AI Innovation Challenge

If you have any questions about the submission process, please email shehroze.saharan@georgebrown.ca and aiinnovation@georgebrown.ca.

Strong but unselected proposals may receive feedback and encouragement to resubmit in future challenges. 

George Brown uses an enterprise-grade version of ChatGPT with advanced security and privacy controls. Unlike standard ChatGPT, this version:

• Ensures that no data or prompts entered into the system are used by OpenAI to train their models.
• Provides additional protections through an enterprise agreement between George Brown and OpenAI, offering safeguards tailored to institutional needs.

George Brown has implemented strict guidelines to prevent personal data from being entered into ChatGPT Edu unless absolutely necessary and approved. Any projects that involve personal data will undergo careful review to ensure compliance with privacy laws and institutional policies.

No, George Brown is not implementing ChatGPT polytechnic-wide, nor is AI being used to replace jobs. Yes, we recently signed an agreement with OpenAI. Our agreement with OpenAI is limited to a small number of licenses for learning and experimentation. This initiative is about understanding AI’s potential and learning what works and what doesn’t work. George Brown is taking a measured, responsible approach to AI. Our goal is to learn, adapt, and ensure that AI is used in ways that benefit faculty, staff and students.  It is not as a means for workforce reductions and is not related to the current mitigations.

Building our own LLM is impractical and cost-prohibitive for an institution like George Brown. Training a competitive AI model requires massive computational resources, specialized AI expertise, and ongoing maintenance—far beyond what most organizations outside of major tech companies and research institutions can support.

For context, training a state-of-the-art LLM requires:

• Hundreds of high-performance GPUs, which are expensive and in limited supply.
• Millions of dollars in computing power, with top models costing hundreds of millions to train.
• Continuous updates and retraining to stay relevant, requiring a dedicated AI research team.

Even if we were to train a smaller, open-source AI model, we would still need U.S.-made hardware (e.g., NVIDIA chips) and cloud services (e.g., Microsoft, Google, or AWS)—making full independence unrealistic.

Instead, by leveraging existing enterprise AI solutions, we gain access to cutting-edge technology at a fraction of the cost and with built-in security and compliance features, allowing us to focus on responsible and practical AI adoption at George Brown.

George Brown is exploring AI solutions from multiple sources, including Canadian companies and other global providers. OpenAI was the first viable enterprise-grade solution, but we are not exclusively committed to it and will continue evaluating Canadian and international alternatives. That said, avoiding U.S.-based technology entirely is nearly impossible. The leading AI infrastructure—hardware, cloud computing, and software—comes from U.S. firms like NVIDIA, Microsoft, and Google. Even training our own AI would rely on American-made processors and platforms. By collaborating with OpenAI, George Brown gains early insights into AI’s role in education while keeping flexibility to adapt as technology evolves.

It’s true that training large AI models requires significant energy, especially during the initial development phase. However, it’s important to note that George Brown is not training its own AI models. We are using pre-trained models (like ChatGPT), which require much less energy to operate once deployed.

Also, the efficiency of AI technology is rapidly improving. Advances in hardware, such as more energy-efficient GPUs and AI accelerators, along with smarter software optimizations, are significantly reducing the power required to run AI systems. For example, OpenAI and other leading AI organizations are now prioritizing green computing practices and cloud infrastructure powered by renewable energy. A recent study by Epoch found that energy use per AI operation is dropping dramatically over time as technology evolves.
URL:  https://epoch.ai/gradient-updates/how-much-energy-does-chatgpt-use

At the same time, AI can actually support George Brown’s sustainability goals by helping us work more efficiently. For example:

• Automating paper-based processes to reduce waste.
• Optimizing space, energy, and resource usage through smart scheduling and planning tools.
• Analyzing environmental data to help drive better sustainability decisions.

In short, while AI does have an environmental footprint, it also has the potential to reduce George Brown’s overall environmental impact when applied thoughtfully. That’s why our approach is focused on learning and experimentation—so we can understand how to use AI in ways that are both effective and responsible.