OpenAI launches Asian operations in Tokyo to avoid being lost in translation

Local customers get early access to Japanese-language GPT-4

On Sunday, OpenAI announced the launch of its operations in Asia, beginning with a office in Tokyo, Japan. It's OpenAI's third outpost beyond the United States, following offices in London and Dublin.

"We chose Tokyo as our first Asian office for its global leadership in technology, culture of service, and a community that embraces innovation," declared a statement penned by CEO Sam Altman.

The ChatGPT-maker is "committed to collaborating with the Japanese government, local businesses, and research institutions to develop safe AI tools."

At the helm of the Tokyo office is Tadao Nagasaki, who was appointed president of OpenAI Japan.

Nagasaki left AWS last month after over 12 and a half years. One of his goals while at AWS was to make generative AI more accessible and useful for Japanese customers.

The newly appointed OpenAI president wrote that he is "[looking] forward to contributing to the local ecosystem, while exploring how AI can help with Japan's societal challenge."

Altman revealed that as a first step in the region OpenAI is providing local businesses with early access to a version of GPT-4 optimized for the Japanese language.

"This custom model offers improved performance in translating and summarizing Japanese text, is cost effective, and operates up to 3x faster than its predecessor," claimed the CEO.

He later named Daikin, Rakuten, and Toyota Connected as customers already using ChatGPT Enterprise in their businesses, and that will therefore benefit from a local OpenAI presence. He also detailed an effort to spread ChatGPT use into 21 local governments.

The emergence of decent quality non-English generative AI has become a priority for many nations and developers. Sovereign AI is seen as necessary to ensure data security and reduce dependencies on offshore providers, and to improve domestic productivity.

South Korean web giant Naver noted recently that English and North American cultures "are extremely overrepresented in the pre-training corpora" for existing mainstream large language models (LLMs).

Naver has developed its own family of LLMs named HyperCLOVA X, which it claimed performs better at cross-lingual reasoning in Asian languages than other models. The Korean web giant also claimed the models displayed competent "multilinguality" – the ability to work in languages other than those it was trained to handle.

The Indian government has also flagged its intention to spin up an indigenous LLM. Last month it pledged $1.24 billion to efforts which – in addition to funding a supercomputer with at least 10,000 GPUs – will see the opening of an AI innovation center tasked with foundational model development.

Japan is not the only Asian country OpenAI wants to work woth. Last September, Altman told an Indonesian audience he would "like GPT5 to be very good at even smaller languages and dialects" and expressed willingness to incorporate Indonesian dialects into OpenAI's models.

In exchange for his achievements, Altman was granted the country's first ever gold visa – which entitles him to reside in Indonesia for up to ten years, should he so choose. ®

More about


Send us news

Other stories you might like