Project Bringing Taiwan's Languages Into the AI Age

A tech consortium is teaching chatbots Taiwan's local languages, in the hopes they won't be left behind in the AI age.

Transcript

00:00ChatGPT may have a solid grasp of Mandarin Chinese,

00:04but put in a query in Taiwan's traditional characters,

00:07and you may well get a response in the simplified characters China uses to write instead.

00:12And if you want to use other local languages with chat,

00:15well, you're out of luck.

00:17It just hasn't been trained much beyond the basics.

00:20But that might not be the case for much longer.

00:23One tech consortium, the Information Management Association,

00:26has launched a project called Taiwan Tongues,

00:29designed to make sure Taiwan's languages stay caught up with the AI boom,

00:33including one of the most widely spoken, Taiwanese Taiyi.

00:47Getting corpuses as vast as those used to train chatbots in English or Mandarin might be difficult,

00:52but the project has made a start.

00:55Already, it's fed chatbots over five million characters of input,

00:58in part, it's able to do so through partnerships with authors who write in these languages,

01:03and who waive copyright, so their tongue has its own place in the expanding AI universe.

01:08This is not only a role, but also a role,

01:12the government is excited about this project.

01:24The government is excited about this project.

01:27If all goes to plan, chatbots will soon have an autonomous, open and reliable source of data

01:45on how Taiwan's people really talk.

01:47Fu Hua Hong and John Van Trieste for Taiwan Plus.

01:51We made this project.

01:52We gave it a chance to complete our infrastructure.

01:53Let's go.

01:57Are you ready?

01:58The government is planning.

02:00Good job.

02:01We're going.

02:05All right.

02:06Here we have 5 million questions,

02:08we have 4 million views,

02:10and we have got 4 million views.

02:12We have a loved one.

02:13We are really comfortable with theButton.