AfricAIED 2024 Hackathon

Challenge: Build an open-source, AI-powered tool that enables students to prepare for the National Science and Maths Quiz (NSMQ) using any of or a combination of the following open-source outputs from the Brilla AI project: dataset, code, and models

Background
The National Science and Math Quiz (NSMQ) is an annual live science and mathematics competition for senior high school students in Ghana in which 3 teams of 2 students compete by answering questions across biology, chemistry, physics, and math in 5 rounds over 5 progressive stages until a winning team is crowned for that year. 

Brilla AI is an open-source project by Kwame AI Inc. that is building an AI to compete live and win the National Science and Math Quiz.  The Brilla AI team built and deployed the first AI Contestant in the 30-year history of the NSMQ in October 2023. This hackathon seeks to build upon the Brilla AI project to make real-world impact in education across Africa. 

AfricAIED is an annual 1-day workshop that brings together educators, researchers, entrepreneurs, policymakers, and AI experts to discuss and collaborate on innovative ideas, best practices, and future developments in leveraging AI to enhance learning experiences and outcomes in Africa.

Motivation
There is a lot of inequity in preparations for the NSMQ where resources such as study materials, great teachers, etc, are available only to the big-name high schools that then almost always make it to the semi-finals and finals, and win. This problem is quite representative of the inequity in Ghana’s educational system. This challenge aims to crowdsource open-source AI-powered tools to democratize preparation for the NSMQ and give all schools a fairer chance at winning the NSMQ. More broadly, the tools could be extended to improve science and math learning across Africa.

Challenge
Build an open-source, AI-powered tool that enables students to prepare for the NSMQ using any of or a combination of the following open-source outputs from the Brilla AI project: 

(1) dataset consisting of past quiz questions and answers, and open-source textbooks
(2) code,
(3) models for speech-to-text (STT) — transcribes speech with Ghanaian accent, text-to-speech (TTS) — speaks with a Ghanaian accent (we have a model for the NSMQ Quiz Mistress’ voice), and question-answering (QA). 

NOTE: You are allowed to use other open-source code, models, and datasets additionally. You are NOT allowed to use commercial models such as GPT-3.5.

Ideas:

• Audio-based assessment system where a question is asked in the Quiz Mistress’ voice, accepts audio answers, and says whether it’s correct or wrong along with an explanation. Examples include:

  • Simulates an assessment for one riddle question in round 5 for one student. The tool reads the clues of a riddle, a student indicates they are ready to answer, the student says the answer, the system transcribes and compares the answer to the correct one, and says whether it’s correct or wrong. If the answer is correct, it determines and says the obtained score based on the clue on which they answered. If it is wrong, it continues reading the clues.

  • Simulates the whole Riddle round (4 questions)

  • Simulates multiple rounds

• Audio-based question and answering system where you ask a question and get an answer and explanations 

• Multiplayer quiz assessments and tracks metrics on how quickly questions were answered, the number of wrong answers, etc.

• Taking tests on specific topics, from specific years, specific rounds, etc.

• Track the performance of other teams determined by AI such as how quickly they answered, the number of wrong answers etc.

Important Dates

• 21st May 2024: Competition start

• 17th June, 2024, 11:59 PM: Submission Deadline

• 1st July, 2024: Announcement of Winners

• 8th July 2, 2024 - Presentations by the 3 winning teams at AfricAIED 2024

Registration

• Teams should be a minimum of 2 people and a maximum of 5 people

• Register here to take part and agree to the competition's official rules

• Registered participants will be added to a Slack group where they will receive technical support from the Brilla AI team 

Technical Requirements

Below are the technical requirements for the hackathon. Please make sure to read the official rules here. Failure to comply with them will result in disqualification and potential legal action (for the requirements related to the ethical use of the AI voice models).

• The Google account of the team lead will be given access to the (1) NSMQ Dataset and (2) Brilla AI Voice Models

• The NSMQ Dataset consisting of past quiz questions and answers CANNOT be redistributed or publicly shared and should be deleted after the period of the hackathon

• The Brilla AI Voice models consist of 2 text-to-speech models that have been trained to “speak” in the voice of Isaac Sesi and Professor Elsie Kaufmann to enable participants to use them for example as a quiz master/mistress, competing student, or AI contestant for their tools.  For ethical reasons, you are NOT allowed to use any of these voice models outside the context of generating audio for education for use in this hackathon. You are NOT allowed to redistribute or publicly share the  models and they should be deleted after the period of the hackathon

• The tool should help students with NSMQ preparation

• The tool should make use of AI

• The tool should make use of at least ONE of the open-source outputs from the Brilla AI project:

  • (1) dataset consisting of past quiz questions and answers, or open-source textbooks 

  • (2) code, 

  • (3) models for speech-to-text (STT) — transcribes speech with a Ghanaian accent, text-to-speech (TTS) — speaks with a Ghanaian accent (we have a model for the NSMQ Quiz Mistress’ voice), and question-answering (QA).

• The tool should use only open-source software or models. You are NOT allowed to use commercial models such as GPT-3.5, Claude, Gemini, etc.

• The tool should have a quality higher than that of the demo tool that we released for inspire ideas (Video here and Repo Here) to be eligible for any of the prizes

• All code and files submitted for evaluation should be released on GitHub with an  Apache 2.0 license, which is a permissive, open-source license so it is compatible with the Brilla AI open-source project

Submission

Submit the following in the submission form:

(1) link to the project’s GitHub repo which contains a README with all necessary details to setup and run the tool, and an Apache 2.0 license 

(2) 3-minute video explaining and demoing your system, and 

(3) technical report (max 3 pages) using this template describing the 

(a) Problem: specific problem regarding NSMQ preparation that your system addresses, 

(b) Solution: your solution along with details of the architecture of your technical system, and various design decisions made and 

(c) Impact: potential impact of your solution regarding improving education first for NSMQ preparation and STEM education broadly

Prizes

The top 3 submissions will win cash prizes and an opportunity to present their solutions at AfricAIED 2024 on 8th July in Accra, Ghana.

1st Place: $1,000
2nd Place: $500

3rd Place: $250

Judging Criteria

1. Impact - How likely is your tool to make a difference in education?

2. Innovativeness - How innovative and novel is your tool?

3. Accessibility - How easily can your tool be used by the target population? For example, can it be run on a computer without the internet?

Technical Resources

1. 1. Brilla AI Repo: https://github.com/brilla-ai/brilla-ai

   2. NSMQ Prep AI Demo:

      • Video: https://drive.google.com/file/d/1F8QVQ2fglujNiaq7G4huUedfSkrA-tSG/view?usp=sharing

      • Repo: https://github.com/brilla-ai/brilla-ai/tree/main/examples/riddle-prep-buddy-v2

   3. Speech-to-Text Starter Notebook: https://github.com/brilla-ai/brilla-ai/blob/main/speech-to-text/STT_Starter_notebook.ipynb

   4. Text-to-Speech Starter Notebook: https://github.com/brilla-ai/brilla-ai/blob/main/text-to-speech/TTS_Starter_Notebook.ipynb

   5. Question Answering Starter Notebook: https://github.com/brilla-ai/brilla-ai/blob/main/question-answering/QAML_Starter_Notebook.ipynb