AI World Cup

12.1 Event Concept and Philosophy#

Why Competition?#

The AI Fellowship exists to build community. So why competition?

Because competition, done right, is one of the most powerful tools for community building that exists. It creates shared stakes, common timelines, stories worth telling, and moments of genuine triumph that unite people. The ImageNet Large Scale Visual Recognition Challenge (2010-2017) drove a decade of computer vision progress because it gave researchers a common benchmark and a reason to show up. Kaggle built one of the largest data science communities on earth around structured competitions. RoboCup has inspired a generation of robotics engineers through the concrete goal of building a robot that can beat the human World Cup team by 2050.

The AI World Cup is The AI Fellowship’s version of this idea - a global, chapter-anchored competition that gives the community a shared north star every year, creates chapter-to-chapter connection and friendly rivalry, surfaces the community’s best work to the world, and gives members something to train for, build toward, and remember.

Why “World Cup”?#

The World Cup format is deliberate. It is instantly understood globally, it implies local qualification and global final, it implies national/regional pride without nationalism, and it implies that what happens in your city matters for what happens on the global stage. It makes a local hackathon in Nairobi feel like it means something beyond Nairobi - because it does.

12.2 Competition Structure#

The AI World Cup follows a three-tier structure:

LOCAL QUALIFIERS (Run by individual chapters) ↓ REGIONAL SEMIFINALS (Run by Central in coordination with regional chapter clusters) ↓ GLOBAL FINAL (Run by Central; hosted in a rotating city)

Tier 1: Local Qualifiers#

• Who runs them: Individual chapters, following the standard hackathon runbook (Document 11) with AI World Cup-specific additions.
• Format: 1-2 day hackathon. Teams of 2-5 compete on the global challenge brief (issued by Central).
• Output: Each qualifying chapter sends its top 1-3 teams to the regional semifinal. (Number determined by chapter size tier - see §12.6.)
• Timeline: Qualifiers happen within a 6-week window set by Central (typically Month 3-4 of the 6-month calendar).

Tier 2: Regional Semifinals#

• Who runs them: Central, in coordination with a host chapter.
• Format: 1-2 day hybrid hackathon (in-person preferred; virtual track available for chapters without travel budget).
• Regions: Initially 4 regions: Africa, Europe/Middle East, Americas, Asia-Pacific. Regions may be adjusted as the community grows.
• Output: Each regional semifinal sends its top 2-3 teams to the Global Final.

Tier 3: Global Final#

• Who runs it: Central team.
• Format: 2-day in-person event. All finalist teams are present.
• Host city: Rotates annually. Chosen from among strong chapters based on venue availability, visa accessibility for international participants, and chapter capacity to co-host.
• Output: Winners announced live. Results published globally. Chapter standings updated.

12.3 Eligibility#

Who Can Compete#

• Any current AI Fellowship member in good standing (pledge signed, chapter active)
• Members who have joined at least 30 days before their local qualifier
• No academic or professional background required - open to all skill levels

Team Sizes#

• Minimum: 2 members
• Maximum: 5 members
• Mixed-chapter teams: Allowed for regional and global stages. At local qualifier, at least 2 of 5 team members must be from the hosting chapter.

Age Rules#

• Minimum age: 16 (with parental consent form on file - see Document 11 safeguarding protocol)
• No maximum age
• Youth category (16-21): Optional for chapters to run a dedicated youth track at the qualifier stage. Central may recognize a Youth Champion award at the global level.

12.4 Challenge Format Options#

Central selects one of the following formats for each year’s competition. The format is announced when the 6-month calendar opens (Month 1).

Option A - Single Global Challenge#

All teams, worldwide, work on the same problem statement. Creates maximum comparability and a clear global narrative. Best for early years when the community is still establishing shared norms.

• Example: “Build an AI system that helps smallholder farmers in any country make better crop management decisions.”

Option B - Category Tracks#

Teams self-select into a track (e.g., AI for Health, AI for Education, AI for Climate, AI for Creative Industries). Each track has its own challenge statement. Winners per track, plus an overall winner.

• Allows broader participation; harder to compare across tracks; more complex logistics.

Option C - Open-Ended with Rubric#

Teams propose and build any AI application. Judged purely on the rubric (see §12.5). Maximum creativity; harder to create a compelling narrative.

• Best for advanced years once the community has established strong judging culture.

Recommendation for Year 1: Option A. A single challenge creates unity and simplifies logistics. Move to Option B by Year 3 as chapter diversity increases.

12.5 Judging Criteria and Rubric#

All AI World Cup judging uses this standardized rubric. Chapter qualifiers must use this rubric (weights are fixed for AI World Cup; chapters may adapt for non-qualifier hackathons).

Total: 100%

Judging Panel Composition#

• Minimum 3 judges per panel (always an odd number to prevent ties)
• At least 1 judge must be a practitioner with technical AI experience
• At least 1 judge should have a non-technical background (product, social impact, policy) to assess real-world impact dimension
• Judges declare conflicts of interest in writing before judging begins; conflicted judges recuse themselves from specific team evaluations

12.6 Timeline Template - 6-Month Pre-Event Calendar#

Key Central deadlines#

• Challenge brief: Published Day 1 of Month 1 (no delays - chapters need time to plan)
• Qualifier confirmation: Chapter must confirm their qualifier date by end of Month 2
• Results submission: Within 7 days of each qualifier
• Global Final date: Fixed in the calendar 12 months in advance

12.7 Chapter’s Role in Mobilizing and Hosting Qualifiers#

Chapters are essential to the World Cup’s success. Their responsibilities:

Mobilization#

• Promote the qualifier to all chapter members starting Month 1
• Actively form teams - don’t wait for people to self-organize
• Run team formation events (workshops, meetups) in Months 1-2 to help people find co-founders
• Create chapter-level hype: leaderboards, team spotlights, countdown posts

Hosting#

• Follow Document 11 event runbook for the qualifier hackathon
• Use the standardized submission format and judging rubric (no adaptations for AI World Cup qualifiers)
• Record all submissions and scores in Central’s shared tracking sheet
• Submit event report and qualifier results to Central within 7 days

Chapter size tiers and qualifying spots#

12.8 Prizes and Recognition#

The AI World Cup prioritizes recognition and opportunity over cash. The community is global and diverse - cash prizes create currency inequality and tax complications. Recognition travels better.

Global Final - Winning Team#

• AI World Cup Champion Trophy (physical, rotating - returned next year and re-engraved)
• Featured profile on theaifellowship.com for 1 year
• Speaking slot at next year’s Global Final (sharing what they built and learned)
• Letters of recognition from The AI Fellowship (for university applications, grant applications, portfolios)
• Access to Central’s partner network for mentorship and project support (e.g., intros to relevant companies or researchers)

Regional Semifinal Winners#

• Regional Champion recognition on website and social media
• 1 qualifying spot to next year’s Regional Semifinal (team can reconstitute)

Local Qualifier Top Teams#

• Chapter recognition: featured on chapter social media and in #wins-and-shoutouts
• Certificate of achievement

Chapter Standing#

Central maintains a rolling chapter standing that aggregates qualifier performance over time. This is a community metric - it celebrates active chapters and creates healthy inter-chapter motivation. Standings published annually.

12.9 Anti-Cheat and Integrity Rules#

AI competitions require integrity norms that traditional software competitions did not. The question “what counts as fair use of AI tools?” must be answered explicitly.

What Is Permitted#

• Using any publicly available AI tools, APIs, models, or libraries (OpenAI, Anthropic, Hugging Face, etc.)
• Using pre-trained models as a foundation and fine-tuning them
• Using AI coding assistants (GitHub Copilot, Cursor, etc.) - must be disclosed in submission
• Referencing publicly available datasets
• Building on open-source repositories - must be attributed and materially extended, not just repackaged

What Is Not Permitted#

• Submitting work done entirely or primarily before the competition began (code freeze is at the start of the hackathon window)
• Misrepresenting the nature of the AI system (“our model” when it is entirely a third-party API call with no original contribution)
• Using proprietary data without authorization
• Submitting another team’s work (any member of a disqualified team is banned from future AI World Cup participation for 2 years)
• Undisclosed use of paid AI services that create an unfair resource advantage (if a team uses significant paid compute, this must be disclosed - judges may factor this in)

Process for Integrity Complaints#

1. Any team, judge, or observer may file an integrity complaint with Central during or within 7 days of an event.
2. Central reviews the complaint with the relevant chapter Director and at least one independent judge within 48 hours.
3. The accused team is notified and given 24 hours to respond.
4. Decision communicated within 72 hours of the complaint. Decision is final.
5. If disqualification occurs, the next-ranked team advances.

12.10 How Results Build Chapter Reputation#

The AI World Cup is not just a competition - it is a data source about chapter health and capability.

Chapter Standing metric is calculated from#

• Qualifier participation rate (% of members who competed)
• Qualifier advancement rate (% of teams that advanced to regionals)
• Regional and global performance of chapter teams
• Event quality score (from Central’s event report review)

Standings are#

• Published annually on theaifellowship.com/chapters
• Referenced when allocating hosting rights for regional and global events
• Used by Central to identify chapters that need support vs. chapters that should be mentoring others

High-standing chapters get first consideration for hosting regional semifinals. This is a meaningful incentive - hosting a regional event is a major community and visibility opportunity.

12.11 Inspiration and Precedents#

The AI World Cup draws on a strong lineage of community-building through structured competition:

ImageNet Large Scale Visual Recognition Challenge (ILSVRC)#

Showed that a well-designed annual benchmark, openly available, with a clear leaderboard, could organize a field’s energy and produce breakthrough results. The Fellowship is not trying to advance academic research - but the principle of “a shared challenge creates shared progress” applies directly.

Kaggle#

Demonstrated that a platform for data science competition could build one of the most active technical communities in the world, with genuine skill-building and career impact for participants. The World Cup borrows Kaggle’s spirit: competition as education and community, not just ego.

RoboCup#

An inspiring model for long-horizon community mission. The stated goal (beat the human World Cup team by 2050) is specific enough to organize energy, ambitious enough to inspire, and long enough to sustain a multi-generational community. The AI World Cup doesn’t need a 30-year goal - but it borrows RoboCup’s lesson that the format itself (the “cup” structure with local qualifiers and a global final) creates pride, identity, and continuity.