Bill Gates said at the Gates Foundation's "Goalkeepers" event that combining "more generosity" with "smart use of AI" can speed up the fight against inequality.
比爾蓋茲在蓋茲基金會的「守門人」聚會上表示,「增加慷慨」與「聰明運用AI」結合,可加速對抗不平等。
He urged the world's richest countries to expand international aid and use AI for "life-saving innovation."
他敦促全球最富裕國家擴大國際援助,並把AI用於「救命創新」。
He also said in a talk with commentator Van Jones that humans have not yet built the right values.
他也在與評論員范瓊斯的對談中說,人類尚未建立正確價值觀。
He also said humans still do not have a way to make moral rules control AI behavior.
他也說,人類還沒有讓道德準則控制AI行為的方法。
This yearly event usually takes place during the week of the UN high-level meeting.
這場年度活動通常在聯合國高階會議那一週舉辦。
This year's opening noted that 26 of the world's 34 richest countries cut development aid to poor countries last year.
今年開場提醒,全球34個最富裕國家中,有26國去年減少了對貧窮國家的發展援助。
The UN also warned that progress on the Sustainable Development Goals is "slow and uneven." These goals aim to cut poverty, improve health, and protect the environment by 2030.
聯合國也警告,2030年前減少貧窮、改善健康與保護環境的永續發展目標,進展「緩慢且不均衡」。
Gates pointed out that these cuts are happening at a time when we already have some of the best innovations in history.
蓋茲指出,這波縮減發生之際,我們已擁有史上一些最好的創新。
The foundation promoted a new alliance with 60 members, including Anthropic, Google, and the OpenAI Foundation.
基金會宣傳一個新聯盟,成員共60個,包括Anthropic、Google、OpenAI Foundation等。
The alliance works to make AI easier to use in underrepresented languages.
聯盟致力讓AI在代表性不足的語言中更易使用。
The goal is to reach over 3 billion people in five years by coordinating existing efforts.
目標是透過協調現有努力,在五年內觸及超過30億人。
Foundation CEO Suzman said this work must move "at full speed."
基金會執行長蘇茲曼說,這項工作必須「全速前進」。
This is true even as some large AI companies urge the industry to slow down work on advanced models.
即使部分大型AI公司正敦促業界放慢先進模型的開發。
This partnership comes right after last week's "Gatekeeper" report.
這項合作緊接在上週的《守門人》報告之後。
The report promised to invest 1 billion US dollars in key AI work.
報告承諾投入10億美元於AI重點工作。
The report also warned about a lack of representative language data. It could make a model mistranslate a pregnant Malawian woman's "my water broke" as "I threw away the water."
報告也警告,缺乏代表性的語言資料,可能讓模型把馬拉威孕婦說的「破水了」誤譯為她「把水丟掉了」。
Mozilla Data Collective CEO Luis-Chung said the "original sin" is that many AI tools are trained on language data scraped from the internet.
Mozilla資料集體執行長路易斯-鍾說,「原罪」在於許多AI工具是用從網路抓取的語言資料訓練而成。
Suzman said details such as the alliance's governance are still being finalized.
蘇茲曼表示,聯盟治理等細節仍在敲定。