Update on Feb 20, 2021

Google also fired Margaret Mitchell, who worked closely with Timnit Gebru, on Feb 19, 2021. Apparently the story will still go on.

I find it more and more scientists work are being affected by corporate  and politics. After all we all need resources to do our work, and not all have the vision to revolutionalise the works that are necessary but hard to see results immediately.

Original post:

I read an interesting piece of article about AI ethics, which talks about a research paper that Timnit Gebru published. I think there are some very insightful findings of a particular AI training model called "Large Language Models" and is part of natural-language processing. The paper is the reason why she was being forced to depart Google.

The paper talks about drawback of the Large Language Models, in which because of the way it works (by analysing a lot of web pages, articles, etc) on the Internet, it can be skewed easily, as quoted in the article:

The result is that AI-generated language will be homogenized, reflecting the practices of the richest countries and communities.

And on the other hand, the AI generated language is being too good, that is it is hard to distinguish it from a normal piece of writing written by any average person.

I think that by combining the effects above, it is very easy for a large population nation, with a non-English language basis, could in theory easily manipulate and cloud people's judgement in the Internet via automated translation tools.

Sounds familiar...