Thank you Emmanuel for your comment!

I agree with you. If GPT-'s training cost around $12 million, Wu Dao 2.0 was probably on the same order of magnitude (even taking into account MoE-based cost improvements).

But they didn't publish anything regarding costs. I didn't even find it in the original Chinese article. I talk about the cost problems in this and other articles.

Cheers :)

Writing at the intersection of AI, philosophy, and the cognitive sciences | alber.romgar@gmail.com | Get my articles for free: https://mindsoftomorrow.ck.page