00:00In the U.S., these Taiwanese expats are recording Taiwanese tai-gi phrases, one line at a time.
00:06The goal is to build a dataset for localized artificial intelligence.
00:11Most AI language models are trained on languages with large populations of speakers, like English or Mandarin.
00:18Researchers say languages from Taiwan, including Taiwanese tai-gi, are often left out.
00:24Now, one project is trying to close that gap.
00:30Tai-gi is a very familiar, veryζζηζθ¦.
00:33It's not like their language and language.
00:34It's a lot of information on Taiwan, but it's a lot of information on Tai-gi.
00:39The project is called Tai-gi Speech, and it focuses on teaching AI to recognize spoken Tai-gi.
00:46The issue is also practical.
00:48For Taiwanese communities overseas, some rely on Tai-gi, not Mandarin, for their daily life.
00:54Because we're using AI, we understand the process of understanding.
01:00This is like a family, they're only 8 tai-gi to a young person.
01:03If you use a language, they don't understand.
01:06We don't have to say anything.
01:08So we can do this for many people who use Tai-gi or a language.
01:13We can use it for many people.
01:14So this is a lot of support.
01:17Researchers say solving this problem could have a wider impact.
01:20They're testing whether AI can learn a language with limited data.
01:24In America, I can't be able to understand the same thing.
01:28The process of understanding is not the same.
01:30I can have this opportunity to tell people that Taiwan has a different language.
01:36I hope that Taiwan has a different approach.
01:39If we can use the best of the information to help our people with our people with our people,
01:43we can use the same tools to help our people with our people with our people with our people.
01:47It comes amid Taiwan's push for sovereign AI, plans to train models in traditional Chinese characters
01:54and Taiwan's local and indigenous languages, reducing reliance on foreign data sets.
02:00Some researchers say this could help prevent outside AI tools from spreading misinformation.
02:05It is absolutely important that we have our own AI model, just in case that more cognitive warfare
02:15or information manipulation can compare with through all this usage of AI to Taiwanese society.
02:25The Tai-gi speech project is backed by teams in Taiwan and the U.S.
02:29And researchers say the aim is to make sure that Taiwanese Tai-gi speakers are not left out of the
02:35AI revolution.
02:37As that technology grows, the language it includes or overlooks will shape who keeps up.
02:44Ryan Wu and Lily LaMantina for Taiwan Plus.
Comments