Aurora50, the GCC’s Diversity, Equity, and Inclusion impact agency, leads companies toward fairer workplaces and strives to be recognized not only as real-life diversity champions but also as major influencers in the virtual world, advocating for an equitable future.

In response, the campaign plays a crucial role in promoting gender equity in the virtual world of AI. By addressing biases in AI, it works towards a more balanced and inclusive representation in the digital world.

Fixing the bAIs uses AI to fix AI.

The campaign created a vast database of diverse and inclusive images, featuring women from various races, ages, body types, and professions, ensuring everyone is seen and valued in the digital world.

The primary objectives of the campaign were to actively counter gender bias in AI, foster awareness about the issue, and reshape societal perceptions to promote gender-equality.

It's not just a campaign; it's a catalyst for change.

We used AI to fix AI itself.

We turned the tables on AI using its own tools—Midjourney, DALL-E2, and Stable Diffusion. Crafting over 80,000 inclusive and diverse images of women, we injected diversity directly into AI's veins. Our website, fixingthebias.com, became the hub, disseminating these game-changing images to AI training datasets. We didn't just address a problem but created a lifelong solution. With a global army of supporters, we birthed a perpetual community, a lifelong generator of images on our discord channel, dismantling AI bias, one prompt at a time. This isn't just fixing; it's a revolution in every pixel.

Our PR strategy was laser-focused, launching on International Women's Day to hook onto organic conversations around gender equality and leveraging wider noise around the impact of AI and its impact on society.

We developed a range of materials to reach 3rd party advocates from written content and opinion articles to visual and video to use across earned, owned and shared digital platforms.

This initial burst of activity helped garner interest in our participation in the Committee on Artificial Intelligence (CAI) Plenary meeting, linked to the EU's AI Act.

The campaign's core idea, reshaping AI perceptions with inclusive images, gained continued media attention and drive conversation, creating a positive narrative on gender equality.

Once we had ignited the debates we then developed a stakeholder target list including KAY FIRTH-BUTTERFIELD, the World's First Chief AI Ethics Officer, and JANA NOVOHRADSKA, UNESCO Representative and other policymakers to join our movement.

The project launched on International Women's Day, 8th of March 2023.

Using AI image-generation tools, we built an extensive collection of diverse and inclusive images featuring women in various professions. We removed gender from their metadata (file names, tags, and descriptions) and added this powerful database to our website fixingthebias.com, catapulting these images to stock image banks, social media, and AI training datasets. Additionally, we created a lifelong Discord community that continuously uses Midjourney and generates images to fix the bias, one profession at a time.

We have created more than 80,000 images and trained 9 datasets, including LAION-5B (world's largest dataset of 5.85 billion image-text pairs, the biggest openly accessible image-text dataset in the world and the primary dataset for AI image-generating tools such as Midjourney, DALL·E 2 and Stable Diffusion). Major stock image companies contribute by incorporating our images for free download.

Fixing the bAIs is being pointed out as a lasting solution to fix gender bias in AI at world summits and councils, such as the Council of Europe and CAI. Acknowledged by AI scientists at Ceimia and Mila (largest concentration of deep learning academic researchers globally), the project has inspired white papers affirming its significance. Its impact extends to educational institutions worldwide, showcased at schools and universities. The project contributed to passing of first regulation on Artificial Intelligence, the EU's AI Act, which deems bias in datasets unacceptable, adopted by 52 countries at the EU parliament. Nine datasets, including Liaon 5B (the world's biggest openly accessible image-text dataset) have undergone training. We also set a World Record for the most pledges received for a gender equality in the workplace campaign in 24hours. Campaign reached 4.4 billion people and earned $4.5 million in media, campaign has gained substantial impact and recognition.