Initially, John Passmore was excited about ChatGPT.
The serial founder has been in the field of artificial intelligence since at least 2008. He remembers a time when experts declared that it would be decades before something like ChatGPT would arrive in the world. Time passed, and that day has now come.
But there is a catch.
ChatGPT, one of the world's most powerful artificial intelligence tools, struggles with cultural nuances. This is especially troubling for Black people like Passmore. In fact, this oversight has angered many Black people who felt they weren't adequately represented in the algorithm that is touted to one day save the world. ChatGPT as it stands provides overgeneralized answers to specific questions that cater to specific communities, and its training appears Eurocentric and Western biased. This is not uncommon: most AI models are not built with people of color in mind. But many Black founders are adamant that they don't want to be left behind.
As Black founders like Passmore seek to capitalize on OpenAI's cultural missteps, a number of Black-owned chatbots and ChatGPT versions focused on Black and brown communities have popped up in the past year.
“If you ask this model generally who the most important artists in our culture are, you're likely to hear Leonardo da Vinci or Michelangelo,” Passmore said of ChatGPT. “Because of its bias toward the trajectory of European history, the model says nothing about India, China, Africa, or even African Americans.”
So Passmore launched Latimer.AI, a language model that provides customized answers to reflect the experiences of Black and Brown people. Erin Reddick launched ChatBlackGPT, a chatbot also focused on Black and Brown communities. Globally, there's Canada-based Spark Plug, which is essentially ChatGPT for Black and Brown students. Africa is also seeing great innovation in this space, with language models emerging that serve the 2,000+ languages and dialects spoken in a continent that Western AI models still overlook.
“We are the custodians of our own stories and experiences,” Spark Plug founder Tamar Huggins told TechCrunch. “We need to build systems and infrastructure that we own and control to ensure our data is ours.”
Personalized AI is here
Generalized AI models cannot easily capture the African American experience because many aspects of their culture are not present online. Current algorithms scrape the internet for sources of information, but many traditions and dialects in African American culture are passed down orally or in person, creating a gap between what an AI model understands about a community and the nuances of what actually happens.
That's one of the reasons why Passmore built Latimer.AI not by training it on user-generated data scraped from the internet, but by focusing on accuracy and using sources like the Amsterdam News, one of the oldest Black newspapers in the U.S. As he did so, he began to notice differences between his model and ChatGPT's.
He recalled that people had once asked ChatGPT about the Underground Railroad, a route enslaved black Americans used to travel to northern states to escape slavery. While ChatGPT's model mentioned runaway slaves, Latimer.AI adjusted the wording to say “slaves” or “freedom seekers,” which is more in line with social adaptations when discussing people who were formerly slaves.
“There are nuances in the language the model uses depending on the training data, and the model itself is only thinking about Black and brown people,” Passmore said.
Meanwhile, Erin Reddick's ChatBlackGPT is still in beta and set to launch on June 19. Her product, as the name suggests, is a chatbot that lets you ask questions about Black culture and receive tailored answers. “At the core of what we do is truly community-driven,” she says.
Image credits: ChatBlackGPT and Stefan Youngblood
She's currently building the tool, asking users how they want it to look and behave. She's also partnering with educational institutions, including historically black colleges and universities (HBCUs), to teach their students and help train the algorithms. “I want to create holistic learning opportunities for black and brown people to safely explore AI,” she said.
“The algorithm prioritizes Black sources and speaks to a body of knowledge that is more immediately relatable than the average experience,” she told TechCrunch, adding that like Passmore's products, technically anyone can use them.
Tamar Huggins also built Spark Plug to provide a more tailored experience for Black and Brown communities. Her platform translates educational materials into African American Vernacular English (AAVE), an ethnic language associated with African American communities. This dialect has traditionally been passed down orally and directly, rather than studied and written down like standard English, meaning that an AI model (or person) that learns it solely from the internet will be less accurate. Capturing AAVE accurately is important not only so that chatbots can use it to respond, but also so that students can more easily write their prompts so the AI can return the results they need.
Image credit: Spark Plug (screenshot)
“We create content that resonates with Black students so they feel engaged in their education, which is crucial for high engagement and academic success,” Huggins says. “When given the opportunity, big tech companies almost always put profits over people, so we've carved out our own path in AI.”
Huggins trained the algorithm using the writings of black authors from the Harlem Renaissance, black authors in education, and even her own teenage daughter's phrasing to capture the essence of AAVE. Huggins also works with educators, linguists, and cultural experts to review and validate Spark Plug's output. Her product is also not built on ChatGPT, which is a proprietary model that means users have control over their own data.
Passmore also plans to build other foundational models for Latimer.AI, and now that more students are using ChatGPT every day to complete their assignments, he's working to expand his company to schools, especially HBCUs.
“This is a better AI companion for a lot of the jobs that are being put to Black and brown kids,” he said.
Diaspora unity
Africa is often overlooked in the current AI movement. For example, according to the 2023 Artificial Intelligence Index report, only 0.77% of all AI journals in the world originate from sub-Saharan Africa, while East Asia and North America account for 47.1% and 11.6%, respectively. In terms of population, Africa accounts for about 17% of the world's population compared to North America, which accounts for only 7%. When it comes to gathering information and experts on AI, sub-Saharan research is highly unlikely to be utilized, which could impact the development of global AI tools.
In Africa, much progress has been made in creating more inclusive language models that are better suited to the Black Diaspora, but as of now, current AI models from ChatGPT to Gemini cannot fully support the 2,000+ languages spoken across Africa.
To address this issue, Yinka Iyinolakan created CDIAL.AI, a chatbot that can speak and understand almost all African languages and dialects, with a specific focus on speech patterns rather than text.
Iyinolakan shared with TechCrunch the same sentiment held by many Black Americans: that the underlying AI models are primarily extracted from internet data and the most commonly spoken languages. Like African American descendant cultures, many African languages and traditions are not present on the internet because Africa is a culture that has historically been passed down orally rather than written. This means that AI models are not well informed about African cultures and are not self-learning enough, leaving knowledge gaps.
Image credit: CDIAL.AI website
At CDIAL.AI, Iyinolakan brought in more than 1,200 native speakers and linguists from across Africa to pool their knowledge and insights to build what he calls “the world's first multilingual speech-first large-scale language model.” The company plans to expand over the next 12 months to add more languages and build models that support text, speech, and images.
He's not alone here: Google recently awarded Kenya-based Jacaranda Health a $1.4 million grant to build out its machine learning services to work with more African languages, and Intron Health recently raised millions to expand its clinical speech recognition to more than 200 accents spoken across Africa.
“Silicon Valley wants to believe that's what artificial intelligence is all about,” Iinorakan said, “but to 'understand' artificial intelligence, which is what every company is aiming for, you need to embrace a third of the world's knowledge.”
Moving forward
The introduction of AI chatbots isn't the only innovation Black founders are trying to tackle.
Steve Jones founded a company called pocstock to create stock images of people of color after decades of a lack of minority representation in stock images, which is one reason why today, when models are asked by users to generate photos of everyone from doctors to pop singers, they primarily get images of white people.
“All platforms and tools should be trained on racially inclusive and culturally accurate data. Otherwise, [perpetuate] “To address the issue of bias currently facing our society as a whole, over the past five years Pockstock has developed a proprietary visual tagging system that collects diversity data and contributes to a database that companies use to train AI models to help create a more inclusive picture,” Jones told TechCrunch.
But some improvements are happening: Jones said he's noticed that the big stock imaging companies that provide information to AI companies are making more of an effort to increase the diversity of their content. Passmore also sees a bright future, saying personalized AI is the future anyway, and the more an AI model interacts with a user, the better it can understand that particular person's wants and needs, “and I think that will eliminate a lot of bias,” he said.
In the future, there may be room for more culturally specific AI models, especially as more Black-owned alternatives emerge — after all, the world is vast and more nuanced, and there’s no point trying to fit it into a single black box.
“I hope that as soon as possible in the next economic boom, we'll see more founders of color developing their own AI platforms and creating new AI jobs,” Jones said. “AI will create trillionaires, and I want to see people of color in those positions, not just as consumers, but as producers.”
This article has been updated to reflect Spark Plug's training content and that Spark Plug and Latimer.AI have their own underlying models.