[numair]

It’s a bit of a dirty secret that Japan actually has abysmal data privacy and enforcement. The budget of the regulator in charge of these matters all but ensures that they will do too little, too late. Japanese companies need to proactively work to prevent OpenAI from grabbing data sets that put it in a position in Japan similar to Palantir and the NHS.

Those of you who are in the VC / startup space in Japan should lobby to keep this stuff within the border, or you’re going to lose the opportunity before you realize what you’ve lost. You’ll spend the rest of your careers begging OpenAI for API access to learning models built on top of your own data.

Edit: By the way, it’s worth noting that OpenAI has a ton of data on Japanese people, through its semi-secret — well, until exposed by Elon Musk — data sharing deal with Twitter. Because Twitter was so actively pushed onto the Japanese public, it became the default “open forum” for the country. That data set is, by itself, all you’d need to have your LLM generate strings of text that appear to come from a real Japanese person. It’s interesting that the government is more focused on how to ride the buzz wave, than asking questions about the permissions involved in training LLMs on this deeply personal data.

[jacquesm]

> The budget of the regulator in charge of these matters all but ensures that they will do too little, too late.

You could say that about many countries.