Asean Business logo
SPONSORED BYUOB logo

ChatGPT needs more time to localise for different countries and cultures: OpenAI CEO

Jamille Tran

Published Fri, Dec 8, 2023 · 12:00 PM
    • OpenAI CEO Sam Altman spoke via video link at AI Day 2023 in Ho Chi Minh City earlier this week.
    • OpenAI CEO Sam Altman spoke via video link at AI Day 2023 in Ho Chi Minh City earlier this week. PHOTO: VINAI

    [HO CHI MINH CITY] Sam Altman, the chief executive officer of US artificial intelligence giant OpenAI, said it will take “a little while” for ChatGPT to adapt to the diverse needs of different countries, cultures and languages.

    Speaking via a video link from a conference in Vietnam, he called for patience as the company goes through what he described as “unnaturally fast growth”.

    “We just went through one of the craziest periods for any tech companies in history,” said Altman at the AI Day conference in Ho Chi Minh City. “We didn’t have the time that other companies normally have, like five or 10 years, to go and build all of that out.”

    He pointed out that OpenAI went from a small research lab to become one of the world’s most valuable technology companies in the space of just a year.

    This meant that there was hardly any time to make its ChatGPT tool accessible for everyone, especially as there was a need to ensure compliance with the local laws of the countries where the AI chatbot service is provided.

    “There will be some limits, there will be some countries that just don’t allow us to operate,” said Altman, who is now back at the helm at OpenAI after a sudden ouster and reinstatement.

    A NEWSLETTER FOR YOU

    Friday, 8.30 am

    Asean Business

    Business insights centering on South-east Asia's fast-growing economies.

    “We still have a lot of work to do, (for example) to figure out how to make future versions of ChatGPT work much better in non-English languages,” he added.

    AI Day is an annual event that is organised by VinAI, the AI research arm of Vietnam’s largest private conglomerate Vingroup. The two-day conference was attended by thousands of global AI experts and industry players.

    Recently, VinAI open-sourced its self-developed generative model series called PhoGPT, a 7.5 billion parameter large language model. Named after the traditional pho noodle dish and akin to OpenAI’s ChatGPT, PhoGPT is pre-trained specifically for the Vietnamese language and culture.

    In a speech at the event, Vietnam’s Deputy Minister of Planning and Investment Tran Duy Dong said he has confidence that Vietnam holds “substantial potential” in AI and the capacity to keep pace with global innovation in the sector.

    At one of the panel discussions, an executive from Amazon Web Services (AWS) – a subsidiary of Amazon that provides on-demand cloud computing platforms – said the company’s Generative AI (GenAI) startup accelerator programme received the highest number of applicants from Singapore, Vietnam and the Philippines earlier this year.

    “We are starting to see the emergence of a group of GenAI-first startups that are building for a global audience while being based in South-east Asia,” said Kang Kai Yong, the startup business development manager at AWS.

    He said this underlines the fact that the region offers a cost advantage to companies with a strong pool of engineering talent and a young population that’s willing to develop and adopt innovative AI products.

    There are opportunities for localisation and “a race to proprietary data”, said Kang, with more startups looking to leverage data sources to provide tailored AI solutions for a specific country or enterprise use.

    One example is how US-based AI software startup Landing AI entered into a partnership last year with FPT Software, the largest IT services company in Vietnam. The two have worked together to implement computer vision solutions for clients in industries such as manufacturing and automotive.

    “You can’t go out and get proprietary data off the Internet,” said Landing AI chief operating officer Dan Maloney. “You need a (local) partner to line up with (to access proprietary data sources), and to train domain-specific high-value large vision models.”

    Copyright SPH Media. All rights reserved.