There was a question and a lot of debate about using AI to just *generate* content in the target language. *No* translation. I *disagree* with a lot of smart people on this. It reminds me of those early self-driving car inventors who dreamed of lobbying the government to change the roads to make driving easier to automate. (Плохому танцору и ______ мешают...) Actually in both roads and content, that already exists -- railroads, "controlled language", templates for generating finance articles and product descriptions... It makes problems so easy automate that you don't even need AI. If you can do that, great, you should! But most content, and most technology, doesn't work like that. The 0.0...1% of content that gets translated with a human touch today is relatively valuable - that's why somebody got it translated even though it's painfully slow and expensive. As a buyer or user, when you want a translation, you generally don't want just some content. You want that article, that video, that contract, that post, that comment, that medical record, that app, that book... It'd be a joke to impose on humanity that everything we create be so bland and structured that it can be trivially generated in a hundred languages. Most translation is just fundamentally downstream of content creation, and always will be. Dealing with that is half the job. C'est la vie. Not too different than transport, medicine or any other technology that has to deal with the world as it is and would hardly be needed anyway if it weren't that way. So I file this idea under what Peter Thiel calls "indefinite optimism". It's not pessimism, but it's not a concrete vision of the future. And, like pessimism, it often becomes an excuse for doing nothing. Or for not succeeding at anything. Which is why it's so popular. Back to work.
There was a question and a lot of debate about using AI to just *generate* content in the target language. *No* translation.
I *disagree* with a lot of smart people on this.
It reminds me of those early self-driving car inventors who dreamed of lobbying the government to change the roads to make driving easier to automate.
(Плохому танцору и ______ мешают...)
Actually in both roads and content, that already exists -- railroads, "controlled language", templates for generating finance articles and product descriptions... It makes problems so easy automate that you don't even need AI. If you can do that, great, you should!
But most content, and most technology, doesn't work like that.
The 0.0...1% of content that gets translated with a human touch today is relatively valuable - that's why somebody got it translated even though it's painfully slow and expensive.
As a buyer or user, when you want a translation, you generally don't want just some content. You want that article, that video, that contract, that post, that comment, that medical record, that app, that book...
It'd be a joke to impose on humanity that everything we create be so bland and structured that it can be trivially generated in a hundred languages.
Most translation is just fundamentally downstream of content creation, and always will be. Dealing with that is half the job. C'est la vie. Not too different than transport, medicine or any other technology that has to deal with the world as it is and would hardly be needed anyway if it weren't that way.
So I file this idea under what Peter Thiel calls "indefinite optimism". It's not pessimism, but it's not a concrete vision of the future.
And, like pessimism, it often becomes an excuse for doing nothing. Or for not succeeding at anything. Which is why it's so popular.
Back to work.