecast)
Compiled 2026-11-01. Papers retain their publication year; disputed research questions are not graded by a single position.