So the title should read that the AI companies are gouging some of their customers because they can.
But that’s not what the underlying study shows. It found that LLMs (including the ones you can self-host, not just the premium models) require more tokens to answer the same prompt in foreign languages versus in English. This is almost certainly a consequence of English being over-represented in the training data (and to a lesser extent, tokenizers that were designed with English in mind). It’s not some intentional and nefarious price gouge.
Don’t get me wrong, I know for a fact that the frontier labs are engaged in all sorts of nefarious anti-consumer shit — but this ain’t that.
But that’s not what the underlying study shows. It found that LLMs (including the ones you can self-host, not just the premium models) require more tokens to answer the same prompt in foreign languages versus in English. This is almost certainly a consequence of English being over-represented in the training data (and to a lesser extent, tokenizers that were designed with English in mind). It’s not some intentional and nefarious price gouge.
Don’t get me wrong, I know for a fact that the frontier labs are engaged in all sorts of nefarious anti-consumer shit — but this ain’t that.