Oji Udezue has constructed merchandise that made thousands and thousands in Silicon Valley, serving as chief product officer at Typeform and Calendly, product lead at Twitter, and head of product at Atlassian. On Moonshot’s Emerging Tech stage, he returned to the query he says retains drawing him again to Africa: who’s a know-how truly for?
Frontier AI, he argued, is being priced out of attain for many of the continent, and the Enzi Prize is his try to change who can afford to use it. “It’s for the person in your village who’s not educated,” he stated. “Otherwise, we would be the most selfish generation that ever existed.”
His argument rests on a quantity. Running a frontier, “Fable-class” mannequin prices roughly $50 per million tokens right now, by Udezue’s benchmark. The prize goals to carry that down to about $5, and ideally to a a lot decrease quantity. “We say 100x because maybe we’re hedging our bets,” he informed the room.
To pursue that aim, the Enzi Prize Foundation, a Delaware non-profit backed by the Africa Deep Tech Foundation, TechCabal, Future Africa and Talent Nation, is providing a $10 million prize purse as a part of a roughly $14 million initiative.
The goal is to make AI cheaper and fluent in African languages and contexts. According to Udezue, an affordable mannequin that scores effectively on normal checks is of little use if it has by no means heard of a tropical illness or the issues “people in my village know that are not anywhere on the internet”. Entries should first meet a minimal normal — what he referred to as “a master’s in intelligence and a PhD in African context and African languages” — earlier than competing on cost.
Submissions, whether or not cloud programs or fashions that run on a telephone, are ranked on a public, open-source leaderboard, with rotating, sealed questions to stop groups from gaming it. Winners retain their mental property however should publish an open, royalty-free model for use in Africa and the Global South.
The mannequin is the Ansari X-Prize: supply a reward for a desired breakthrough and let competitors attract funding to obtain it. “If I put up $10 for you to chase,” he stated, “people will spend $9 to chase 10.” He famous {that a} $20,000 pilot had already “brought a lot of talented people out of the woodwork”, producing the demos proven earlier that afternoon. “What would $10 million do?”
The purse just isn’t but totally funded. Udezue stated he had spoken with OpenAI and Google that day, and estimated that successful the prize might take three to six years. His closing attraction was much less for cash than for perception: “Let’s make inference part of African innovation.”
