ChatGPT on fire as Africa conference explores AI's challenges and opportunities for poor countries

May 7 - More than 2,000 researchers and engineers from around the world gathered in Rwanda this week to debate different visions for the future of artificial intelligence, amid growing concern about its promise and perils.

ChatGPT ignites fervor as African conference explores AI's challenges and opportunities for poor countries

One vision is to build more powerful systems like ChatGPT that aim to surpass human intelligence to improve worker productivity and boost economic growth. Another vision is to create smaller, more targeted AI solutions to address global challenges like climate change, improving healthcare and protecting biodiversity.

This is the first major AI research conference to be held in Africa this year, and one of the highlights of the conference is the increased cross-fertilization of ideas. Organizers hope that researchers from the U.S., China and wealthy companies will become more aware of the societal issues facing Africa, while allowing African researchers to participate in the discussions and provide their unique insights.

Yoshua Bengio, known as one of the "godfathers of AI" and a member of the conference organizers' board of directors, advocated for bringing AI technology to Africa: "It's clear that in order for everyone to enjoy the potential benefits of AI, we need everyone to be involved. need everyone to be involved."

The conference comes just months after ChatGPT, a chatbot, began to catch fire, adding urgency to the discussion of AI's trajectory and its short- and long-term implications. the release of ChatGPT set off a global generative AI (AIGC) frenzy, prompting major tech companies such as Google and Baidu to scramble to launch their own similar products, software that can generate text and images that sparking fears of job displacement and the unchecked spread of misinformation.

OpenAI, the developer of ChatGPT, was not present at the AI research conference in Kigali, Rwanda, but it was still the center of attention. Will it be embraced or resisted for the direction it is taking? The global AI community is polarized in its views.

Recently, Geoffrey Hinton, another "godfather of AI," announced that he was leaving Google to speak more freely about the risks of AI development. In a media interview, Hinton said he was concerned that the technology posed a threat to the long-term survival of humanity.

At the conference, many researchers from Africa and other developing countries said they were instead more worried about AI posing a direct challenge to their societies. They fear that current trends in the AIGC model could exacerbate the dominance of the U.S. and China in AI development, which would create two problems: either Africa would be left behind or would have to deal with the problems it poses. Such consequences include facing more disinformation in African elections and the gradual disappearance of their native languages in digital technology.

Many researchers are also concerned that the development of beneficial AI solutions, which could help improve the basic quality of life for people around the world, will lag behind.

Girmaw Abebe Tadesse, an Ethiopian researcher in Microsoft's Nairobi, Kenya, office, highlighted key data issues, such as filling out the wrong medical forms, that have hindered the progress of AI in developing countries and have yet to help improve maternal health and reduce child mortality.

Tadesse also shared some success stories, such as using a combination of high-quality data and statistical analysis to find that child mortality in southern Nigeria is lower than the country's average, which has allowed researchers to target their investigations more closely, identify causes and design solutions for other locations.

Other participants also presented applications such as using satellite imagery to understand ethnic disparities in access to parks and sanitation facilities, using sensors and mathematical models to improve grid maintenance, and using computer vision to detect agricultural diseases.

Bengeo said he hopes this year's conference will allow researchers to focus more on artificial intelligence for social good applications rather than simply pursuing profits. "We talk a lot about AI risks, but we don't talk enough about how to make better use of AI," he said.

Conference organizers said there were 261 attendees from Africa since the conference, compared to 16 in 2019. For years, the largest and most prestigious annual AI research conferences have typically been held in the United States or Canada because of their proximity to Silicon Valley. Due to visa issues, African researchers are often unable to attend the conference due to difficulties in obtaining visas. This results in a lack of African perspectives in the development of one of the most powerful and transformative technologies.

Prominent researchers, including artificial intelligence ethicist Timnit Gebru, point out that research tends to focus on a few dominant players in Silicon Valley and lacks researchers who are not Western or from marginalized groups.

Gebru, who grew up in Ethiopia and later came to the U.S. as a refugee, founded the nonprofit Black in AI in 2017 to bring more diversity to the artificial intelligence community.

Gebru moderated a panel discussion last Friday on the limitations of large language models in dealing with African languages. These models are the underlying technology used to build ChatGPT, but African languages are increasingly excluded from the digital world due to a lack of data needed for current AI techniques. Gebru said she co-authored a paper criticizing the explosive growth in demand for resources and the environmental impact of such models, which also underpin Google's search engine. Previously, Gerbru was fired from Google, causing public concern. Within Google, Google characterized her departure as a resignation.

In another panel discussion, Vukosi Marivate, a professor of data science at the University of Pretoria in South Africa and chair of the conference program, said African researchers are struggling with the threat of losing their native languages. "We are in a race against time" before English dominates and African languages cease to exist, he said after the event.

Despite the fact that Kigali is thousands of kilometers from Silicon Valley, the lure of large language models and AIGC tools such as ChatGPT could not be escaped. Researchers filled the lecture halls for related conferences, sitting on the floor, leaning against the walls, and others rushing straight from the airport to sit on their suitcases.

Many researchers said they were surprised by OpenAI's findings, prompting them to actively expand their models as well. One doctoral student in attendance said he plans to shift his study focus to AIGC, joining the most exciting research directions. "This will be the last conference that is not dominated by large language modeling studies," he said.