AI Summit
Climb from base camp to the summit by using AI wisely at work, in twelve questions, and what a fork can customise.
The player climbs a mountain from base camp to the summit by answering twelve questions about using AI at work: hallucinations, verification, bias, personal data, human review, the right tool for the job and who is accountable. Each right answer gains a step of altitude; a wrong one costs nothing, and the climb goes on. A quiz played solo in 6 to 10 minutes.
Play it now, for free and without an account: the AI Summit game page.
Rules
- Each question shows its answers: one is right. On the keyboard, A to D (or 1 to 4) answer, F plays the 50:50 and H the hint; with a mouse or a finger, the answer cards.
- After the answer is locked in, the suspense, then the reveal: the right answer and the explanation, before "Continue".
- Lifelines, once per climb: 50:50 removes two wrong answers, Hint shows a clue. They never change the score; they are reported per question to the organisation.
- Timer per question (30 s by default): a question left unanswered counts as wrong.
- Camps at steps 4 and 8 are celebrated on the way up; only a perfect climb reaches the summit.
- The game keeps its own end screen: result and altitude, the skills acquired and to work on, and a review of each question (your answer, the right answer, the explanation).
The score is the quiz: one point per right answer at the first try, out of twelve, recomputed by the server from the answers (the platform's common quiz scoring). The altitude reached is shown as game points, never in the score. The pass mark (60% by default) is the one sent to the LMS.
Skills measured
| Skill | Questions |
|---|---|
ai.critical.hallucination: spotting hallucinations |
confident tone, invented sources |
ai.critical.verification: knowing when to check |
quotes, figures, primary sources |
ai.critical.bias: detecting bias |
indirect clues, "typical" images |
ai.usage.personal-data: protecting personal data |
rewording an email, minimum data |
ai.usage.human-review: reviewing before publishing |
a draft you review |
ai.usage.good-fit: knowing when AI fits |
low-stakes language work |
ai.usage.right-tool: choosing a script over a model |
exact arithmetic on many rows |
ai.usage.accountability: keeping a human accountable |
who answers for a message or a report |
What a fork can customise
The editor of a fork has four spaces (Questions, Appearance and Rules, and Brand).
Questions: the twelve questions, their answers, the right answer, the explanation, the hint, the skill and the difficulty; add, remove or reorder questions, import or export them as CSV. AI Summit presents the five question types of the platform: one answer, several answers, an order, pairs and a value in a range. The ladder always has one step per question of the game.
Rules:
| Rule | Default | Effect |
|---|---|---|
summitAltitude |
8,848 m | Altitude of the summit; the steps lengthen towards the top. |
safeStepA, safeStepB |
4, 8 | Camps (step numbers, 0: none). |
fiftyFifty, hint |
on, on | The two lifelines, once per climb. |
timer, timerSeconds |
on, 30 s | Timer per question. |
sessionDurationSeconds |
no timer | Maximum game length. |
passingScorePercent |
60 | Pass mark sent to the LMS. |
Appearance: every interface text (welcome screen, buttons, end screen), in the fork's source language (other languages are translated with a file).
Brand: main, secondary, background, surface and text colours, font, corners and logo, applied live in the game, with the AA contrast checked as you choose. The logo shows at the bottom left of the game. See where the brand applies.
Where it runs
AI Summit is the first game of the platform that runs isolated in its own frame: its code and
rendering run in a sandboxed frame served by play.maketools.ai, with no access to the player's
account, cookies or the rest of the site; the platform keeps the session, the score and the end of
the game. For a network that filters addresses, allow maketools.ai and play.maketools.ai
(or *.maketools.ai, see network access). In an LMS (SCORM 1.2 or 2004, opened in a window or shown in the LMS page),
the detail of each question and one objective per skill reach the LMS
(LMS integration).
Review the content
The twelve questions are listed with their answers, explanations and hints in
games/ai-summit/content-review.md (generated by bun run content:review in the game package),
in English and French.